Jensen Huang has stepped down as head of NVIDIA, sending shockwaves through the company. The stock is down by almost a third from its top. The board placed the CFO in charge rather than an engineer. Two House chairmen want Huang in a closed-door interview. Reports indicate NVIDIA lacks sufficient insurance to protect the company from lawsuits. The new interim leader has clashed with Huang’s daughter. A trader placed roughly $8 million on a bet that NVIDIA shares would fall a month before the resignation; that position is now worth $196 million on paper.
Huang was the strongest public voice in the United States for open-weight AI models. NVIDIA had committed $26 billion over five years to build them, along with a $13 billion Hugging Face deal that has not yet closed. With the person who drove those efforts now gone, questions have arisen about the future of that open-model push. While some online commentary has speculated about events, a viral letter from an NVIDIA engineer threatening to leave for another project included at least one fabricated name.
Board Places CFO in Charge and Prioritizes Financial Risk
The board did not install an engineer or product leader as interim CEO. It put Colette Kress, the chief financial officer, in the role while she continues as CFO and the company searches for a permanent CEO. This choice signals a focus on money, risk management, and keeping investors calm.
A recent filing showed more than $530 billion in commitments and guarantees, up from about $184 billion the prior quarter. Much of that reflects supply commitments—the large volumes of wafers and memory that Huang had previously flagged as a concern. Reports state that Kress has been reducing the riskiest financing arrangements. One example is the OpenAI Ohio data center, where a draft backstop of $250 billion was cut to $120 billion. Smaller cloud companies that expected NVIDIA funding are now concerned those commitments may not materialize.
Some investors view the shift positively: fewer high-risk guarantees, greater attention to margins, and a cleaner balance sheet. At the same time, a finance-oriented interim leader is likely to examine costs closely. A $26 billion multi-year plan to develop free open-weight models is the type of expenditure a careful CFO would scrutinize. The board has said it will consider both internal and external candidates for the permanent CEO position.
Huang reportedly described Kress as “a deer in the headlights.” She had questioned aspects of his dealmaking, particularly customer financing arrangements. Huang was known for large, circular-style deals and handshake approaches; a CFO is expected to take a more cautious stance. Kress reportedly halved NVIDIA’s initial financial commitment to OpenAI, with the reduced deal signed on August 17. Huang had warned of enormous chip, wafer, and memory purchase obligations as major cloud providers planned to use alternative chips and reduce dependence on NVIDIA. The company paused some AI cloud credit support deals in late August, and financial support arrangements now face executive and board review.
Internal Power Shifts and Family Involvement
According to reports from The Information, internal structures began changing immediately after Huang’s resignation. Kress assumed the interim CEO role and has clashed with Huang’s daughter and son-in-law, both of whom remain at the company. The family succession arrangement that had been in place collapsed.
The son-in-law, Nico Caprese, a former member of Switzerland’s national ski team who joined NVIDIA in 2024, was promoted rapidly to vice president of global AI infrastructure growth. He had been in discussions with Neo clouds. The long-serving CFO repeatedly limited financial risk, stating that certain commitments were too large. Huang’s daughter Madison has seen her influence decline; she reportedly advised him during a recent scandal to replace the existing public-relations lead with a new expert.
For investors, the change reduces nepotism risk and places greater emphasis on financial discipline over expansive customer support. Reckless credit extensions and guarantees for AI startups and Neo clouds are expected to decline under more conservative management. Smaller Neo cloud companies that relied on prior arrangements, including those linked to Caprese, have expressed concern.
Open-Weight Models, the Hugging Face Deal, and Future Direction
Huang strongly supported open-source and open-weight AI, describing the ecosystem as beneficial for safety, innovation, diffusion, and sovereignty. As frontier labs moved to develop their own chips and reduce reliance on NVIDIA, supporting open models also served as a diversification strategy. With his departure, the continuation of that approach is uncertain. The $13 billion Hugging Face transaction has not closed. A representative from Hugging Face stated on X that there was no cause for worry, though the statement itself has been noted as a possible indicator of underlying concern.
NVIDIA had planned to spend $26 billion building open-weight AI models. Earlier in the year the company unveiled Numitron 3 Ultra, described as America’s smartest open-weights models and 30 percent cheaper to run. The flagship features 500 to 550 billion parameters. The model received mixed reactions and was viewed as somewhat underwhelming, yet Huang continued directing resources toward open models and the potential Hugging Face acquisition. Whether a finance-focused interim leader will maintain the same level of investment or adopt a more restrained approach remains an open question.
Investigations, Anthropic Links, and Related Claims
Most politicians who have called for investigations of Huang in recent weeks have extensive connections to Anthropic. NVIDIA is one of Anthropic’s major rivals. NVIDIA has advocated open-source approaches and freedom; Anthropic has favored closed-source models and government regulation. Anthropic is donating $20 million to Public First Action, an effort framed as bringing benefits for science and technology. Critics view the donation as a method of increasing pressure on open-weight models. Anthropic has previously stated that open-weight models are dangerous and should be subject to governance.
Public First Action has launched a $15 million conservative AI consensus project. The initiative includes more than $7 million in advertising featuring congressional leaders on AI topics, along with a national polling program on conservative Americans’ views of AI. Two congressional members have requested that Huang appear for a closed-door interview focused on ties involving China. Some commentary asserts that Anthropic coordinated a broader campaign against Huang and that Anthropic co-founder Jack Clark provided details about Huang’s private life to the New York Times. Clark is said to hold two million shares in Anthropic—two million more than Dario Amodei.
Beyond disagreements over chip exports to China, NVIDIA had been preparing to compete more directly. In June the company announced the Numitron 3 Ultra open-weight model with 550 billion parameters. In an interview four months earlier, Huang stated that Numitron could already compete with Claude Opus and was approaching the capabilities of frontier models. From Anthropic’s perspective, NVIDIA supplies the GPUs essential for training and running models while simultaneously developing its own large models. The AI industry has been compared to a five-layer cake, with NVIDIA positioned to participate across all layers.
Insurance Gaps and Broader Concerns
Reports indicate that NVIDIA failed to secure directors and officers (D&O) insurance. For a company of its size, the omission has been characterized as a significant oversight, leaving the firm exposed to potential lawsuits following Huang’s departure. Kress has continued to manage the interim role amid the internal clashes already noted.

The resignation raises practical questions for the open-weight community. The unfinished Hugging Face deal would have given that platform the backing of NVIDIA’s resources. Open-weight models are valued by many developers because they allow companies to run AI systems locally, keeping data sovereign rather than sending it to external cloud providers. With the strongest public advocate for that approach no longer at the helm, and with financial scrutiny increasing, the scale and pace of NVIDIA’s open-model efforts may change.
The board’s choice of a careful financial leader over a deal-making founder has produced clear short-term effects on commitments and risk appetite. Whether those changes ultimately strengthen the company’s balance sheet or slow its open-weight initiatives will depend on the permanent CEO selection and subsequent decisions. Investors, partners, and the broader AI community are watching closely to see which priorities prevail.




