Analysis: 75% of China's overseas computing power comes from Southeast Asia; if Jensen Huang is ousted, new computing power leasing may face further pressure.

November 12, according to The Wall Street Journal, Nvidia's board is preparing to oust CEO Jensen Huang and may meet as soon as today to decide his fate. Nvidia has not yet announced a decision. If the company further tightens GPU supply or customer vetting for Southeast Asian cloud providers after the leadership change, the first risk Chinese AI companies may face is not a disruption of existing computing power, but whether the next batch of high-end computing power can be rented at the originally planned scale, price, and timeline.

, Singapore-based cloud service provider Megaspeed deploys Nvidia GPUs in data centers in Malaysia, Indonesia, and other locations, allowing Chinese customers to rent computing power remotely without importing chips. A Bloomberg investigation shows that as of November 2025, Megaspeed had imported at least $4.6 billion worth of Nvidia equipment since its founding. The U.S. think tank Institute for Progress estimates that by the end of 2026, overseas AI computing power accessible remotely to Chinese companies will be roughly equivalent to 1.05 million H100 GPUs, with about 75% of the related data centers located in Southeast Asia.

This channel had already been facing tightening before the ouster news emerged. According to a November 3 report by the UK's Financial Times, Nvidia has verbally asked relevant cloud providers to sell at least 51% of their computing power services to non-Chinese customers, with the requirement applying to future contracts and not affecting already signed leasing agreements. Therefore, even if Southeast Asian cloud providers can still buy Nvidia chips, as long as business restrictions targeting Chinese customers are further tightened, the additional computing power Chinese companies can rent may also decrease. On the other hand, if cloud providers receive reduced GPU quotas or face delayed deliveries, the launch of new computing power may also be postponed.

The risks are more concentrated in newly signed orders, as well as renewals and capacity expansions that require re-contracting. Companies relying on a single overseas cloud provider may be unable to expand their clusters as originally planned even if they can continue using existing computing power; if the number of cloud providers that meet supply conditions and are willing to take orders decreases, companies may also face higher quotes and longer delivery cycles.