The AI semiconductor race is changing. For years, TSMC, Samsung Electronics, Intel and SK hynix competed primarily through technology, manufacturing capacity and capital investment. But on August 22 last year, the U.S. government agreed to invest $8.9 billion in Intel for roughly a 9.9% stake, moving American semiconductor policy beyond subsidies to direct ownership in a strategic chipmaker.
Intel has also secured roughly $7 billion in private investment from Nvidia and SoftBank as part of its turnaround. After reporting a $18.8 billion net loss in 2024, the company moved to cut around 20,000 jobs, sell stakes in non-core assets including Altera and Mobileye, and refocus on advanced process technology and foundry operations. Intel’s recovery has therefore become a test of whether state backing can help rebuild a globally competitive U.S. manufacturing base.
Washington Moves From Subsidies to Equity
Before taking an equity stake, Washington had committed up to $7.86 billion in direct CHIPS Act funding for Intel facilities in Arizona, New Mexico, Oregon and Ohio. The equity investment marks a deeper intervention: advanced semiconductor capacity is increasingly being treated as both industrial infrastructure and a national security asset.
Sujai Shivakumar, Charles Wessner and other experts at the Center for Strategic and International Studies have argued that industrial policy has effectively become standard practice across advanced economies. Leading-edge chipmaking now requires such large amounts of capital and carries such high technological risk that maintaining domestic production capacity is increasingly difficult for companies to shoulder alone.
Government backing, however, cannot substitute for commercial performance. Intel CEO Lip-Bu Tan has stressed that there will be “no more blank checks” for its next-generation 14A process, signaling that further investment must be supported by customer commitments and economic viability.
Intel’s 18A process entered production in 2025, while 18A-P has moved into risk production this year. Intel says 18A delivers up to 18% better performance at the same power, 38% lower power at the same performance and roughly 30% greater chip density than Intel 3. The real test is whether 18A and 14A can attract external customers while achieving competitive yields and costs.
Samsung Builds on Two Fronts
Samsung Electronics is following a different model. Rather than receiving direct Korean government equity investment, the company is expanding manufacturing in both Korea and the United States, backed by public incentives and infrastructure support.
Samsung plans to invest more than $37 billion around Taylor and Austin, Texas, with up to $4.745 billion in direct U.S. CHIPS Act funding. Yet Korea remains the center of its manufacturing strategy. Under its 2026–2040 investment plan, Samsung expects to invest 2,450 trillion won overall, including roughly 2,100 trillion won in semiconductors, with about 1,650 trillion won directed toward the Yongin cluster and other domestic semiconductor hubs.
Its technological advantage may increasingly depend on integrating memory and foundry capabilities. Samsung began shipping samples of 12-layer HBM4E in May, combining 1c DRAM with a logic base die produced on a 4-nanometer foundry process. As HBM evolves into a more integrated component of AI computing, Samsung’s ability to combine DRAM, HBM and logic manufacturing could become strategically more important.
SK hynix Converts HBM Leadership Into Scale
SK hynix is taking a more focused route: turning its leadership in AI memory into manufacturing scale. The company was estimated to hold about 58% of the global HBM market in the first quarter, while SK Group Chairman Chey Tae-won said in June that SK hynix plans to double its overall wafer production capacity over the next five years.
That strategy gained another 54.3 trillion won on August 7, when the board approved 35.2 trillion won for the Yongin Y2 fab and 19.1 trillion won for the Cheongju M17 fab. Y2 will produce advanced DRAM, including HBM and targets its first cleanroom operation in June 2029, while M17 is scheduled to begin NAND cleanroom operations in December 2028.
The company’s longer-term investment plan reaches approximately 1,100 trillion won across Yongin, Cheongju and a planned semiconductor base in southwestern Korea. It is also investing about $4 billion in an advanced HBM packaging and R&D facility in West Lafayette, Indiana, targeted to begin operations in the second half of 2028.
Manufacturing Scale Becomes Strategic
The strategies differ, but they point to the same reality. Intel is using state and private capital to rebuild advanced U.S. foundry capacity; Samsung is expanding its memory and foundry footprint across Korea and the United States; and SK hynix is using large-scale investment to reinforce its HBM leadership.
In AI semiconductors, technological leadership alone is no longer enough. Suppliers must deliver HBM, logic chips and advanced packaging for millions of AI accelerators at the scale and timing customers require. The ability to convert technology into reliable manufacturing capacity is becoming a decisive source of market power.
But fabs alone do not create competitiveness. TSMC founder Morris Chang has warned that rebuilding semiconductor manufacturing in the United States could become “a very expensive exercise in futility.” His warning points to a deeper reality: semiconductor capacity depends on engineers, equipment and materials suppliers, power and water, customers, yield management and manufacturing know-how built over decades.
Korea Needs a National Semiconductor System
The U.S. and Korean models now diverge most clearly in the role of government. Washington has moved from CHIPS Act grants and tax incentives to export controls, national security policy and direct equity investment. Korea remains more corporate-led, with Samsung and SK hynix committing the capital while the government supports investment through taxation, infrastructure and regulation.
Each model carries risks. Direct state investment can distort markets, while Korea’s approach could leave power, water, talent and permitting unable to keep pace with corporate investment. The larger lesson is that AI semiconductor competitiveness can no longer be separated from national industrial capacity.
Intel’s revival therefore poses a broader question for Korea. Having world-class chipmakers is no longer enough. Korea must connect power, capital, talent, R&D and advanced packaging so that technological leadership can translate into manufacturing scale and durable market power.
The next phase of the AI semiconductor race will not be decided by process technology alone. It will be decided by which countries can turn corporate innovation, manufacturing capacity, infrastructure and national strategy into the most effective semiconductor ecosystem.
