Huang quits Nvidia: what changes for GPU buyers, and what doesn't

Nvidia announced on November 12 that Jensen Huang had resigned as its president, chief executive and board member, effective immediately. The Register, citing the company's 8-K filing, reports the board accepted the resignation on November 11. Colette Kress, CFO since September 2013, is interim CEO and keeps the finance job. Harvey C. Jones, a former CEO and executive chairman of Synopsys, becomes the board's first-ever chair, and the board will look at internal and external candidates for a permanent CEO.

The exit follows two weeks of reporting on Huang's relationship with Alice Huang, a former executive at Singapore cloud company Megaspeed, an Nvidia customer. The New York Times reported US authorities concluded she was a Chinese intelligence asset; Reuters could not verify that and China denies it. Megaspeed denies illegally diverting chips to China. Nvidia shares fell 14.4% to $148.80 just after the open, according to The Hill, then recovered much of the loss; Investing.com showed them down 5.35% at 11:26 a.m. Eastern.

For an IT buyer the stock price matters less than three questions: will orders ship, will the roadmap hold, and does any of this reach a contract.

What Nvidia has promised

One concrete sentence. Jones said the board expects operations, customer commitments and the product roadmap "to continue without interruption." Kress added that the company's focus remains on serving customers. That is the extent of Nvidia's public commitment to customers so far. It has said nothing specific about GPU allocation, lead times, DGX or HGX delivery dates, cloud capacity deals or Nvidia AI Enterprise licensing.

One fund manager is not worried about the roadmap: Jeremiah Buckley of Janus Henderson told CNBC he thinks it is "solidified through 2028." The alternatives are thin in the short run too. Daniel Newman of the Futurum Group told CNBC that if every customer switched to AMD tomorrow, AMD could not supply them. CNBC says Anthropic, Meta, Microsoft, OpenAI and Oracle all run AMD's Helios racks, which compete with Nvidia's Vera Rubin racks.

What has already moved

Some buyers are not waiting. CRN reported on November 11 that an enterprise customer in a highly regulated industry told its systems integrator it intended to pause a purchase of Nvidia-based servers. Another partner told CRN that three of his major customers asked for calls about security, and he wants Nvidia to state that deployed systems are not at risk. Both partners spoke anonymously. Three other partners told CRN they were waiting to see how things played out, and one called the episode a non-event. Before the resignation, Nvidia's board said on November 8 it had "no reason to believe" national security or company secrets were compromised.

The risk that reaches contracts

The CEO change does not alter a purchase order. A rule the Commerce Department is reportedly finalizing could alter where your cloud GPUs sit. According to The Information, which cited two people familiar with the matter, as relayed by Investing.com, the rule would stop Chinese AI companies from remotely using advanced chips and would apply to seven more countries, including Malaysia, Singapore and Thailand. It could be announced this month but remains under interagency review. Renting chips housed in Singapore and Malaysia data centers to Chinese firms is legal under current rules, Fortune reports.

Reuters reported, citing three unnamed people, that Chinese AI firms are now trying to rent Nvidia capacity outside China, and some have contacted data center operators about reserving it. Smaller AI labs are considering approaching Southeast Asian providers of rented computing power, two of the people said. For a buyer of GPU capacity in those markets, that means possible competition for the same racks now, and possible compliance changes at your provider later.

What IT teams should do

• Ask your Nvidia partner or OEM for written confirmation of delivery dates on open orders. The board's sentence is a statement of intent, not a contract term.
• If security review is the worry, ask Nvidia through your account team for a statement on deployed systems. Partners are asking for the same thing.
• If you rent GPU capacity in Singapore, Malaysia or Thailand, ask the provider how a remote-access rule would affect it, and whether your contract lets them move or suspend capacity.
• Listen to the Q3 fiscal 2027 results on Tuesday, November 17, at about 1:20 p.m. PT, with the call at 2 p.m. PT. It is the first chance to hear management on supply under Kress.
• Do not cancel a roadmap commitment on today's news. Hold new capital decisions until after the results if timing allows.

Sources: Nvidia's announcement (8-K exhibit), CNBC, CRN on partners and customers, Reuters, Investing.com on the Commerce rule, Fortune on Megaspeed, The Register, The Hill. Related: AWS sold out of 2027 capacity: the options, Lambda's $4 billion round.