Nvidia just made a historic decision: halting production of its H200 chips destined for the Chinese market and redirecting all manufacturing capacity at TSMC toward Vera Rubin, its next-generation chip. According to the Financial Times, the company already has 250,000 H200 chips in inventory and doesn't expect significant sales to China in the near term. After following the chip war for years, this move marks a point of no return.
What exactly happened
Although the Trump administration approved H200 sales to China earlier in 2026, shipments remained stalled due to a double blockade:
- US side: Export control regulations with "guardrails" that slow every shipment
- China side: Beijing hadn't formally approved imports and was considering requiring domestic chip purchases alongside any H200 imports
Nvidia CFO Colette Kress confirmed it on the latest earnings call: "While small amounts of H200 products for China-based customers were approved by the US government, we have yet to generate any revenue, and we do not know whether any imports will be allowed into China."
The numbers that matter
| Metric | Detail |
|---|---|
| Orders from Chinese companies | 2+ million chips for 2026 |
| Nvidia's current inventory | ~250,000 H200 chips |
| Price per H200 chip | ~$27,000 |
| ByteDance's planned investment | $13.8 billion in Nvidia chips |
| Combined CapEx top 5 cloud | $700 billion in 2026 |
| Vera Rubin vs chips allowed in China | 22x more powerful |
Why Nvidia prefers Vera Rubin
The logic is simple: if China isn't buying H200s, that TSMC manufacturing capacity is worth more producing Vera Rubin, which already has confirmed buyers:
- Microsoft: Deploying Vera Rubin NVL72 rack-scale systems in next-gen data centers
- CoreWeave, AWS, and Google Cloud: Deployments scheduled for H2 2026
- Global AI demand: Hyperscalers are investing $700 billion this year
Vera Rubin is not available to Chinese buyers under current export rules and delivers 22x the performance of chips Nvidia is permitted to sell in China. Each new generation widens the technology gap. In my experience analyzing the semiconductor market, this is the largest divergence I've seen between what China can access and what the rest of the world is deploying.
How this affects you
- If you invest in Nvidia (NVDA): The stock dipped on the news but recovered as analysts recognized Vera Rubin has higher margins. This is seen as net positive medium-term.
- If you use AI services: Models trained on latest-gen chips (Vera Rubin) will be faster and cheaper. Microsoft, Google, and AWS will have superior hardware in H2 2026.
- If you work in tech in China: Chinese companies will have to rely on domestic alternatives like Huawei Ascend, which lag significantly in performance.
What could change
Analysts are monitoring a scheduled meeting between Trump and Xi Jinping in late March. If they reach a diplomatic agreement, Nvidia could restart H200 production in approximately 3 months. However, I've been following these negotiations for a while and every time progress seems likely, new restrictions appear.
Common issues
Will Nvidia lose money from this?
Short-term, it loses potential China revenue (~$54B in orders). Long-term, Vera Rubin has better margins and guaranteed demand from Western hyperscalers. Wall Street sees this as a smart pivot.
What alternatives does China have?
Huawei Ascend 910B/C is the primary domestic alternative, but it performs significantly less than the H200. Chinese companies like ByteDance and Alibaba are developing their own chips but are years away from matching Nvidia.
Should I buy or sell Nvidia stock?
We don't provide financial advice, but the analyst consensus maintains NVDA as a "buy" with an average price target of $180+. AI demand remains the largest market driver.
