Jensen Huang lauds China's AI models as Nvidia gears up to resume chip exports

Nvidia’s selective reopening of chip export channels has thrust Chinese AI demand back into the global spotlight. The move, which allows certain high-performance semiconductors to flow into mainland China under tightened restrictions, signals a recalibration rather than a full thaw. For months, the market had braced for a complete severance of access to Nvidia’s advanced processors, but the partial resumption suggests Washington is weighing commercial realities against security concerns. The mainland AI supply chain, long a linchpin of semiconductor geopolitics, now finds itself at the center of a policy tug-of-war that shows no signs of stabilizing. The immediate impact is tangible. Chinese cloud providers and AI startups, which had been stockpiling alternatives from domestic suppliers like Huawei, are now scrambling to recalibrate their procurement strategies. Nvidia’s H800 and B200 chips, though modified to comply with export controls, still offer a performance edge over local substitutes. Yet the reopening is not a blanket approval. Shipments are being vetted on a case-by-case basis, with end-user verification protocols that add weeks to delivery timelines. For companies racing to deploy large language models, that delay can mean the difference between market leadership and obsolescence. What a casual observer might miss is the deeper anxiety rippling through China’s AI ecosystem. The issue is less about any single shipment than about whether model developers can plan capacity with any confidence while policy rules keep shifting. A startup that invests in Nvidia-based infrastructure today could find its supply cut off tomorrow, stranding capital in hardware that cannot be repurposed for domestic chips. This uncertainty is already distorting investment decisions. Some firms are hedging by splitting orders between Nvidia and local suppliers, a strategy that increases costs and complicates software optimization. The selective reopening also exposes a paradox in U.S. export controls. By allowing some chips through, Washington acknowledges that a total ban would cripple Chinese AI development but also accelerate domestic substitution. Huawei’s Ascend 910B, for instance, has gained traction in data centers, and its successor is expected to narrow the gap further. The partial flow of Nvidia chips may actually slow that substitution, keeping Chinese developers tethered to foreign hardware while the U.S. retains leverage. It is a delicate balancing act, one that assumes Chinese firms will not simply hoard Nvidia chips as a hedge against future restrictions. On the ground in Shenzhen and Shanghai, the mood is pragmatic. Engineers are training models on whatever hardware is available, mixing Nvidia GPUs with domestic alternatives in hybrid clusters. The performance penalty is real but manageable for many applications. The real bottleneck is software ecosystem lock-in. Nvidia’s CUDA platform remains the gold standard for AI development, and migrating to domestic alternatives requires rewriting code and retraining teams. That transition is underway but will take years. The coming months will test whether the selective reopening is a genuine accommodation or a tactical pause. Chinese AI demand is not going away. If anything, the country’s push for sovereign AI capabilities is accelerating, driven by government mandates and corporate ambition. Nvidia’s partial return to the market may buy time for both sides, but it also raises the stakes. Every shipment that lands in China is a data point in a larger geopolitical calculus, one where the rules of engagement are being written in real time.

Jensen Huang lauds China's AI models as Nvidia gears up to resume chip exports CNBC

The move has implications for the global chip supply chain beyond the immediate players.

The development adds to a wider China semiconductors story in which companies are being judged on execution, capital access, regulatory fit and the credibility of their regional expansion plans.

For business readers, the important question is whether this becomes an isolated announcement or part of a more durable operating pattern across customers, financing channels, partners and public-market expectations.