
Japan Just Bet $16 Billion on a Chip Startup Nobody Has Heard Of. It Wants to Break the TSMC Monopoly.
Japan approved another $4 billion for Rapidus Corp, bringing total state investment to $16.3 billion. The target: 2nm chips by 2027.
The AI Post newsroom — delivering AI news at the speed of intelligence.
Japan just doubled down on the most audacious bet in the semiconductor industry. The government approved an additional 631.5 billion yen ($4 billion) in subsidies for Rapidus Corp., a chip startup that most people outside the semiconductor world have never heard of. Total state backing now stands at 2.6 trillion yen, or $16.3 billion.
The goal: manufacture 2-nanometer chips by 2027 at a new fabrication facility in Hokkaido. If Rapidus pulls it off, Japan will have a domestic source of the most advanced AI chips on Earth, breaking its dependence on TSMC in Taiwan and Samsung in South Korea.
If. That is the biggest "if" in the chip industry right now.
Why This Matters for AI
Right now, the entire AI industry depends on a single point of failure: TSMC in Taiwan. Nvidia designs the GPUs. But TSMC manufactures them. So do the chips from AMD, Apple, Qualcomm, and almost everyone else making advanced silicon. Taiwan sits 100 miles from mainland China. If Beijing ever moves on Taiwan, the AI industry does not have a plan B.
Rapidus is Japan's attempt to build one. The company was founded in 2022 with backing from Toyota, Sony, NEC, NTT, and other Japanese industrial giants. IBM is providing the process technology. The partnership is real. The question is whether $16 billion is enough to compete with TSMC, which spent $36.2 billion on capital expenditure in 2025 alone.
The Skeptics Have a Point
Rapidus has never manufactured a chip commercially. It is attempting to leapfrog from zero production to the most advanced node in the world. No company has ever done that. Intel, which has been making chips for 56 years, still cannot reliably produce at the leading edge. Samsung, with decades of experience, is struggling with its 3nm yields.
The Japanese government is not blind to this. The latest subsidy review focused heavily on execution milestones at the Hokkaido site. Tokyo wants proof that the 2027 manufacturing deadline remains on track. The fact that the money keeps flowing suggests Rapidus is hitting its marks. Or that Japan has decided this is too important to let fail.
The Global Chip Subsidy Arms Race
Japan is not alone. The United States passed the CHIPS Act to bring $52 billion in semiconductor investment home. The EU committed $47 billion through its own European Chips Act. South Korea pledged $470 billion in private and public investment over the next decade. China is spending even more, with Alibaba powering up 10,000 domestically made AI chips just last week.
But Japan brings something unique: it already dominates critical parts of the chip supply chain that nobody talks about. Japanese companies produce roughly 50% of the world's semiconductor materials and over 30% of the equipment used to make chips. Tokyo Electron, Shin-Etsu Chemical, and SUMCO are not household names, but without them, TSMC and Samsung cannot operate.
Rapidus is Japan betting that it can convert supply chain dominance into manufacturing dominance. The raw ingredients are there. The question is whether the recipe works.
What Happens If It Works
If Rapidus hits its 2027 target, the AI chip map gets redrawn overnight. Nvidia, AMD, and Qualcomm would have a second foundry option for their most advanced designs. The Taiwan risk premium drops. AI companies get more supply, potentially at lower cost. Japan becomes a strategic pillar of AI infrastructure alongside Taiwan and South Korea.
If it fails, Japan is out $16 billion and the world remains dependent on TSMC for every chip that matters. Given that Iran just moved the Middle East closer to a shipping crisis, and Trump just committed $10 billion to Microsoft infrastructure in Japan, the timing of this bet looks less like a gamble and more like a geopolitical insurance policy.
Sixteen billion dollars. A startup with zero commercial production. A deadline 11 months away. The most ambitious semiconductor bet since TSMC itself was founded in 1987. Japan is not messing around.