Jiyoung Sohn, Asa Fitch
Samsung 005930 -0.94%decrease; red down pointing triangle Electronics plans to more than double its total semiconductor investment in Texas to roughly $44 billion, according to people familiar with the matter, a significant breakthrough in the U.S.’s quest to make more of the world’s cutting-edge chips.
The South Korean company’s new spending will be concentrated in Taylor, Texas, located just outside of Austin, where Samsung is building a semiconductor hub and has other nearby existing operations, the people said. The additions include a new chip-making factory, and a facility for advanced packaging and research and development.
Samsung is one of just three firms, along with
Intel and Taiwan Semiconductor Manufacturing, capable of producing advanced logic semiconductors vital to artificial intelligence and national defense. These companies sit at the heart of the Biden administration’s push to strengthen the U.S.’s chip-making capabilities, as Washington simultaneously seeks to undercut Beijing’s tech advances.
To help finance the broader Texas expansion, Samsung is expected to receive billions of dollars in subsidies from the U.S. Chips Act, the people said. Talks with the Commerce Department remain ongoing, though Samsung is expected to receive one of the largest payouts given to a single company.
An event to announce Samsung’s broadened investments is expected to be held on April 15 in Taylor, according to the people familiar with the matter. Samsung declined to comment. The Commerce Department declined to comment, saying that it is unable to discuss any specific company projects.
Samsung’s additional investments add to the $17 billion that the company had previously committed more than two years ago to Taylor for a cutting-edge chip-making plant.
The factory broke ground in 2022, with plans to start mass production as early as this year. The costs of building this first chip-making plant have increased due to inflation and other factors, requiring several billion dollars in extra investment, according to people familiar with the matter.
The second Taylor-based chip factory is expected to cost more than $20 billion, the people said. Samsung’s R&D efforts are expected to be warehoused inside those two plants. The size of the investments in these two factories could shift depending on market conditions.
The planned facility for advanced packaging—a key final step in the production of high-end AI chips like those made by Nvidia—will have a price tag of roughly $4 billion, the people said.
Earlier on Friday, Samsung said it expects a 10-fold increase in its first-quarter operating profit, to roughly $4.9 billion. This tops industry analysts’ estimates, as the chip industry awakens from a protracted downturn.
Samsung’s supersize chip investment in Texas adds fresh momentum to one of President Biden’s marquee domestic agendas as he seeks re-election in November. Many of the highest-profile projects have seen costs rise and face delays.
U.S. chip-making dominance has been a priority for Washington, which has earmarked tens of billions of dollars in subsidies to woo back local production that had migrated to Asia.
The Chips Act, which passed two years ago, set aside $53 billion in grants for projects like Samsung’s—and the money has only recently begun to flow.
Intel was awarded $8.5 billion for several chip plants planned in the U.S. last month, following a $1.5 billion grant to GlobalFoundries for projects in New York and Vermont in February. TSMC and Idaho-based Micron, a memory chip maker, are also expected to receive grants under the program.
The spree of projects aims to bolster domestic supplies of critical semiconductors. The U.S. share of chip-making declined to about 12% in 2020 from 37% in 1990, a fall that is increasingly seen as a national-security liability in an age when chips underpin advanced weapons, cyberwarfare and AI.
In a speech in February, Commerce Secretary Gina Raimondo said that based on the level of industry interest in Chips Act funding, the U.S. was on track to produce roughly 20% of the world’s most advanced logic chips by the end of this decade.
Samsung’s big bet in Texas is comparable to those made by its chief rivals. TSMC is building two chip-making factories in Arizona, with a projected investment of $40 billion. Intel’s total investment in U.S. projects in the next five years is expected to exceed $100 billion.
Samsung began semiconductor operations in Austin in 1996, starting with memory chips. The site later transitioned into the contract chip-manufacturing business. The Taylor chip making plants are expected to be filled with equipment for producing the industry’s cutting-edge logic chips for customers.
Two years ago, Samsung floated the prospect in filings made to the Texas controller’s office of potentially investing upward of $200 billion toward building 11 new chip-making plants in Texas over the next two decades.
Samsung, the world’s largest memory maker by revenue, is also racing alongside South Korea’s SK Hynix and Micron for leadership in high bandwidth memory, a critical component of artificial-intelligence computing.
HBM can speed up computing times by stacking multiple DRAM memory—used commonly for helping devices or servers multitask—on top of each other and merging them as one.
HBM has become the go-to type of memory to work in tandem with graphic-unit processors made by the likes of Nvidia that power AI computing. To enable faster data-processing speeds, the two types of chips are currently bundled together using “2.5-D” packaging techniques. Major chip makers including TSMC, Samsung and Intel are investing in 2.5-D packaging as well as next-generation 3-D packaging.
Earlier this week, South Korea’s SK Hynix announced its plans for a $3.9 billion facility in West Lafayette, Ind., for advanced chip-packaging, mainly for HBM. SK Hynix is the exclusive partner to Nvidia for the most-advanced HBM currently in the market.
Samsung’s planned advanced chip-packaging facility is expected to carry out packaging for HBM and provide 2.5-D and 3-D packaging technologies, according to people familiar with the matter.
At one of his company’s events last month for AI developers, Nvidia Chief Executive Jensen Huang said his company was in the process of determining whether Samsung’s next-generation “HBM3E” product could be a viable option. He stopped by Samsung’s booth and autographed one of the chips. He wrote: “Jensen approved.”
Write to Jiyoung Sohn at jiyoung.sohn@wsj.com and Asa Fitch at asa.fitch@wsj.com
No comments:
Post a Comment