The semiconductor industry continues to forge ahead with new facility plans and partnerships. Just last week, OpenAI, the company behind the breakout generative artificial intelligence (AI) chatbot ChatGPT, announced its intention to collaborate with Japan’s top semiconductor minds. This new partnership aims to avoid a possible AI chip capacity challenge from the overwhelming demand.
Brad Lightcap, COO of OpenAI, says Nvidia is the leading force behind the high-performance graphics processing units (GPUs) essential for generative AI. Nvidia currently holds 80% of the market, which is already tight, and Lightcap says it will be crucial that the industry avoids a shortage.
“Our priority is making sure that we don’t end up in a world where there’s so much demand for AI, but we don’t have enough capacity,” said Lightcap. Given how robust demand is now, with forecasts predicting it will extend for several years, we might have already reached that point with some AI components.
AI Demand Contributing to Shortage Conditions
In a recent letter to customers, memory giant Western Digital confirmed that there is a shortage of large-capacity HDD and SSD products. The overall demand from the AI market has caused HDD prices to surge, leading to ongoing adjustments for NAND-Flash and hard drive products.
Due to AI's requirements for data storage and memory suppliers' strategic production cuts over 2023, HDD prices have skyrocketed. Industry sources say that from 3Q23 to 1Q24, HDD prices rose by 10% to 20%.
While robust, this jump is smaller than SSD products over the same period. This results from SSD’s growing popularity over HDD in mainstream consumer PC markets for storage devices below 2TB. SSDs have access speeds that surpass HDD by almost ten times, allowing SSDs to narrow the cost gap when NAND-Flash saw declines.
TrendForce reports that there will be a substantial increase in Q2 for NAND-Flash contract prices, likely falling between 13%-18%, with enterprise SSDs landing on the high end of the spectrum. Kioxia and Western Digital are boosting production capacity utilization rates due to demand, while others remain conservative. There will be an overall dip in Q2 NAND-Flash purchasing, likely influenced by decreasing supplier inventories and production cuts. Western Digital’s supply shortage is evident, and the market demand will remain high for the foreseeable future.
Industry sources believe there will be a supply shortage for large-capacity HDD products through this quarter and possibly the year. Western Digital’s recent letter to customers only affirms this prediction.
Similarly, HDD and SSD prices could see another price increase of around 5% to 10% in 2Q24. According to TrendForce's sources, suppliers are unlikely to expand production immediately to sustain the upward price trend. Further stabilization depends on AI’s developments and whether HDD fits AI’s requirements or is outdone by enterprise SSDs or other high-performance components. Demand is already higher than previously forecasted.
Western Digital communicated that pricing adjustments will be frequent in the future and that the company has a limited ability to handle unplanned demand and orders. Customers are encouraged to notify their representatives early to accommodate changes better.
Samsung Receives CHIPS Act Funds
Intel, TSMC, BAE Systems, and others have received their portion of the highly desired United States CHIPS Act. TSMC was the most recent recipient, making headlines with its plans to build a third facility within the U.S. to join its other two fabs. Now, Samsung Electronics is joining these chipmakers with the Biden Administration’s announcement.
Samsung Electronics will receive $6.4 billion in funding to help construct a new semiconductor plant in central Texas. Samsung will invest $40 billion to build new plants in Taylor and expand its pre-existing facility in Austin. The plants will be used for semiconductor research and development and semiconductor packaging.
A semiconductor packaging facility will be another essential part of the developing U.S. semiconductor ecosystem. U.S. semiconductor companies still have to ship components to Taiwan to be packaged.
"For the first time, Samsung can conduct core research and development in the United States of America, support the future, manufacturing at scale, and advanced packaging all in Texas," Commerce Secretary Gina Raimondo told reporters on a conference call.
The Samsung plants will also benefit national security endeavors, as they manufacture chips directly for the U.S. Department of Defense (DoD).
"It will boost our production capacity for chips in critical U.S. industries, including aerospace, defense, and autos. And doing so will bolster our national security," Lael Brainard, Biden's top economic adviser, said.
As the third largest recipient of CHIPS Act funds, the Texas state government and the Semiconductor Industry Association (SIA) have expressed their excitement about Samsung’s investment. The White House estimates the construction of the new plants will create 17,000 jobs and require over 4,500 manufacturing jobs within the next decade.
In an official announcement, Samsung Semiconductor CEO Kye Hyun Kyung said, “We’re not just expanding production facilities; we’re strengthening the local semiconductor ecosystem and positioning the U.S. as a global semiconductor manufacturing destination. To meet the expected surge in demand from U.S. customers, our fabs will be equipped for cutting-edge process technologies for future products like AI chips and help bring security to the U.S. semiconductor supply chain.”
To support the demand for local talent, Samsung plans to “support and participate in the CHIPS R&D program such as NSTC and NAPMP to create new opportunities for the innovation of semiconductor industry. However, for our semiconductor ecosystem to succeed, it is absolutely critical that our efforts here attain active support from the public sector, including more inclusive STEM education, more eligibility for Investment Tax Credit programs, and more…”
More announcements on CHIPS Act recipients are expected to come soon. Each one is certain to strengthen the growing U.S. semiconductor industry in this next decade.
Domestic semiconductor facility plans continue to forge ahead within the United States. As CHIPS Act subsidies are doled out, new facility plans are on the table. It has been a long road, but the U.S. is finally underway in building a strong foundation for its domestic semiconductor ecosystem.
Not everything is sunshine and rainbows across the board. Due to the prolonged delay in awarding CHIPS funding, some companies have scaled back plans, if not canceled them altogether. It is still too early to tell how much of a detriment the extended wait time will have on U.S. semiconductor dreams.
TSMC Goes All in as SkyWater Technology Folds
After months of waiting for CHIPS funding, TSMC is finally receiving its portion of subsidies. Over the last few years, thanks to the CHIPS Act proposal, TSMC has been developing a vast expanse of the Arizona desert to produce its most advanced semiconductors. Apple has already booked out most of the chip capacity produced at TSMC’s Arizona location, making it a significant player in U.S. semiconductor goals and the broader economy.
TSMC’s Arizona plant plans have suffered setbacks throughout their creation. Initially, the problems were the lack of local talent, an indicator of the greater global semiconductor labor shortage. To remain on schedule, TSMC brought in some of its domestic workforce to train the incoming U.S. staff. This set back plans and upset local government officials, who stressed the importance of TSMC hiring domestic U.S. employees.
As TSMC grappled with this problem, it was besieged by another pressing issue: the postponed arrival of the coveted CHIPS funding. In response, TSMC pushed out its expected production start date from 2026 to 2027 or 2028. In late February, TSMC Chairman Mark Liu said Arizona’s progress depends on “how many incentives the U.S. government can provide.”
Now, TSMC is moving full speed ahead. The Biden Administration announced that the company would receive $6.6 billion in federal grants. With these funds, TSMC says it plans to invest another $25 billion to expand operations in Arizona, bringing the total number of TSMC facilities in the state to three.
The three plants will produce TSMC’s most advanced chips and, hopefully, its upcoming 2nm fabrication process. These chips will be used in consumer electronics and artificial intelligence (AI) applications. The $6.6 billion in grants includes an allotment of $50 million in workforce development and an additional authorization for up to $5 billion in government loans alongside manufacturing tax credits.
TSMC expects some of the labor challenges it encountered to lessen with each project. Mark Liu said, "Even though we encountered challenges in Arizona for our first fab construction ... we believe the construction of our second fab will continue to be much smoother."
As TSMC prepares for expansion, chipmaker SkyWater Technology is scaling back.
In a recent announcement by SkyWater Technology, plans to build a $1.8 billion semiconductor R&D and fabrication facility in Indiana have been temporarily scrapped. A Purdue spokesperson stated that SkyWater released its option on the land it planned to develop but would remain a partner with research opportunities in the works.
“As the details of the CHIPS initiative have unfolded over the last year, we have taken the opportunity to reassess our plans for new fab construction in conversations with the state of Indiana,” SkyWater Technology said. “While we don’t have a definitive plan targeting new fab construction in Indiana, we remain committed to growing the microelectronics ecosystem in the U.S., including Indiana, with emphasis on providing fab access and production support for emerging and strategic technologies.”
While disappointing, SkyWater Technology’s withdrawal isn’t a surprise. In the original announcement, Purdue stated that the project depended on receiving funds from the CHIPS and Science Act. However, Purdue isn’t starting back at square one with SkyWater’s pullback. South Korean memory giant SK Hynix plans to invest $4 billion in a semiconductor facility at Purdue Research Park.
Productivity and Prices are Up with AI
AI is having another stellar year. Continued demand for AI applications and growing competition between Nvidia and its rivals have led to a rapidly developing market. AI’s influence has aided the recovery of DRAM and NAND-flash markets alongside strategic production cuts by their manufacturers. Continued implementation of artificial intelligence will only bolster demand in the coming years, strengthening as the benefits of AI become more well-known.
Research by the International Data Corporation (IDC) shows that applying generative AI to various marketing tasks may result in an estimated productivity increase of 40%. IDC’s research results show that generative AI could handle over 40% of the collective work of marketing teams and potentially 100% of specific marketing tasks. These results will vary from company to company, but the productivity gains already show strong guidance for marketing teams of all sizes.
"In the next five years, GenAI will advance to the point where it will handle more than 40% of the work of specific marketing roles," said Gerry Murray, Research Director at IDC's Enterprise Marketing Technology practice. "Because of the rapid evolution of GenAI capabilities, marketing leaders must prepare their staff for fundamental changes to roles, skills, and organizational structure."
With the growing use cases for AI, such as IDC’s marketing productivity results, components used in these AI applications are also increasing in demand.
Over the past several weeks, there has been a developing shortage of enterprise solid-state drives (SSDs) within the NAND-Flash market. The shortage has been correlated to the AI boom and “subsequent construction of data centers by global big tech companies, with a particularly explosive increase in demand for storage devices needed for data centers.”
Following the ongoing trend by memory manufacturers over 1Q and 2Q of 2024, Samsung Electronics is expected to raise the price of its enterprise SSDs by 20% to 25% in 2Q24. With the surge in demand exceeding previous marker forecasts, Samsung Electronics is going beyond its original planned price increase of 15%.
A semiconductor industry insider told DigiTimes, “Server companies seeking to expand their storage capacity are rushing their SSD orders recently, and some products are even experiencing shortages, leading to considerations for increased production.”
The decision to increase SSD prices is also a result of accelerated AI-related storage server expansions by Nvidia and Tesla. Major server companies, including Dell Technologies and Hewlett Packard Enterprise (HPE), are competitively purchasing SSDs.
Continued reports highlighting AI implementation's benefits will bolster further aggressive purchasing trends among other companies, leading to higher prices for these components throughout the year.
Over the last several years, countries have worked overtime to improve domestic semiconductor manufacturing. The global semiconductor shortage emphasized the gaps left behind by declining chip ecosystems. The lack of diversity on a worldwide scale, with most chip manufacturers centered in one geopolitical area, further supported new domestic initiatives.
Access to semiconductors has become necessary for a country’s economy and national security. Because semiconductors are needed in almost every modern product, it’s pertinent for countries to protect this supply. Unfortunately, this need has contributed to a growing trade war between the two giants over the last few years. The chipmakers they’re trying to court are suffering from the fallout.
More terrifying than the back and forth between governments are devastating natural disasters, like the earthquake that occurred last week in Taiwan.
Earthquake Shakes Taiwan
One of the largest earthquakes in 25 years struck Taiwan on Wednesday morning, April 3rd. According to the Central Weather Administration, the 7.4 earthquake hit 15 miles south of Hualien before 8 a.m., and 76 aftershocks were recorded in less than five hours post-quake. In preliminary reports, over 1,000 people were reported injured, with nine tragically losing their lives.
On Wednesday, the quake and its resulting tremors, landslides, and damaged buildings trapped over 150 people. A silver lining in the terrifying aftermath was the location of the earthquake’s epicenter. On the eastern coast, Taiwan’s larger cities, including its capital, Taipei, were out of the earthquake’s danger zone on the island's west side. A tsunami warning was issued soon after the quake but was downgraded quickly.
Chipmaking giant TSMC quickly moved staff out of certain facility areas and evacuated others. United Microelectronics Corporation similarly halted production and evacuated some of its facilities in Hsinchu and Tainan. Other chipmakers or semiconductor industry players, such as ASE Technology Holding Co., are assessing the damage to their facilities.
TSMC said in a statement, "Safety systems are operating normally. To ensure the safety of personnel, some fabs were evacuated according to company procedure. We are currently confirming the details of the impact.”
Like many countries along the Pacific Ocean’s Ring of Fire, Taiwan is no stranger to earthquakes. However, Taiwan is more prone to quakes than others due to its proximity to the convergence of two tectonic plates. U.S. officials have long called for Taiwanese semiconductor companies, especially TSMC, to diversify operations off the island due to its earthquake vulnerability. A single tremor can destroy entire batches of precision-made semiconductors.
However, Taiwan hesitates to lose what has been deemed its “silicon shield” should its semiconductor manufacturing prowess decline. Since the global semiconductor shortage, Taiwan has been expanding its international footprint with new expansion projects in Japan and the U.S. These facilities will take several more years to reach full speed.
In the earthquake's direct aftermath, early assessments of building safety and production have been minimal. TSMC reports that personnel were safe and have returned to their workplace without concerns.
"A small number of tools were damaged at certain facilities, partially impacting their operations. However, there is no damage to our critical tools, including all of our extreme ultraviolet (EUV) lithography tools.”
Micron Technology, which also has operations in Taiwan, said its staff were safe and that once damage evaluations were completed, consumers would be informed of changes in delivery commitments.
New Developments in Trade War
The trade war between the U.S. and China continues to escalate. Over the last year, new restrictions or other export controls have been announced every few months. In 2023, the Cyberspace Administration of China (CAC) said Micron Technology had failed to pass a network security review. As a result, Micron products were banned from selling key domestic infrastructure products.
Luckily, according to experts, the impact on Micron’s sales was expected to be limited as most of Micron’s products were used in consumer electronics such as smartphones. Similarly, Micron planned to work alongside Chinese authorities to solve the problem.
Some security experts believed Micron’s ban was retaliation against earlier actions by the U.S. that limited the sale of advanced semiconductor manufacturing equipment and chips. In 2023, with AI's rising popularity, the U.S. placed restrictions on the sale of specific AI-capable components, such as Nvidia’s GPUs. China then restricted two elements critical to some chips: gallium and germanium.
There has been a new development in the trade war escalating tension.
According to a recent report by the Financial Times, Beijing has instructed official institutions in China to refrain from using PCs and servers equipped with microprocessors from Intel and AMD. Similarly, Beijing plans to have businesses reduce procurement of Microsoft Windows operating systems and database software outside of China.
Neither government officials nor companies have officially responded, but Microsoft and Intel have declined to speak on the situation. The Financial Times says Chinese authorities have “requested state-owned enterprises to promote localization internally.” Intel and AMD dominate nearly all global PC processor market shares.
The move is not entirely unexpected. With increasing restrictions, both countries are focused on establishing a secure and self-reliant domestic semiconductor ecosystem. Chinese officials released new guidelines on the latest PC, laptop, and server procurement standards in late December. Government departments at the township level and above were mandated to incorporate standards for purchasing “secure and trustworthy” processors and operating systems.
Most of the 18 approved processors came from Huawei and Phytium, with China-based processor manufacturers utilizing a hybrid architecture of existing components and self-developed designs. It will take a lot of work for China to achieve its goal of transitioning entirely to domestic processors. Some of its domestically produced processors, such as Zhaoxin’s KaiXian (KX)-7000, are years behind other manufacturers in terms of technology.
Likewise, as Huawei and SMIC seek patents on self-aligned quadruple patterning (SAQP) processes for chip production, experts believe they will eventually need ASML’s advanced lithography to remain competitive. One can only speculate how hard that will be should these restrictions become more stringent in the coming years.
After the harsh decline in the semiconductor industry last year, the DRAM and NAND-Flash market has had to pick itself up by its bootstraps. Seeing the most significant drops in consumer demand, leading memory manufacturers, including Samsung Electronics, SK Hynix, and Micron Technology, engaged in strategic production cuts to quickly prevent inventory overhang from worsening.
Even now, the same manufacturers and others are still planning to continue strategic production cuts and partial facility shutdowns. The goal is to keep supply and demand tight until the market enters its rebound stage in 2H24. Despite increased sales, the market is still relatively poor, which could be attributed to the typical low-sales season following end-of-year holiday shopping.
DRAM and NAND-Flash are seeing significant upticks in procurement now that buyers are stockpiling specific components. Artificial intelligence (AI) demand, specifically for high-bandwidth memory (HBM), is propelling it to become a major player in DRAM for 2024.
HBM Demand is a Rising Star for DRAM Rebound
HBM is rapidly increasing its market share within DRAM. TrendForce data in a recent report reveals that the DRAM industry will have allocated “approximately 250K/m or 14% of its total capacity to producing HBM TSV, with an estimated annual supply bit growth of around 260%.”
HBM is expected to increase its revenue share from around 8.4% in 2023 to 20.1% by 2024, a major spike.
The growing demand for HBM is likely due to two major factors. Firstly, HBM’s use in AI applications. Secondly, in 1Q24, unfulfilled DDR5 orders triggered aggressive buying trends among procurement teams. Buyers were stockpiling components because of the uncertain market outlook and undelivered products. HBM has a much longer production cycle than DDR5, which may have triggered similar strategies as buyers considered their long lead times.
TrendForce reports that “the die size of HBM is generally 35%-45% larger than DDR5 of the same process and capacity. The yield rate (including TSV packaging) for HBM is approximately 20% -30 % lower than that of DDR5, and the production cycle (including TSV) is 1.5 to 2 months longer than DDR5.”
The complex packaging required of HBM, which consumes three times the DRAM capacity for DDR5, has indirectly led to capacity constraints for non-HBM. This tight supply-demand has contributed to DRAM’s improvement in the market, as tight supply allows manufacturers like Samsung to increase product prices.
If TrendForce’s forecast rings true and HBM rises to 250K/m, Samsung Electronics total production capacity for HBM is 130K, while SK Hynix’s is 120K. Micron Technology’s CEO, Sanjay Mehrotra, says Micron’s HBM capacity has been fully allocated for 2024 and 2025’s capacity booked too.
Micron expects that its revenue forecast will surge 58% year-on-year (YoY), massively surpassing Wall Street expectations. This is a big turnaround from losing $2.3 billion in the same period last year. This year, reduced production and explosive AI growth have immensely benefitted upstream memory manufacturers.
Mehrotra said, “We believe Micron is one of the biggest beneficiaries in the semiconductor industry of the multiyear opportunity enabled by AI.”
New Production Method for China-Made 5nm Chips?
According to a recent Bloomberg report, patents for a chip production process called “self-aligned quadruple patterning (SAQP)” have been submitted. Huawei and presumably China’s Semiconductor Manufacturing International Co. (SMIC) are behind the submission to produce chips using this 5nm-class process technology.
SAQP is a process that has been used before. Unfortunately, the results were not overly spectacular. This process is cited as a contributing factor to the failure of Intel’s 1st generation 10nm-class process technology. Intel used the SAQP technology to eliminate reliance on high-end lithography, specifically extreme ultraviolet (EUV) lithography.
ASML explicitly dominates EUV lithography production. When one manufacturer holds most of the market share or is the only manufacturer, bottlenecks and shortages become more frequent. However, due to SAQP’s failure to measure up, Intel reverted to more conventional manufacturing processes, such as AMSL’s lithography machines.
Unlike Intel, Huawei and SMIC have little choice otherwise, having but cut off from leading-edge chip manufacturing tools due to export restrictions. SAQP technology is the only tool they may have to increase transistor density.
This doesn’t mean that Huawei and SMIC’s endeavor is automatically doomed to failure. The SAQP process could enable SMIC to build chips on sub-10nm technologies related to SMIC’s rumored 5nm fabrication process.
Huawei’s SAQP technology patent is reported to involve “etching lines on silicon wafers multiple times to boost transistor density, reduce power consumption, and potentially increase performance.” If this technology works as planned, it could help increase the production of more sophisticated chips by SMIC, such as the Kirin 9000S. Bloomberg also notes that SiCarrier, a state-backed chipmaking gear developer working with Huawei, has also been granted a patent for multi-patterning.
TechInsights Vice Chairman Dan Hutcheson believes that even if these multi-patterning technology processes help Chinese OCMs manufacture 5nm-class chips, they will eventually need the aid of EUV machines. The help of advanced lithography machines is necessary to remain competitive in the market and go beyond 5nm nodes. Without EUV lithography, the cost per chip using SAQP technology might exceed economic feasibility. Eventually, these companies must procure or develop their own EUV technology.
Advances in chip technology are critical to modern-day country economies. Consumers within China would like to possess competitive products, such as Apple iPhones, which require smaller and smaller chip nodes. SMIC must catch up to supply Huawei with the chips to accomplish such a feat, requiring advanced chipmaking process technologies.
Whether Chinese manufacturers will eventually be able to purchase EUV machines or develop their own is still up in the air.
Mitigation efforts continue to eliminate remaining inventory overhang and excess electronic component stock. Many manufacturers plan to continue with strategic production cuts or engage in partial shutdowns to keep supply-demand tight for the coming market rebound.
Organizations are seeking new ways to expand their global operations during the transition from glut to growth. Likewise, countries with small domestic semiconductor ecosystems or those eager to build more stable supply chains are considering new proposals to strengthen such areas.
Samsung NAND-Flash Price Could Increase by 20%
According to a report by the South Korean news outlet, The Chosun Daily, Samsung Electronics is coming out of the other side, after enduring the worst of the 2023 market downturn. Its strategic decision to reduce production on some of its product lines has paid off. Since late 2023, memory chip prices have consolidated at a price bottom and are now increasing.
In a recent report, an anonymous semiconductor industry source says that Samsung plans to increase its NAND-Flash chip prices by 20%. This is to restore the profitability of its memory chips. Price hikes are a common trend amongst DRAM and NAND-Flash suppliers this year, since forecasts were made in early 1Q24 regarding the likelihood of growing component costs.
“The first-quarter price negotiations between major memory manufacturers such as Samsung Electronics and SK Hynix and their customers have not yet been concluded,” said the industry source. “However, customers are rushing to secure supplies as the price of NAND flash has been steadily increasing, and fears spread that NAND flash cuts will continue this year.”
With growing demand for mobile, PC, and server products, Samsung plans to renegotiate prices with major customers in these areas for March and April. Samsung will push for an increase of 15% to 20%.
After months of declines in the memory sector and strategic production cuts, NAND-Flash prices have steadily increased over the last five months. Despite being in the middle of a traditional low-demand season, TrendForce data reports buyers are stockpiling NAND-Flash to establish safe inventory levels. This action is also reflected in the DRAM market, where undelivered stock for DDR5 products has increased buyer stockpiling.
With buyer attitudes changing, Samsung’s decision to push for higher prices to minimize previous losses is expected. While the NAND-Flash market is not as consolidated as the DRAM sector, NAND-Flash is still dominated by five major players, with Samsung and SK Hynix holding the greatest market share. Samsung continues to remain in the top position.
If Samsung goes forward with the planned increase, the others will likely follow suit.
UAE Investors Interested in OpenAI Plans
Earlier this year, Sam Altman, CEO of ChatGPT creator OpenAI, came out with an ambitious and expensive proposal: a multi-trillion-dollar venture to establish an artificial intelligence component supply chain. This proposal would help OpenAI achieve its goal of creating its own semiconductor chips in-house, reducing OpenAI’s reliance on Nvidia, the current king of AI chips.
OpenAI made a deal with Thrive Capital by selling company shares in February 2023 as part of its process and funding efforts. This dramatically increased the company's valuation, skyrocketing the price to over $80 billion—nearly a threefold increase in under ten months.
It is a crucial step in the right direction, but OpenAI needs a lot more to achieve its goals. It's one of the reasons Altman is courting leadership in the United Arab Emirates (UAE). UAE’s state-backed group, MGX, is reportedly in discussions to back OpenAI’s venture, according to two people with “knowledge” of these talks.
MGX, the group behind the potential UAE investment, is an AI-focused fund recently formed and headed by UAE’s National Security Advisor, Sheikh Tahnoon Bin Zayed al-Nahyan.
The fund was also created in collaboration with “G42 and Mubadala, the former having already entered into a partnership with OpenAI in October 2023 as a part of the company's Middle East expansion.”
Altman says the partnership with G42 plans to bring AI solutions that “resonate with the region's nuances.” The UAE and others are attempting to establish themselves as hubs for emerging technologies such as cryptocurrencies and the metaverse.
Sources say the fund hopes to create “a structure that will put Ahu Dhabi at the center of this AI strategy with global partners worldwide.” The UAE’s recently launched free zone will help support companies involved in digital and virtual assets such as blockchain and Web3.
TSMC is Looking to Japan to Increasing Advanced Chip Packaging Capacity
During 2023, despite the overall market decline, there were some bottlenecks, specifically with components used in artificial intelligence. TSMC saw dramatic drops for several product lines but could not meet the high demand for advanced packing orders or its chip-on-wafer-on-substrate (CoWoS) packaging.
This advanced packaging technology is used in Nvidia’s flagship GPUs, which propelled the company into the trillion-dollar club over 2023. TSMC’s CoWoS is a high-precision technology that stacks chips on each other to boost processing power while saving space and power consumption. All of TSMC’s CoWoS capacity is located within Taiwan.
According to two unnamed sources, Reuters says that TSMC wants to expand its advanced packaging capacity with new facilities within Japan. Deliberations are still in the early stages, but should they evolve, this move would be a welcomed boon to Japan’s domestic semiconductor revitalization efforts.
TSMC’s ongoing attention in Japan might indicative its future global expansion plans. TSMC just built one plant and is moving forward with a second plant in Kyushu, same as the first. TSMC is partnering with Sony and Toyota for these facilities, with a total investment expected to reach over $20 billion.
As a leader in semiconductor materials and equipment makers, Japan is uniquely poised to take a more prominent role in advanced packaging should it be given the opportunity. However, TrendForce analyst Joanne Chiao said that should TSMC move forward with building advanced packaging capacity in Japan, it would be limited in scale.
Besides TSMC, Intel and Samsung want to establish advanced packaging capacity within Japan. Intel declined to comment on specific plans, but two separate sources have spoken about Intel’s interest in establishing an advanced packaging research facility within the country. With government support, Samsung Electronics is already working on setting up its own facility in Yokohama.
The renewed interest in manufacturing capacity within Japan is helping the country’s revitalization efforts in semiconductor production capacity. Like other countries, Japan hopes to regain its lost domestic semiconductor manufacturing ecosystem ever since the dramatic effects of the global semiconductor shortage.
The semiconductor industry is slowly gaining traction after a year of poor consumer demand. Artificial intelligence spurred a wave of orders for specific components, especially Nvidia’s GPUs, but it couldn’t draw the entire industry out of its decline. Many manufacturers, including memory giants Samsung Electronics and SK Hynix, engaged in strategic production cuts to soften the harsh blow.
According to experts, these mitigation efforts have helped reduce the inventory overhang currently plaguing the semiconductor industry. However, more corrections are needed in the next few months for a truly successful market rebound in 2024. That said, semiconductor demand is ramping up, a great sign after a year of low sales.
Global Semiconductor Sales are up 15% in January 2024
The Semiconductor Industry Association (SIA) has good news for the semiconductor industry. Year-on-year (YoY) analytics indicate that the semiconductor industry saw sales increase 15.2% YoY in January 2024 compared to January 2023. Sales in January reached $47.6 billion, up from $41.3 billion in 2023. However, January fell behind December 2023 by 2.1%, or $48.7 billion, in total sales.
The slight decline from December to January is easy to explain. December, a massive holiday shopping season, is one of the best months for companies globally. Numerous businesses, such as Apple, release new products around late Q3 and early Q4, sparking interest and demand. The fact that January demand didn’t drop too drastically is good news.
“The global semiconductor market started the new year strong,” said John Neuffer, SIA President and CEO. “With worldwide sales increasing year-to-year by the largest percentage since May 2022. Market growth is projected to continue over the remainder of the year, with annual sales forecast to increase by double-digits in 2024 compared to 2023.”
Regionally, year-to-year sales were up across China, the Americas, Asia Pacific, and others, with exceptions in Japan and Europe. Similarly, month-to-month sales were down across all markets, as expected in the post-holiday shopping season.
Monthly sales data is compiled by the World Semiconductor Trade Statistics (WSTS) organization, with the SIA representing 99% of the U.S. semiconductor industry and 2/3rds of non-U.S. chip firms.
This news comes right after the SIA’s most recent analytics showing that the semiconductor industry saw an 8.2% decline in global sales revenue in 2023. This noticeable decline contributed to some strategic production cuts and new facility construction delays as chipmakers, distributors, and others prioritized retaining capital during the worst of the slowdown.
The industry isn’t out of the woods yet, but it is gaining ground on clawing its way back to a rebound.
Tower Semiconductor Halts Operations
While the semiconductor industry sees positive numbers indicating returning consumer demand, excess electronic component inventory remains. Mitigation efforts continue, with some chipmakers taking more drastic steps than others. Samsung Electronics and SK Hynix made headlines last year for their decision to engage in production cuts after memory saw the worst of the market decline.
According to research firms, the first half of the year will be critical in efforts to reduce the remaining industry overhang. Customers are still reducing orders as they clear inventory that has accumulated over the past two years. Considering the fragile state of the semiconductor industry recovery, memory manufacturers, despite seeing increased orders in DRAM and NAND-Flash, are continuing production cuts to keep supply tight.
Similarly, Tower Semiconductor is planning a 3-week shutdown for most of its operations in Newport Beach, as Data Center Dynamics reports. Nearly 700 employees will be taking leave from April 1st to 7th. Further shutdowns are expected from July 1st to 7th and October 7th to 13th. The announcement also details that most of Tower’s employees will be affected by the 3-week furlough period.
According to IJIWEI, Tower Semiconductor management announced that its equipment will remain in a “warm shutdown,” and power to unused plant areas will be shut off. This is typical for semiconductor facilities as they cannot be switched “on and off” from no production to full production the same way an automotive manufacturing line can.
Despite Tower Semiconductor’s planned shutdown and sales falling behind other chip behemoths like Intel and TSMC, these actions do not indicate the company’s performance. Tower Semiconductor produces components for major clients like Broadcom and enjoys a high market position in rapidly growing fields like electric vehicles (EVs).
Likewise, Bloomberg recently reported that Tower Semiconductor has proposed a $9 billion investment to establish a facility in India. This planned facility is a step in Tower Semiconductor’s overall strategic goal of expanding its global operations and producing 80,000 wafers per month.
Like so many other chipmakers, Tower Semiconductor has needed to adjust due to lingering inventory challenges and last year’s decline. Given the company's massive room for growth, it is unlikely that Tower Semiconductor will continue this trend in the future. Tower is preparing itself for the coming market rebound during its transitional period.
The semiconductor market is going through a transition. Consumer demand hasn’t returned in full yet, and excess inventory mitigation remains a priority among chipmakers. There are still several months until the market rebound many hope for is expected to arrive. In the meantime, low customer demand is affecting everyone, even AI-chip leader Nvidia.
Nvidia’s H100 Is Being Resold By Customers
After a year of explosive growth and demand, Nvidia’s highly coveted and elusive products are beginning to see challenges in the company’s AI-chip dominance. The popular and scarce Nvidia data center GPU, H100, is experiencing a noticeable reduction in lead time and improved market conditions.
Over 2023, despite poor market conditions, Nvidia experienced spectacular growth after the AI boom, thanks to ChatGPT’s success. Customers, including tech giants like Microsoft and Google, battled to purchase large stockpiles of Nvidia’s components, including H100. While there was no genuine shortage for these GPUs in 2023, bottlenecks occurred. TSMC struggled to meet production capacity for advanced packaging utilized in Nvidia’s GPUs.
Due to the recent easement in market conditions and part lead time, customers who once purchased large quantities of H100 chips are beginning to resell them.
In a report by Tom’s Hardware, the H100 saw a reduction in delivery wait times from 8-11 months to 3-4 months, indicating a relief in supply pressure. While there is still an ongoing surge for artificial intelligence, major cloud suppliers such as AWS, Google, and Microsoft Azure offering access to AI computing services have also contributed to enterprises' decisions to offload their GPUs.
Reduced scarcity and high maintenance costs have led to enterprise customers selling their stockpiles, but this change does not indicate a more significant market trend. Current orders for AI remain robust, with large-scale artificial intelligence model computation keeping overall demand greater than the supply. Ongoing AI popularity and integration efforts have prevented a steep drop in price for H100 GPUs.
According to the report, customers are now prioritizing price and practicality when leasing AI computing services from cloud service providers rather than immediately going all in on their own AI services. Likewise, competitors' alternative solutions to the H100 have since emerged on the market, offering comparable performance and software support. These new products provide some customers with more affordable pricing, contributing to a more equitable market condition.
In new data by TrendForce, research suggests that the 2024 market landscape for high-demand AI servers powered by Nvidia, AMD, or other top-tier ASIC chips will be influenced by North America’s cloud service providers. Around 60% of the global demand will be divided between Microsoft, Google, AWS, and Meta, with Microsoft leading the charge with 20.2%.
However, despite the saving grace, Nvidia is still contending with some challenges. These problems go beyond Nvidia’s stronghold in AI, as even with competition from AMD and others, Nvidia’s GPUs hold 70% of the AI market. The issue lies with limited growth, attributed to U.S. restrictions on technological exports to China. Nvidia, trying to enter China’s market with chips that work around restrictions, such as H20, might not be able to challenge Huawei, which has capitalized on Nvidia’s market absence.
Similarly, while Nvidia holds the AI throne now, AMD is quickly rising as a competitor with cost-effective strategies. AMD offers similar products at 60%-70% of Nvidia’s comparable model price. This has allowed AMD to break into the market effectively, carving out a section against Nvidia’s pressure. Microsoft, one of Nvidia’s largest clients, is expected to be “the most enthusiastic adopter of AMD’s high-end GPU MI300 solutions in 2024.
The M1300 series, debuted by AMD last year, offers similar software solutions that developers have praised Nvidia for. Should H100 continue to see reselling throughout the year, it will probably not indicate poor customer demand but rather the growing competition Nvidia faces.
Newport Wafer Fab Acquired by Vishay
Since the summer of 2021, the former Newport Wafer Fab factory in the UK has been awash in national security concerns and other controversies. In July 2021, Nexperia, the Dutch-based technology company subsidiary of Shanghai-listed Wingtech, purchased a majority stake in the plant. The action led many ministers in the UK government to express concerns over national security if a Chinese company controlled the firm.
As a result of these concerns, the UK government forced Nexperia to sell 86% of its stake in what was then known as Newport Nexperia. Most of the problem seemed to originate from the worry that China would be able to undermine the UK’s domestic semiconductor manufacturing ability by transferring knowledge of critical technology used in modern digital devices to a Chinese-owned entity.
In November 2023, Vishay announced it would buy the plant outright from Nexperia for $177 million. However, just like before, the purchase wasn’t a simple handoff. The deal still required national security clearance by the UK’s Cabinet Office. This clearance would take another four months before Secretary of State Oliver Dowden issued the consent order.
The Newport Wafer Fab is the UK’s largest manufacturer of semiconductors and mainly supplies components for the automotive industry. Vishay is eager to expand the site and prioritize research and development of compound semiconductors.
Vishay and the UK government will have to keep in close communication, as stated in some of the conditions that came with clearance for the purchase to be approved. If Vishay plans to enter future agreements to sell, transfer, or lease to a third party that would give them access to the property, the company must inform the UK government. Likewise, ministers want the intellectual property and sensitive information now owned by Vishay upon the Newport Wafer Fab purchase to be tightly controlled.
In an official company statement, Toni Versluijs, Country Manager Nexperia UK, said that Nexperia would have liked to continue its strategy with the Newport Wafer Fab location but agrees that Vishay’s solid customer base and capabilities will ensure the site’s future success.
UK’s Economy Minister, Vaughan Gething, said: "I am pleased that the long-overdue decision to permit the acquisition of Newport Wafer Fab by Vishay International has now been taken. Today's news brings security to a hugely talented workforce after a long period of uncertainty, and I hope they can look forward to a new sense of optimism."
The semiconductor industry is poised for a market rebound in 2H24. New semiconductor manufacturing facilities are entering construction or preparing to begin chip production. Due to low consumer demand in 2023, some chip manufacturers have readjusted their schedules, no longer expected to commence production in 2024, as previously suggested. In the United States, the government's delay in awarding the CHIPS Act’s highly coveted subsidies has led to slowdowns in construction as manufacturers wait to receive additional funding.
That’s not to say all countries are facing similar delays. While some manufacturers have scaled back in the United States, waiting for more suitable timeframes, they are ramping up production in other global locations. Geopolitical volatility has pushed for greater diversification, and many chipmakers are eager to add new locations worldwide to their corporate footprint.
With artificial intelligence expected to buoy demand as a primary growth driver in the coming months, these new facilities can’t come soon enough.
TSMC Joins Japan’s Domestic Chip Push
TSMC isn’t just bringing its power to the desert of Arizona in the U.S. Last month, TSMC opened its first chip fabrication plant in Japan, with a second one in the offing. Construction on the second plant may begin at the end of 2024 and represents TSMC’s continued push toward diversification outside of Taiwan.
The new chip fabrication plant in Kumamoto is slated to begin production in late 2024. The facility was constructed by Japan Advanced Semiconductor Manufacturing Inc. (JASM), majority-owned by TSMC. Established in 2021 with support from the Japanese government, Sony Semiconductor Solutions, and Japanese automotive components maker Denso Corporation, the goal was to help power Japan’s semiconductor ecosystem.
Japan, once a powerhouse in semiconductors, lags ten years behind chip-making leaders TSMC and Samsung, according to the Center for Strategic & International Studies. Trade challenges and disputes with the U.S., missed opportunities during the PC revolution, and other scenarios have contributed to Japan’s decline. Since the global semiconductor shortage, Japan has been striving to strengthen its semiconductor presence, much like the European Union and the United States.
With the ongoing trade war between the U.S. and China, Taiwanese companies like TSMC are eager to diversify production outside Taiwan. Japan has become a great option.
Earlier this year, TSMC, Sony Semiconductor Solutions, Toyota, and Denso agreed to further invest in JASM for the second planned plant. It will produce semiconductors for automotive, industrial, and consumer uses for high-performance computing-related needs. Morris Chang, TSMC’s founder, believes that the resurgence of Japan’s semiconductor industry will begin with the inauguration of TSMC’s first facility, enhancing the country’s resilience.
The Japanese government, represented by Minister Ken Saito, has shown “strong support for TSMC's investment, viewing it as a model for semiconductor industry growth. The new plant is already positively impacting the local economy in Kyushu, offering higher-than-average wages and promising to contribute to the region's economic development,” according to Reuters’ latest article detailing the plan.
Chang is very optimistic about TSMC’s success in Japan, citing the country’s history of production quality and previous positive encounters, including his 1968 joint venture with Sony and Texas Instruments. With the rising demand for artificial intelligence, Chang says it’s not just that AI chipmakers are looking for tens of thousands of wafers but “multiple new fabs with formidable semiconductor manufacturing capacity.”
A need that TSMC’s new plant can offer. Should it continue, especially with the Japanese government doling out investments, TSMC’s total subsidy would put it beyond the 1 trillion-yen mark.
Nvidia’s Earnings Report Powers Up AI Sector
Nvidia recently released its earnings report for the end of 4Q2023. It’s a big one, which isn’t unexpected news considering Nvidia’s prolific boom last year thanks to the rising popularity of artificial intelligence. Two days after it released its quarterly earnings report, Nvidia briefly hit $2 trillion in market value for the first time. This high comes directly from the insatiable demand for its components that power most AI applications today.
Beyond Nvidia, the AI craze has made winners out of many AI companies, including Super Micro Computer and Arm Holdings. Nvidia’s shares are up 60% year-to-date, Super Micro Computers have grown by 200%, and Arms’s shares are up 77%. For Nvidia, 2023 was a record year in revenue, coming in at $60.9 billion, up 265% from a year ago.
“Accelerated computing and generative AI have hit the tipping point. Demand is surging worldwide across companies, industries and nations,” said Jensen Huang, CEO of Nvidia.
“Our Data Center platform is powered by increasingly diverse drivers — demand for data processing, training and inference from large cloud-service providers and GPU-specialized ones, as well as from enterprise software and consumer internet companies. Vertical industries — led by auto, financial services and healthcare — are now at a multibillion-dollar level,” Huang said about Nvidia’s recent report.
Huang continued, “NVIDIA RTX, introduced less than six years ago, is now a massive PC platform for generative AI, enjoyed by 100 million gamers and creators. The year ahead will bring major new product cycles with exceptional innovations to help propel our industry forward. Come join us at next month’s GTC, where we and our rich ecosystem will reveal the exciting future ahead.”
Furthermore, Nvidia expects this growth to continue through 2024, into 2025 and beyond. Its earnings report included previous product announcements, detailing which components supported its fantastic rise over 2023. The markets that contributed the most to Nvidia’s climb were data centers, gaming, professional visualization, and automotive sectors. In 2024, similar exciting launches are bound to come.
The AI craze will help drive the semiconductor market to rebound and Nvidia will be there to provide the next generation of leading-edge AI chips.
The semiconductor industry is on its way toward a market rebound. As recovery approaches, funding for semiconductor programs and facilities continues to be proposed and given out to chipmakers across the globe. In the United States, the multi-billion-dollar CHIPS Act that sparked over $200 billion in investments from chipmakers is finally starting to be divvied up.
The market downturn in 2023 and tax concerns have delayed that process until now. With the slow response of the U.S. government to hand out the CHIPS funding, some chipmakers are starting to readjust their projected production schedules as concerns over CHIPS incentives rise. This could evolve into a larger problem, since market growth is expected to return before global production capacity has been sufficiently increased.
GlobalFoundries Receives CHIPS Funding
The U.S. government has recently awarded GlobalFoundries $1.5 billion to subsidize the company’s semiconductor production. This is the first major announcement of the CHIPS Act’s funding allocation since its passage. The $39 billion fund for domestic semiconductor production continued to be divvied up in the coming weeks. GlobalFoundries will use its allotment to expand and upgrade its facilities in New York and Vermont.
These facilities are responsible for producing components utilized mainly in automotive and defense industries. The coming improvements to both factories are especially welcome after the impact of the global semiconductor shortage. Both automakers and defense contractors ran into numerous problems when they could not secure necessary stockpiles.
"The chips that GlobalFoundries will make in these new facilities are essential to our national security," Commerce Secretary Gina Raimondo told reporters. Later, Raimondo told Reuters that the agency is in active talks with numerous applicants and expects to make several announcements by the end of March.
"We're in the process of really complicated, challenging negotiations with these companies," Raimondo said. "These are highly complex, first-of-their-kind facilities. The kind of facilities that TSMC, Samsung, Intel, are proposing to do in the United States -- these are new-generation investments -- size, scale complexity that's never been done before in this country."
Despite the good news, production delays have haunted chipmakers over the last year. TSMC and its widely publicized $40 billion Arizona plant have faced several obstacles. Last summer, TSMC had to push back its production start date from 2024 to 2025 when it needed more workers with the expertise to install sophisticated equipment.
More recently, TSMC stated that its second plant will not produce chips until 2027 or 2028 rather than 2026 due to concerns over tech choices and federal funding. TSMC Chairman Mark Liu, said that Arizona’s progress depends on “how much incentives the U.S. government can provide.”
TSMC isn’t the only chipmaker encountering problems either. Intel, Microchip Technology, and others have needed to adjust production schedules to manage infrastructure due partly to the sales slump in microelectronics last year. New facilities are massively complex. To ensure organization and focus, funding and construction timelines were delayed to prioritize surviving the worst of the market decline last year.
Meanwhile, concerns have risen over the success of the CHIPS Act. The U.S. government has been locked in negotiations with major chipmakers over the funding amount and timing. These impediments could lead to rival countries with subsidy programs, such as the EU and East Asia, obtaining other projects.
“The longer the U.S. government waits to distribute benefits, the more other geographies are going to snap up these investments, and more leading-edge investments will be made in East Asia,” said Jimmy Goodrich, Senior Advisor for the RAND Corporation. “So, the clock is ticking.”
White House and Treasury Department spokespersons say that the delays have been due to questions over tax credits and the overwhelming number of companies that submitted interest statements for the grants. After the shock of the low demand drop from last year, many organizations have been left with inventory overhang and no immediate need for new factories.
“Companies are rethinking how and what and when investments will occur,” said Thomas Sonderman, the Chief Executive of SkyWater Technology.
Microchip and Intel have adjusted purchases of factory tools, waiting until business conditions improve before considering investing further. Operations have slowed but not ceased entirely. Others, such as Micron Technology and Texas Instruments, are pushing ahead despite the market downturn. Rather than delay the inevitable, Micron’s Vice President, Scott Gatzemeier, said that construction projects should be based on future demand over current conditions.
“Once you start, you don’t want to stop,” he said. “Renting massive cranes and other equipment and securing construction workers are big expenses that might need to be repeated if a project is halted.”
Arm Considers AI as SIA Forecasts Major Growth
SoftBank, the controlling stakeholder behind world-renowned chip designer Arm, is considering a new strategy to take on the reigning king of artificial intelligence (AI) chips, Nvidia. This new AI strategy will require a hefty $100 billion investment, broken up by $30 billion from SoftBank and $70 billion from Middle Eastern institutions.
Codenamed Izanagi, the recent news increased SoftBanks’ share price by 5%. Arm’s chip designs are crucial for dozens of chipmakers, including Nvidia. Due to this, antitrust concerns could become prominent if Arm partners with any specific chipmaker, echoing Nvidia’s failed takeover of Arm two years ago.
Something else that sounds familiar is the recent trend of turning to the Middle East for investments. OpenAI’s Sam Altman is currently attempting to do the same with his proposed multi-trillion-dollar AI supply chain venture. Like Arm, there is concern that Altman could unwittingly put OpenAI in a place of antitrust scrutiny if it ends up being a separate venture from OpenAI. Altman would require the blessing of the U.S. government before receiving investments from the Middle East as well.
Despite the challenges, SoftBank’s push into AI is an expected next step for the chip designer, considering the overwhelming growth in AI over the past year. The semiconductor market is poised to rebound, with memory and AI leading the charge. The Semiconductor Industry Association (SIA) forecasts that the semiconductor industry will see a sales rise of 13% in 2024 after announcing a decline of 8.2% in 2023.
“Global semiconductor sales were sluggish early in 2023 but rebounded strongly during the second half of the year, and double-digit market growth is projected for 2024,” said SIA CEO John Neuffer, “with chips playing a larger and more important role in countless products the world depends on, the long-term outlook for the semiconductor market is extremely strong. Advancing government policies that invest in R&D, strengthen the semiconductor workforce, and reduce barriers to trade will help the industry continue to grow and innovate for many years to come.”
Investments within the semiconductor industry are on the rise. Organizations and governments are financing every part of the electronic components supply chain, from research and development to artificial intelligence manufacturing. The U.S. government continues to fund semiconductor-related research and development under the CHIPS Act, as OpenAI’s Sam Altman embarks on his most ambitious plan yet.
The U.S. Continues its Investment in Semiconductors
To support domestic semiconductor manufacturing, the U.S. passed the CHIPS and Science Act, a great start to revitalizing this neglected industry. Since its creation, the U.S. has seen over $200 billion in investments for U.S. semiconductor manufacturing. Under the law, the allocated $52.7 billion was split into several areas, with $39 billion in subsidies for semiconductor production and $11 billion in research and development.
At the center of this new push for research and development is the National Semiconductor Technology Center (NSTC). The NSTC, in U.S. Commerce Secretary Gina Raimondo words, is “a public-private partnership for the government, industry customers, suppliers, academics, entrepreneurs, and venture capitalists to come together to innovate, connect, network, solve problems and allow Americans to compete and outcompete the world.”
The NSTC will establish an investment fund to help “emerging semiconductor companies advance technologies toward commercialization.” The center will also conduct research and prototyping of advanced semiconductor technology.
Raimondo spoke about the plan with Reuters, discussing the upcoming announcement to award funding to several chip manufacturers. “These are highly complex, first-of-their-kind facilities…These are new-generation investments–size, scale complexity that’s never been done before in the country.”
More recently, the National Center for the Advancement of Semiconductor Technology, a nonprofit created to operate the NSTC, appointed Deirdre Hanford as its CEO. Hanford said the first step will be to “build a community whose members will help define the strategy and investments core to the semiconductor R&D ecosystem.”
Hanford said this represents a “once-in-a-generation opportunity” to create a new institution that will give the U.S. semiconductor industry a boost that Washington hopes for.
Alongside the NSTC, there are three other programs to improve research and development within the U.S. These new programs under the CHIPS Act are the National Advanced Packing Manufacturing Program, the CHIPS Metrology Program, and a CHIPS Manufacturing USA Institute.
"With strategic investments in R&D complementing targeted industry incentives, CHIPS for America will not only bring semiconductor manufacturing back to the US – it will keep it here for good," said Raimondo.
Semiconductor research and development will help fuel economic growth, secure national security, and improve technological competitiveness. The NSTC will play a significant part in that. According to the Semiconductor Industry Association (SIA) and Boston Consulting Group (BCG), five key areas that the CHIPS Act R&D’s funding should strengthen are:
- Transitioning and Scaling Pathfinding Research
- Research Infrastructure
- Development Infrastructure
- Collaborative Development
- Workforce
These projects should create over 400,000 jobs within the semiconductor ecosystem and support thousands more throughout the economy.
OpenAI’s Sam Altman Seeks Trillions to Reshape the Semiconductor Industry
The capabilities of large language models (LLMs) and generative artificial intelligence have taken the world by storm, thanks in part to OpenAI’s ever-popular ChatGPT. The chatbot, released in late 2022, quickly captured the interest of companies and countries alike. Over the last year, everyone from tech giants to small businesses has developed their own version of OpenAI’s successful LLM.
There were some stumbling blocks, mainly regarding Microsoft and Google’s accidental missteps with Bing and Bard, which have since been rebranded and corrected but are not enough to trip up AI’s rise to the top. Over the last year, OpenAI has been working alongside Microsoft to enhance the latter’s AI application, Copilot, and petitioning other technology giants to establish AI regulations and universities to educate students.
Now, OpenAI’s CEO, Sam Altman, is exploring a new area, and it could be one of the company's most significant undertakings yet.
Sam Altman is currently in talks with investors and others, including the United Arab Emirates government, to raise funds for “a wildly ambitious tech initiative that would boost the world’s chip-building capacity, expand its power-to-power AI, and other projects.” Such a zealous action plan could cost as much as $5 to $7 trillion—quite the expense.
AI’s expansion has been limited in comparison to its possible potential due to the scarcity and expensive nature of AI chips. This problem has placed constraints on OpenAI and other AI companies’ growth. As Altman explained, these chips are necessary to train large language models effectively but are far too limited. The number of GPUs needed to help push for greater AI use would need to be significantly expanded, hence the reason for this massive fundraising initiative.
However, an investment of this size is bigger than the current worth of the global semiconductor industry, which reached $527 billion in 2023, down 8.2% from poor sales and economic decline. The semiconductor industry isn’t forecast to even reach $1 trillion in size until 2030 at its current growth rate. This investment is even more significant than the national debt of some countries, like the U.S., with its corporate debt reaching $1.44 trillion. The combined market capitalization of Microsoft and Apple, the two highest-valued businesses in the U.S., is $6 trillion.
However, this plan might be necessary for OpenAI’s continued growth toward reaching human-level artificial intelligence. The support needed for such an extreme level of AI would require a solid and secure global-spanning network of funders, industry partners, and governments to provide funding and energy–of which AI facilities require oceans.
In the last few weeks, Sam Altman has met with U.S. Secretary of Commerce Gina Raimondo, U.A.E.’s Sheikh Tahnoun bin Zayed al Nahyan, a top security official and brother of U.A.E.’s President Mohamed bin Zayed al Nahyan, and the chair of numerous vehicles of Abu Dhabi’s sovereign wealth. Sam Altman has also discussed this initiative with Masayoshi Son, Softbank's CEO, and TSMC's representatives.
The U.A.E. and its growing financial portfolio would be significant players in the venture but is dependent on allowance from the U.S. government. Altman plans on raising money in the Middle East to build new chip-fabrication facilities and have TSMC build and run them. Microsoft supports Altman’s goal in this venture. Whether it will prove to be successful is another question.
A new supply chain that helps support the growing AI sector will be necessary, considering how limited GPU supply is already causing bottlenecks. Whether this challenge will be solved through Altman’s venture, or another initiative is left to time.
The semiconductor industry will see many changes this year. Research shows that 2023 saw continual declines in late Q3 and Q4 despite inklings of oncoming recovery. Memory giant SK Hynix stated in October 2023 that initial signs of chip recovery were beginning to appear. A month later, SK Hynix and Samsung Electronics indicated that signal demand weakness bottoming out was a good sign of recovery in 1Q24.
SK Hynix and Samsung Electronics have cited the rising popularity of artificial intelligence (AI) applications boosted chip demand, preventing the market downturn from growing worse. Despite the recovery outlook, SK Hynix and other memory manufacturers are still planning production cuts to keep supply-demand tight in 1H24. Market research predicts that there will be a rebound, like SK Hynix and Samsung Electronics expected in late 2023, but it could take longer than the original hope of 1Q24.
Market Demand to Rebound Coming Late in the Year After Declines
A recent Semiconductor Industry Association (SIA) report revealed that the semiconductor industry’s revenue was down 8.2% in late 2023 compared to the same time frame in 2022. Despite the downturn, sales peaked during the second half of 2023, reaching $526.8 billion worldwide.
“Global semiconductor sales were sluggish early in 2023 but rebounded strongly during the second half of the year, and double-digit market growth is projected for 2024,” said John Neuffer, SIA President and CEO. “With chips playing a larger and more important role in countless products the world depends on, the long-term outlook for the semiconductor market is extremely strong. Advancing government policies that invest in R&D, strengthen the semiconductor workforce, and reduce trade barriers will help the industry continue to grow and innovate for many years to come.”
Counterpoint Technology Market Research reported similar findings, with global semiconductor industry revenue declining by 8.8% in 2023. This was due to a slowdown in enterprise and consumer spending. Counterpoint Research reported that 2023 was “a year for semiconductor companies to fine-tune their strategies/outlook and manage inventory adjustments to prepare for the impending AI boom.”
William Li, Senior Analyst at Counterpoint Research, said, “In general, we believe artificial intelligence (AI server, AI PC, AI smartphone, etc.) will continue to be a major organic growth driver in the semiconductor industry in 2024, followed by the memory sector’s rebound due to normalizing oversupply situation and demand recovery. The automotive sector could be another driver for the market due to content growth, which was already a key revenue driver for Infineon and STMicroelectronics in 2023.”
Nvidia will continue to lead the charge in market industry regrowth thanks to its high market share of general-purpose GPUs utilized in AI/high-performance computing.
The industry is at the end of its inventory correction cycle, and the current support from consumer demand is solid. Supply constraints may continue to be a challenge this year from outside factors, so it will be important for companies to be cautious despite the positive market forecast.
Challenges That Might Impact the Global Supply Chain in 2024
Stability is still far from reach after months of shortages, excess inventory, long lead times, and sharp price changes. Most research firms believe 2024 will be a year of growth for the semiconductor industry. The expectation of how much growth will come depends on the research firm's optimism. Edgewater Research has expressed a more tepid outlook, viewing 2024 as a transitional year.
In contrast, the International Data Corporation (IDC) and World Semiconductor Trade Statistics (WSTS) are far more optimistic about the market outlook. Most research points to AI and memory components being the main drivers of growth in 2024, with a significant rebound occurring in 2H24.
The semiconductor industry isn’t out of the woods just yet. Sourceability’s Senior Vice President of Sales, Josh Pucci, discussed six of the biggest challenges poised to cause significant supply chain disruptions if left unchecked.
There is a silver lining: opportunities for new avenues for business development and growth for every problem. The six issues EPS News listed in its recent article, “6 Trends Shaping the 2024 Supply Chain,” are:
- Geopolitics: Geopolitical volatility has risen in the last several years. Currently, the world is grappling with strife as the Russia-Ukraine war and Israel-Hamas conflict continue. Houthi rebel attacks on vessels in the Red Sea have delayed logistics as companies reroute around Cape Good Hope, and the ongoing trade war between the U.S. and China has contributed to restrictions on manufacturing equipment and raw materials.
- Climate Change: Billion-dollar weather events are on the rise, leading to costly droughts, fires, floods, and storms that disrupt nearly the entire length of the global semiconductor supply chain. In Panama, shipping companies have been delayed by a historic drought that has led to heavy restrictions on passage through the much-needed Panama Canal.
- Environmental, Social, and Governance (ESG): Sustainability in the semiconductor industry is becoming a major priority as companies work to reduce their overall carbon footprint. Semiconductor manufacturing is known for its, unfortunately, toxic past with dangerous chemicals leeching into groundwater. Intel has been working alongside cleanup agencies in the U.S. to help detoxify its previous facilities.
- Cyber Attacks: It’s said that the next war will be fought on the frontlines of cyberspace and sophisticated malicious malware attacks are ramping up. Semiconductor facilities can present an attractive target to foreign hacking groups as successful breaches can have widespread effects. These attacks can delay shipments, leak intellectual property, and cause production downtime. Worse, these attacks can lead to infected products hurting a company’s brand.
- Excess Inventory: The bullwhip from shortage to months of excess electronic component inventory has contributed to rising concerns about obsolescence risks. Electronic components have expiration dates, and improper storage can make components deteriorate faster. Long lead times, like those seen during the shortage, can pose a risk for holding inventory, as by the time the market rebounds, the stock could be useless.
- Labor Shortages: From a lack of a sufficient talent pool to labor strikes, the semiconductor industry is grappling with an increased demand for higher production capacity but insufficient staff to accomplish this goal. Most countries working to improve domestic semiconductor manufacturing resiliency are also grappling with labor shortages, with an additional one million workers needed by 2030.
These challenges can cause massive disruptions in the fragile, global supply chain. Each delay can result in far-reaching ripple effects that vary in severity, especially with how interconnected the semiconductor supply chain is.
It’s not all bad news. These challenges are the perfect catalyst for continued digitalization and the use of artificial intelligence applications within the semiconductor industry. Automation can help compensate for the lack of skilled labor on production lines while reducing human error and optimizing workflows to lessen a facility’s carbon footprint.
Better inventory management through market intelligence can prevent dramatic inventory overhang and help organizations avoid the possibility of obsolescence. Greater visibility can highlight risks related to geopolitical volatility, and diversification can fortify supply chains against disruptions.
Increased attention to digital tools and digitalization can improve an organization’s cyber security, preventing successful breaches by ransomware. Lastly, continued collaboration between semiconductor suppliers and environmental agencies can help find new alternatives for the toxic chemicals needed in manufacturing.
Countries continue to work toward semiconductor manufacturing resiliency. Since the global semiconductor shortage, new programs and plants have made headlines over the last year. In the coming months, subsidies to support semiconductor manufacturing will be awarded to selected chipmakers.
According to reports, the Biden Administration will soon announce which chipmakers will be receiving funding under the CHIPS Act. Since its creation and passage, the CHIPS Act has garnered the U.S. over $200 billion in private investments toward semiconductor manufacturing. Financial aid from both federal and state governments has continued to captivate interest among industry competitors.
A side effect of new tax breaks and other monetary benefits has contributed to a rise in collaborative partnerships among semiconductor manufacturers. The benefits of these strategic partnerships are invaluable. They could prove essential in overcoming one of the biggest challenges for the electronic components industry in the coming decade: the labor shortage.
Reports Indicate Semiconductor Projects Could Receive Billions in Subsidies
The Biden Administration will soon award billions in subsidies to top semiconductor manufacturing companies. The Wall Street Journal reports that executives expect this announcement to occur during President Biden’s State of the Union Address on March 7th. The likely recipients of these subsidies include Intel and TSMC, which will help these companies complete their latest U.S. semiconductor facilities.
Intel has projects underway in Arizona, Ohio, New Mexico, and Oregon that will cost more than $43.5 billion. TSMC has two plants in Arizona with a total investment of $40 billion, announced last year. Samsung Electronics, which could also be among the recipients, has a $17.3 billion project in Texas. The Wall Street Journal also expects Micron Technology, Texas Instruments, and GlobalFoundries to be among the top contenders for the subsidies.
It’s predicted that the announcements will kick-start the manufacturing of advanced semiconductors for smartphones, artificial intelligence (AI) applications, and weapons systems. This comes after remarks from U.S. Commerce Secretary Gina Raimondo in December, stating that she would make around a dozen funding awards for semiconductor chips within the following year. This will reshape U.S. chip production.
The first award was also made public in December, with over $35 million in funding to a BAE Systems facility in New Hampshire. The facility is said to produce chips utilized in fighter planes and is part of the $39 billion “Chips for America” subsidy program.
While these economic initiatives are an added boon to the United States, lingering concerns remain due to the ongoing labor shortage. Despite investments, some experts worry that semiconductor manufacturing production will take years to meet goals because of the need for more skilled labor.
Universities and semiconductor companies are working together to create educational pipelines, but it could take years, not months, to properly establish. There is hope that these new subsidies will help propel educational efforts to grow the talent pool faster. If not, AI’s growing improvements can help bridge the gap left by the labor shortage.
Intel and UMC Collaborate for High-Growth Markets
Intel and the United Microelectronics Corporation (UMC) recently announced their collaboration to develop a 12nm semiconductor process platform. The new process will address high-growth markets like mobile, networking, and communication infrastructure. This long-term agreement will leverage Intel’s at-scale U.S. manufacturing capacity and experience alongside UMC’s extensive foundry capabilities on mature nodes to enable an expanded portfolio.
To begin, the partnership will focus on Intel’s U.S.-based high-volume manufacturing capacity and experience with FinFET transistor design. This design offers a potent combination of performance and power efficiency, giving Intel and UMC an excellent springboard for creating a product that satisfies customer demand. Intel and UMC will cooperate on design enablement to support the 12nm process through electronic design automation and intellectual property solutions from other ecosystem partners.
The project is forecasted to begin operations in 2027, with the 12nm process being developed and manufactured in Fabs 12, 22, and 32 at Intel’s Ocotillo Technology Fabrication site in Arizona. Intel and UMC will significantly reduce the investment needed to create this new process by using existing equipment.
Since the announcement, Intel and Taiwan have shared positive statements regarding each other’s strengths.
“Taiwan has been a critical part of the Asian and global semiconductor and broader technology ecosystem for decades, and Intel is committed to collaborating with innovative companies in Taiwan, such as UMC, to help better serve global customers,” said Stuart Pann, Intel Senior Vice President and General Manager of Intel Foundry Services. “Intel’s strategic collaboration with UMC further demonstrates our commitment to delivering technology and manufacturing innovation across the global semiconductor supply chain and is another important step toward our goal of becoming the world’s second-largest foundry by 2030.”
Jason Wang, UMC Co-President, said, “Our collaboration with Intel on a U.S.-manufactured 12 nm process with FinFET capabilities is a step forward in advancing our strategy of pursuing cost-efficient capacity expansion and technology node advancement in continuing our commitment to customers. This effort will enable our customers to smoothly migrate to this critical new node, and also benefit from the resiliency of an added Western footprint. We are excited for this strategic collaboration with Intel, which broadens our addressable market and significantly accelerates our development roadmap, leveraging the complementary strengths of both companies.”
The landmark collaboration is a win-win for both companies as UMC can agilely leverage the FinFET capacity without the added pressure of costly capital investments. For Intel, this collaboration is expected to aid the company in its goal of shifting from an integrated device manufacturer (IDM) to a foundry business model. Both companies, especially UMC, will be able to solidify their standing among fierce competition in their respective markets.
Trendforce notes that if the partnership succeeds, Intel may consider “co-managing additional 1Xnm FinFET facilities with UMC, potentially expanding to sites like Ireland’s Fab24 and Oregon’s D1B/D1C.” However, it goes without mentioning that the journey to its tentative production start in 2027 is not without challenges. Intel continues to face hurdles as it attempts to break into the foundry industry, and UMC’s 14nm process has been in development since 2017.
Hopefully, Intel and UMC can strengthen each other's weak points with their individual experience and expertise. It will be interesting to see what develops in this new alliance in the coming years.
Ongoing coverage over the recent price hikes to DRAM contract prices is now shifting to NAND Flash. Prices are going up and up after a low year for the memory market, marked by significant capital losses for manufacturers and strategic production cuts.
New data by TrendForce reveals aggressive price hikes by suppliers for several reasons ranging from market outlook uncertainty to manufacturers offsetting losses. The semiconductor industry will likely see many different production strategies among various suppliers for 2024. Expectations are still optimistic for a market rebound in the latter half of the year.
Meanwhile, countries continue to push for domestic semiconductor manufacturing resiliency in the new year. With recent government funding aid in the United States and facility construction commencing in Europe, Chinese companies are working overtime to boost domestic capabilities. Over the last few weeks, China’s production capacity is expected to boom over the next few years as its imports of semiconductor equipment reach billions.
NAND Contract Prices See Growth Echoing DRAM in 1Q24
In true cyclical fashion, the NAND Flash sector is following behind DRAM. Earlier this year, TrendForce reported that DRAM contract prices were forecasted to rise 13-18% in 1Q24. In 4Q23, there were already expectations that prices across DRAM and NAND would see hikes after a year of drops.
Due to uncertainty about DRAM's market outlook, memory manufacturers are increasing prices as buyers make bold purchasing decisions. Unfilled DDR5 orders are fueling this uncertainty with suppliers continuing production cuts to keep the supply-demand tight. Buyers are stockpiling DRAM components in some industry sectors, such as graphics and consumer.
TrendForce's most recent report reveals that even though the market faces a traditional low-demand season, buyers are increasing their purchases to establish safe inventory levels. The NAND Flash market will see an estimated 15-20% increase during 1Q24.
According to TrendForce, “With demand struggling to keep pace with these rapid increases, future price escalations hinge on the resurgence of enterprise SSD procurement. The first quarter of 2024 will see varied production strategies among suppliers, with some ramping up output early. This could lead to added pressure if anticipated demand growth falls short, moderating price hikes in 2H24.”
So far, enterprise SSD has not seen any demand spikes in the North American market. Lately, Chinese CSPs and server brands have filled the gap NA CSPs left behind. This keeps the market buoyant and encourages buyers to beef up their orders, helping push SSD contract prices to a high of 18-23% during 1Q24.
Meanwhile, eMMC and UFS products are witnessing bold price hikes in both sectors as suppliers' persistent production cuts have greatly limited production capacity. Because of the limited capacity, buyers must accept the price increases to prevent shortages. This is especially true for the UFS market, where critically low smartphone client inventories lead to UFS 4.0 being highly coveted.
Stabilizing smartphone and Chromebook demand is the primary driver of buyers' stockpiling and suppliers' bold pricing strategy. Some UFS series products are even witnessing price jumps of over 30%, with a forecast of a high of 18-23% rise. NAND Flash will see a moderate increase in 1Q24 at 8-13% from diminished buyer enthusiasm.
Uncertainty plagues DRAM and NAND Flash markets, with buyers and suppliers making ambitious moves while observing how the market plays out. Keeping a watch on pricing over the year's first half as excess inventory mitigation continues will be pertinent for buyers and suppliers to prepare for the forecasted market rebound in 2H24.
China Continues Pushing Toward Self Resiliency
Analysts at Barclays have indicated in a recent report that China’s semiconductor manufacturing capacity is expected to double in the next 5 to 7 years. This analysis of 48 chip manufacturers with production facilities in China surpasses previous market expectations significantly. Research suggests that 60% of the expected additional capacity to meet these goals will be able to come online within the next three years.
Currently, 22 wafer fabs are under construction in China. Ten more are planned to be constructed by companies like SMIC, Nexchip, CXMT, and Silan in the coming months. All of them are expected to come online by the end of 2024.
Chinese firms have also accelerated the procurement of crucial chip manufacturing equipment to ensure the timely completion of these facilities. The import value of lithography equipment from the Netherlands, essential in producing advanced semiconductors, has surged by 1050%. It has reached nearly $40 billion, according to Bloomberg. The most recent surge came in December as firms rushed to buy ahead before Dutch restrictions on lithography equipment started.
To circumvent challenges by export controls, many firms are investing in the development of China-made semiconductor manufacturing equipment. Last year, sales revenue for China-based semiconductor manufacturing equipment surged significantly after restrictions passed.
Barclays analysts expect most of China’s production capacity will be held by mature semiconductors (28nm and above) over advanced semiconductors. Research suggests that China will likely lag in advanced semiconductor production for at least another decade. Most mature semiconductors are still used in household appliances and automotive systems, meaning these facilities will be a welcomed addition.
There is concern that the larger production capacity and focus on mature nodes could lead to a market oversupply, but Barclays research suggests it would take several years to reach that point. It would also depend on new trade restrictions, if any arise, and on component quality.
However, TrendForce notes that China’s mature process capacity, rising from 29% to 33% and driven by local production policies, could cause a flood of mature processes to enter the global market and trigger a price war. Such a move would take place around 2027. As China’s mature process capacities emerge in the coming years, prepare for this.
This year, the semiconductor, interconnect, and passive industries are poised for growth. Most analysts forecast a rebound to come fully in 2H24, with excess electronic component inventory remaining a challenge for the year's first half. Mitigation of excess inventory has been making progress, noticeable in 3Q23 and 4Q23, despite low sales. Artificial intelligence (AI) made its mark on the industry in 2023 and will continue to do so in 2024.
A surprise contender for one of the main drivers of market recovery in 2024 will come from the memory sector. Last year, memory giants experienced the steepest declines from the drop-off in consumer demand. DRAM contract prices are rising in 1Q24, fueled by production capacity uncertainty for unfulfilled orders of specific components.
As memory manufacturers strategize for the year, logistics companies face another few months, if not the year, of challenges from disruptions to the two most well-known waterways worldwide.
Expansion in Memory to Account for Undelivered DDR5 and HBM Stock
A recent report from TrendForce made one thing clear about the coming quarter: DRAM contract prices were rising. After months of price drops throughout 2023, there were clues of an upcoming price increase in 4Q23, according to Sourcengine’s Lead Time Report. There is likely an average increase of between 13%-18% across DRAM market segments.
Uncertainty in the market outlook is fueling the current price trend, with buyers in consumer and graphics DRAM making more aggressive purchase decisions to stockpile components.
Unfulfilled DDR5 orders generate market unpredictability, feeding into the recent DRAM surge. Similarly, the industry is pivoting toward DDR5 over DDR4 and is poised to surpass DDR4 in the ongoing pricing rally.
December is the starting point of the memory rebound, with recovery evident in revenue growth among Taiwanese companies such as Macronix, Nanya Technology, and Transcend. Despite the growth, some memory manufacturers plan to continue with production cuts to maintain the supply-demand balance throughout the year. SK Hynix is one such company that plans on continuing production cuts but reducing the scale in 1Q24. The company plans to make similar adjustments to NAND Flash production in 2Q24 and 3Q24.
Over the next few weeks, SK Hynix’s expansion will focus primarily on DDR5 and high-bandwidth memory (HBM) products. Manufacturers believe that over 2024, manufacturers will expand penetration of HBM and DDR5 through each quarter. According to TrendForce, low-margin DDR4 capacity “will be crowded out.”
HBM products are vital in using cloud servers and artificial intelligence applications, such as generative AI and large language models (LLMs). SK Hynix, Samsung Electronics, and Micron Technology have been working toward new-generation HBM components for cloud services.
Logistics Complications Continue in Two Important Canals
Over the last year, disruptions from geopolitical volatility and the energy crisis have led to challenges in logistics. In the early days of the global semiconductor shortage and the Covid-19 pandemic, a peculiar problem arose when a container ship called the Ever Given ran aground in the Suez Canal. The obstruction lasted six days, holding up nearly $60 billion in trade and causing an estimated 60-day shipping delay on an already overtaxed global supply chain.
Now, the world is grappling with challenges impacting two of the world’s most important trade routes, the Panama and Suez Canal. These challenges have already caused a litany of headaches for shipping companies worldwide.
Panama is experiencing one of the worst droughts on record. To combat the ongoing drought, the Panama Canal Authority has placed restrictions on vessels passing through the Panama Canal due to the reduced water levels. Usually, the Panama Canal sees 36 ships pass through the route per day. This has been lowered to 22.
However, since the drought began in December 2023, the situation is improving. The El Niño oscillation in the Pacific Ocean brings a fair amount of warmer ocean temperatures, contributing to more storms and wetter weather. Panama might see a decent amount of rainfall this year, remedying the drought.
“Rainfall and lake levels have been higher than expected,” Air Cargo News reported on the situation in Panama. “Restrictions were set to become stricter with ships to be limited to 18 by February next year,” but due to higher water levels, the Panama Canal Authority will increase the transit number to 24 ships in January.
Conversely, the Suez Canal is grappling with the ongoing Red Sea Crisis. Due to the devastating Israel-Hamas conflict, tensions in the Red Sea from rebel attacks on passing ships have surged. The additional security risks to Suez Canal-bound transport ships have raised the insurance cost.
Reuters says navigation through the Suez Canal is flowing normally, with the canal authority watching the tension and analyzing the impact on shipping. The Suez Canal is used by roughly one-third of all global container ship cargo, and the redirecting of ships costs around $1 million extra in fuel.
The Panama Canal and Suez Canal lack an abundance of alternate routes. Those who avoid either waterway must go around Cape Horn in Chile or Cape Good Hope in South Africa. The increased fuel costs by taking alternate routes could translate to higher prices on some shipped components, similar to what the market experienced during the initial days of the war in Ukraine when Russia cut off natural gas to the European Union.
The effect might not be as dramatic as the energy crisis, but it will be imperative for companies to keep an eye on it in the coming weeks if these matters aren’t resolved. The semiconductor industry is poised for a rebound in the latter half of 2H24, but market demand may be muted and not enough to offset even small price increases. If the market recovery follows a more positive outlook, lead times could grow following shipping delays.
From a global perspective, the difference between new and old is a thin lens. In the chip world, the days of silicon’s dominance are coming to an end. Researchers believe a new form of graphene, merged with silicon carbide, could be the future of chipmaking and quantum computing. They recently unveiled the world’s first graphene semiconductor.
Meanwhile, the international supply chain is applying lessons learned from past disruptions to manage a new wave of uncertainty. As war rages in the Middle East and Ukraine and demand for goods fluctuates wildly, the supply chain is embracing diversity and risk mitigation to bolster confidence.
The World’s First Graphene Semiconductor Could Power Future Quantum Computers
The limits of using silicon to create semiconductors are rapidly approaching—perhaps sooner than many would like to admit. With this in mind, researchers are searching for the next material to carry the industry forward. Graphene is promising thanks to its superior conduction properties. However, its lack of a band gap has limited its ability to be used in chipmaking.
According to a new study published in the journal Nature, this problem might be solved. The solution comes from a source with which many in the industry are already familiar: silicon carbide (SiC).
The research team bonded a single-atom layer of graphene with a silicon carbide layer using their specialized heating and cooling process. Their result? An epitaxial graphene-based semiconductor that could revolutionize chipmaking in the coming years.
In this system, the SiC atoms “donate” electrons to the graphene molecules, which creates a functional band gap. This refers to the minimum amount of energy electrons need to move within a material to allow transistors to switch between “on” and “off” phases in a computer chip. Using SiC allowed the researchers to bypass the limitations of graphene and create the world’s first functional graphene semiconductor.
Since graphene moves electrons much faster than silicon, the chip can operate at terahertz frequencies. Such speeds are a massive leap compared to today’s silicon chips and promises big things for the industry.
Notably, the researchers believe their approach can be easily integrated into existing wafer manufacturing processes. Indeed, many chipmakers are already experimenting with SiC wafers given their increasing importance for clean energy and electric vehicles. With this in mind, it’s realistic to see how the industry could shift towards this new epitaxial graphene as a medium for advanced semiconductors without a revolutionary upgrade to existing processes.
Perhaps even more exciting is the material’s relevance for quantum computing. Today’s leaders in this segment are using a variety of methods to achieve their goals. There is plenty of disagreement on which technology will yield the greatest quantum results. Intel, for example, believes in a silicon-based solution that houses spin qubits inside. Meanwhile, Google, IBM, and Rigetti Computing are focused on a more mainstream superconducting model that requires extremely low temperatures but is tremendously powerful.
In an interview, lead researcher from the graphene chip study, Walt de Heer of the Georgia Institute of Technology, said, “Like light, electrons in graphene have quantum mechanical wave-like properties that can be accessed in devices, particularly at very low temperatures.”
The research team hopes to explore the quantum applications of their discovery in future studies. Though it remains speculative, de Heer says graphene-based chips could outperform superconducting technology. If true, this would be a massive breakthrough in the quantum computing world and could reshape how the industry evolves in the years ahead.
For now, graphene-based semiconductors are an exciting area to explore. More research and significantly more testing are required before mass production can be considered. However, the need for more advanced chipmaking materials is dire, and epitaxial graphene is a promising solution.
Diversity, Risk Mitigation Key for Combating Global Supply Chain Uncertainty Amid Turmoil
Globally, the supply chain is still working to recover from the effects of the COVID-19 pandemic. More recently, though, new factors have introduced further disruption. The conflicts between Russia and Ukraine, as well as Israel and Hamas, have a ripple effect on the supply chain. Ongoing fluctuations in demand also make it nearly impossible for logistics managers to predict future needs. As a result, there is an undeniable element of uncertainty right now as manufacturers, logistics operators, and consumers grapple with the influx of changes.
Fortunately, global risk diversification efforts are beginning to pay dividends. Following the onset of escalating tensions between the U.S. and China in 2018, chipmakers and players across numerous other industries have engaged in strategic realignments to protect their operations from uncertainty. Indeed, this trend of global diversification has been one of the biggest headlines in recent years as companies expand their operations in new locations and avoid regions ripe with conflict—economic or otherwise. This diversity has dampened the impact of military conflicts and gives all parties more flexibility for moving goods around the world on schedule.
With the chip industry at the center of rising global demand for components to power the AI revolution, Taiwan is handling a surge in business. The island’s Taiwan International Ports Corporation saw record-breaking highs in 2023 with even more volume projected for 2024. This comes partially thanks to the opening of its newest terminal, which can berth four container ships at a time.
In Taiwan’s port and others worldwide, lessons learned during the pandemic are improving efficiency and preventing congestion. This includes the increased use of automation and better global planning and operating rules from leading logistics corporations.
Of course, waning demand for goods also plays a role. According to data from several international institutions cited by DigiTimes Asia, global trade volume growth currently lags behind maritime capacity expansion. Today, the supply chain is better equipped to keep up than in years past.
Alongside this, the latest PMI for the U.S. manufacturing and services sectors points to sustained activity moving forward with no more declines on the horizon. While the same isn’t true for the EU, consumer confidence is rising in the region.
The global supply chain will need much more time to forget about the massive disruptions that shook the industry to its core in 2020. In reality, it’s best not to forget those trials or the lessons learned as a result. Even so, the uncertainty will linger as a barrage of political and economic tensions continues to put pressure on manufacturing and logistics.
Fortunately, the global supply chain is now more diverse and robust than ever. There’s no sign that the efforts behind this trend will slow down in the coming years. As companies worldwide emphasize operational security, global collaboration will be increasingly essential, and the broader supply chain will benefit.
Japan’s semiconductor industry appears to have dodged a significant blow following a deadly 7.6 magnitude earthquake as early reports reveal minimal damage to key facilities in the affected region. Experts believe the current temporary shutdowns for inspections will be resolved quickly and operations will resume shortly.
Meanwhile, the memory market has finally turned the page on a tumultuous 2023 and is preparing for a more positive year. While spot prices for DRAM and NAND Flash components remained flat to close the year, analysts have high expectations for prices in the first quarter of 2024.
Japan Earthquake Temporarily Halts Production at Key Chip Plants, Damage Appears Minimal
Following a devastating 7.6 magnitude earthquake that rocked Japan on the first day of the new year, experts are still assessing the impact on the country’s chip industry. At least 94 people have been confirmed dead following the quake that centered in the Noto region of Japan’s Ishikawa Prefecture.
While there is no making up for the tragic loss of life, there is a glimmer of hope for a fast economic recovery. According to a report from TrendForce, the recent earthquake does not appear to have significantly damaged any of the several key chip-related facilities in the affected region.
Ishikawa is home to several important fabs and raw wafer production plants. While these sites are shut down temporarily for authorities to thoroughly investigate each of them for damage, preliminary reports appear positive.
Most of the factories weren’t located near the quake's epicenter and only experienced seismic forces of levels 4 to 5—well within the tolerance of the affected structures. Fortunately, early inspections reveal no damage to essential chip machinery, which should prompt a speedy return to operation.
Shin-Etsu and GlobalWafers both operate plants in the area, which are currently shuttered for inspection. Due to the crystal growth needed for wafer production, seismic activity impacts these facilities more drastically. Fortunately for Shin-Etsu, most of its crystal growth operations are located in Fukushima, so the impact is limited, according to TrendForce.
Meanwhile, Toshiba’s facility in Kaga is also halted for inspection alongside TPSCo’s trifecta of factories in Uozu, Tonami, and Arai. The latter is a co-venture between Tower Semiconductor and Nuvoton. Taiyo Yuden, Murata, and TDK, key MLCC manufacturers in the region, also reported no significant damage.
While the humanitarian impact of Japan’s recent earthquake is still devastating, the country’s disciplined approach to earthquake preparedness prevented the disaster from being much worse. Ever since the brutal 1995 earthquake, Japan has invested heavily in infrastructure designed to be more resilient against strong seismic forces.
Interestingly, the country has also invested in drones to help aid in disaster relief. These robotic fliers can access hard-to-reach or dangerous areas much faster than human workers. Drones often assist in search-and-rescue operations to find people buried underneath the rubble and perform preliminary inspections of certain structures to ensure they are safe for human experts to enter.
As Japan continues its recovery, expect the local chip industry to be back on its feet sooner rather than later. Thanks to the country’s preparation and quick work from disaster recovery teams, the earthquake's economic impact should be minimal, and chipmakers in the area shouldn’t be significantly affected.
DRAM, NAND Flash Spot Prices Hover Due to Sluggish Year-End, But Increases Are on the Horizon
The memory chip market was up and down throughout 2023 and ended the year on a relatively flat note. Per new data from TrendForce, spot prices for both DRAM and NAND Flash components remained unremarkable thanks to the year-end and sluggish demand.
The report outlines that some DRAM suppliers are releasing more stock into the spot market to lower their inventories. This trend has put downward pressure on spot prices. However, buyers have been slow to increase their procurement quantities, keeping prices for the DRAM market in flux.
Meanwhile, NAND Flash spot prices have also remained flat thanks to the lack of driving forces to boost demand. TrendForce reports that purchasing stagnated at the end of the year. However, the NAND Flash market is in the midst of a substantial price correction thanks to restricted supply. Top memory chip makers, including Samsung, SK Hynix, and Micron, have all sustained sharp production cuts over the latter half of 2023 and into the new year to boost prices.
Despite the stagnation in memory spot prices, industry analysts have a sunny outlook on the memory market’s future for 2024. TrendForce data suggests DRAM prices will increase by 13-18% in the first quarter. This jump will largely come thanks to demand generated by AI applications. The market is also currently flooded with unfulfilled DDR5 orders as OEMs race to manufacture devices with the fastest memory chips available today.
Moreover, AI is also expected to spike demand for high-bandwidth memory (HBM) chips. These components play a crucial role in smartphones and other devices featuring on-device AI capabilities. As a result, mobile DRAM is expected to be the strongest category in Q1, with an expected price hike of 18-23%.
Chipmakers will likely need to start churning out more memory modules in 2024. Analyst Roko Kim told Financial Times, “DRAM chipmakers’ inventories are likely to fall short of appropriate levels by the end of the first quarter. Then, they will need to think about increasing output and plant operating ratios.”
Meanwhile, NAND Flash is poised for a surge of its own. TrendForce data projects a price increase of 13-18%. However, unlike the DRAM market, much of this increase is due to price hikes from manufacturers rather than organic demand.
Ultimately, the memory market is sure to have a better year in 2024 than it did in 2023. Between growth driven by AI and price increases as manufacturers seek to offset their recent losses, buyers shouldn’t expect to procure memory chips on the cheap this year—even if current spot prices are stagnant.
Chipmakers are pushing the limits of what’s possible with cutting-edge silicon. However, as buyers continue to demand more and more from the latest chips, creative design and manufacturing solutions are more critical than ever. For some, this means turning to collaboration and joint projects. Samsung and ASML’s recent partnership to build a new $760 million EUV facility in Seoul is a testament to this trend.
Meanwhile, a push for geographic diversification is sparking interest in chips and equipment from places outside traditional hubs. Japan aims to recapture a spot at the top of the industry amid this shift. The country’s hopes center on its darling startup Rapidus, which believes it can close a 20-year gap with the world’s leading chipmakers thanks to its ambitious roadmap and strategy.
Samsung, ASML Partner to Build New $760M EUV Facility in South Korea
Collaboration has recently been the name of the game in the chip industry as players look for new ways to gain an edge. ASML and Samsung recently unveiled plans to jointly build a $760 million advanced chip plant in the latter’s home of South Korea. The Seoul facility will utilize ASML’s next-gen extreme ultraviolet (EUV) equipment to produce high-end advanced semiconductors.
The partnership is part of a broader diplomatic effort between the two sides as they seek to strengthen their collaboration in the chip sector. It comes after South Korean President Yoon Suk-yeol recently visited the Netherlands on a four-day trip, including a tour of ASML’s headquarters alongside Dutch King Willem-Alexander. Yoon said during the visit that his government is committed to working alongside the Netherlands and will provide all necessary support for the project.
ASML has been working to expand its presence in South Korea beyond the four sites it already operates. Given the ongoing trade tensions between the U.S. and China, ASML is investing in geographic diversity with its latest expansion facilities.
Along with the new EUV plant in Seoul, ASML and Samsung are also working together to establish the Korea-Netherlands Advanced Semiconductor Academy. Aimed at addressing Dutch-based ASML’s severe labor shortage at home, the program will give South Korean students and chip workers opportunities for education and employment in the Netherlands.
In a statement, Yoon Suk-yeol’s office said, “The technological innovation led by ASML is becoming a powerful driving force of the Fourth Industrial Revolution around the world, and Dutch semiconductor companies such as ASML and ASM are building new facilities for production, R&D, and talent training in Korea.”
“This will mark a crucial turning point for the Korea-Netherlands semiconductor alliance,” the statement adds.
ASML’s EUV machines are critical for Samsung and other chipmakers wanting to stay at the leading edge of chip production. As the South Korean firm maintains its status as a global leader in memory chip production, EUV advancements will remain vital. With demand for high-performance DRAM chips being driven by applications such as AI and electric vehicles, the need will be even greater in the years to come.
However, Samsung isn’t the only South Korean firm joining hands with ASML. SK Hynix, the second-largest memory chip manufacturer, is working with ASML on hydrogen recycling technology for EUV equipment. The goal is to boost efficiency by reducing power usage and operating costs for EUV machinery.
According to DigiTimes Asia, successful commercialization of the technology could yield annual cost savings of nearly $13 million per EUV machine. That’s an exciting proposition for the chip sector given EUV’s ubiquity for advanced chip production.
The blossoming partnership between ASML and South Korea’s leading chip firms is an intriguing development to monitor. Close collaboration could yield results that significantly impact the wider industry and have widespread ramifications for chip production.
Rapidus is Confident Japan Can Close 20-Year Gap with Global Chip Leaders
Japan’s most exciting chip startup, Rapidus, has taken the industry by storm since it was first announced. With the lofty goal of churning out 2nm chips by 2027 and already planning for 1nm production, Rapidus has set expectations sky-high. However, there’s no denying that its rivals have a two-decade head start and much more experience behind them.
Even so, Rapidus believes it can quickly close this gap and catch up with the likes of TSMC and Intel. At the same time, the startup hopes to reinvigorate Japan’s semiconductor industry and return it to a place of prominence on the global stage.
During the recent Semicon Japan conference, Rapidus Chairman Tetsuro Higashi noted that his firm’s forthcoming plant in Hokkaido will “surely succeed.” Higashi points to several fast-paced shifts in the chip industry as the driving factors.
For one, the arrival of gate-all-around architecture is changing the way advanced semiconductors are designed and manufactured. Experts believe it will begin to replace traditional designs before this decade is over. Higashi notes this shift will unlock new chipmaking advancements which Rapidus intends to capitalize on with its 2nm production.
“Around 2027 or 2028, there will be a point in time where the pendulum for the trend of the technology will start to swing in a different way,” he says, referencing this change.
Higashi adds, “The chip market is moving more toward focusing on products with specific capabilities, instead of general use chips.”
The latter is another critical shift Rapidus is targeting as part of its plan to catch up with industry leaders and boost Japan’s domestic industry. Companies across sectors, but particularly those in tech, are aggressively pursuing in-house development projects. While the likes of Apple and Meta have stolen the spotlight, countless others have shifted their focus to chips tailor-made for their applications. This is a significant trend to monitor in the coming years as traditional chipmakers will need to adapt to meet the changing demands of buyers.
For Rapidus, this transition period over the next ten years or so presents an opportunity. The startup hopes it will be an ideal time to enter the market alongside more established players.
One way Rapidus aims to break in is by harnessing the full support of Japan’s chemical and fab machine suppliers. Though the country’s overall chip sector lags behind other industry hubs, these firms are already key suppliers for the world’s largest manufacturers, including TSMC and Samsung. This will make Japan’s return to global semiconductor relevance a bit easier.
However, the path forward isn’t without challenges. Japan’s chip industry has fallen far behind hubs like Taiwan, South Korea, and the U.S. in recent years. While Rapidus offers a breath of fresh air, it can’t revitalize the industry alone.
Higashi says, “Japan must at all costs create a platform where cutting-edge technology can be born.”
As Rapidus aims to close the gap between itself and global chip leaders, domestic support is essential. Fortunately, the Japanese government has fully bought into the project and has already pledged significant support for the startup. This includes hundreds of billions of yen for its Hokkaido headquarters and future chip research and expansion projects.
Over the coming years, Rapidus will be one of the hottest chip firms to watch as it seeks to achieve a seemingly impossible feat. Whether or not it will draw level with TSMC and Intel anytime soon remains to be seen. However, it seems inevitable that Rapidus will shake up the industry with its ambitious roadmap and play a major role in returning Japan to a place of prominence in the global chip landscape.
Quotengine: Your Ultimate BOM Tool
What’s Your Excess Worth?
The Last Integration You’ll Ever Need
Sourcengine’s Lead Time Report
Sourcengine’s Lead Time Report
Your Lead Time Overview
Subscribe HereStruggling with excess inventory?
Download your free copy!.png)
.svg)
.svg)
.svg){kind=icon}
.svg){kind=icon}