The $5 Trillion AI Risk Sitting in the Taiwan Strait

What’s the biggest risk in the AI boom that almost nobody is pricing in?

Nvidia (NASDAQ:NVDA) depends almost entirely on a single foundry, TSMC, for its most advanced and profitable chips. Those chips are produced mainly in Taiwan, a geopolitically exposed location.

Image by foto qin from Pixabay

What makes this so dangerous?

Nvidia is currently valued at $4.3 trillion – having briefly touched around $5 trillion earlier this year – yet the H100, H200, and Blackwell lines all rely on TSMC’s leading-edge facilities in Taiwan. The island produces over 90% of the world’s cutting-edge chips, and 2025 has brought a spike in Chinese military drills, pressure, and uncertainty — making that single point of failure more acute.

Portfolio beats stock-picking every time. Consider what could long-term performance for your portfolio be if you combined 10% commodities, 10% gold, and 2% crypto with equities.

Let that sink in: the entire AI boom depends on one spot on the map.

Isn’t this everyone’s problem, not just Nvidia’s?

Sure is. AMD, Qualcomm, Broadcom, and nearly every fabless leader rely on the same supply base, although Nvidia’s focus on the highest end-chips makes it more dependent on the leading edge facilities in Taiwan compared to other players who are more spread across older nodes, alternative foundries, and less packaging-intensive products. Overall, this is the single point of failure behind the entire Magnificent 7 AI build-out.

How Dependent Is TSMC on Taiwan?

Over 90% of TSMC’s wafer capacity is still in Taiwan.  Taiwan hosts over 20 fabs including advanced ones like Fab 18 in Tainan with more than 1 million wafers/month capacity and state-of-the-art 5nm and 3nm processes.  The company is expanding in the U.S., Japan, and Germany, but its most advanced nodes — the ones Nvidia needs — stay at home.  The inertia is massive, largely because rebuilding an ecosystem this advanced outside Taiwan is incredibly hard.  TSMC has pledged roughly $160 billion for U.S. fabs, but ramping advanced nodes on foreign soil is a years-long process. Yield, volume, engineering talent — none scale instantly. Moreover, Ironically, the U.S CHIPS Act is actually expected to increase near-term dependence because export restrictions force more cutting-edge work to remain in Taiwan.

How Exposed Is Nvidia?

Nvidia sources 100% of its highest-end GPUs from TSMC. There is no second source for 3-nm or 2-nm class production at scale until 2027 at the earliest. And the packaging story is just as concentrated. Chip-on-Wafer-on-Substrate, the advanced packaging that makes these GPUs possible, is overwhelmingly located in Taiwan as well. TSMC’s CoWoS (Chip-on-Wafer-on-Substrate) packaging is essential for Nvidia’s top-end AI chips, and the capacity for it is entirely in Taiwan. It provides the performance and bandwidth these high-end GPUs need, which is why Nvidia depends on it so heavily.  It’s concentration in its purest form.

Geopolitical Risk: Still Barely Priced In

Tensions in the Taiwan Strait are the highest in decades. 2025 has already seen more frequent and larger-scale military drills, sharper diplomatic pressure, and rising cross-strait friction. In the worst case, China doesn’t need a full invasion — even a limited blockade could halt TSMC exports overnight. A direct conflict would shut down over 90% of the world’s leading-edge chip output instantly, freezing global AI compute and crippling the entire tech supply chain.

If TSMC goes dark for some time, who steps in?

• Samsung is roughly two years behind on leading-edge logic and struggles with the yield consistency that Nvidia requires.

• Intel is still trying to stabilize its latest nodes and is nowhere near TSMC’s N2 scale.

• China’s SMIC is capped by sanctions at 7 nm with no CoWoS alternative. And given the Taiwan–China dynamics, Nvidia using a mainland foundry is likely off the table.

• Global supply of AI training chips would fall by 80% plus overnight.

There is no redundant global system for advanced-node manufacturing.

Valuation Math: What Happens To Nvidia Stock In A Disruption?

Nvidia trades around 43x forward earnings, a multiple that assumes smooth, risk-free supply.

Here’s the math if things break:

• Nvidia is on track for about $300 billion in revenue next fiscal year.

• A six-month disruption could cut that revenue in half to roughly $150 billion.

• With net margins near 50%, that’s a $75 billion hit to earnings.

The impact to valuation is likely to be much more pronounced. Why? Because earnings don’t just get cut – the entire multiple compresses. A supply shock forces investors to re-price Nvidia as a company with very real geopolitical risk. A geopolitical premium could become a permanent feature of semiconductor valuations.

If the World De-Risks, Who Wins?

A global re-shoring wave would create new winners.

• Intel gets another chance because the world suddenly needs every credible fab it can get. Intel’s U.S. focus is a huge plus.

• Samsung benefits from the urgency even if it’s behind.

• ASML, Applied Materials a major chip fabrication suppliers wins regardless of geography – any new fab requires tools.

• U.S., Japan, and Europe become long-term beneficiaries of supply-chain diversification.

The Trefis High Quality (HQ) Portfolio, with a collection of 30 stocks, has a track record of comfortably outperforming its benchmark that includes all three – the S&P 500, S&P mid-cap, and Russell 2000 indices. Why is that? As a group, HQ Portfolio stocks provided better returns with less risk versus the benchmark index; less of a roller-coaster ride, as evident in HQ Portfolio performance metrics.

Invest with Trefis Market-Beating Portfolios

See all Trefis Price Estimates