Wednesday brought the most significant leadership change in AI hardware since the sector took shape. Jensen Huang resigned from Nvidia, six days before Anthropic is due to begin trading on Nasdaq. For operators with AI dependencies, both developments raise the same question: what changes when the person or structure that made something work is replaced by a new one?
Jensen Huang exits Nvidia
Nvidia's SEC 8-K filing, submitted Wednesday, confirmed that Jensen Huang resigned as chief executive and board director, effective immediately. CFO Colette Kress takes the interim chief executive role while also retaining her CFO duties. Harvey C. Jones was named the company's first permanent board chair. A formal search for a permanent successor is under way, with internal and external candidates both in scope.
Nvidia's formal announcement gave no reason for the departure. Reporting by PBS NewsHour and Variety ties the departure to an internal investigation, first disclosed on 9 November, into allegations that Huang had maintained a personal relationship with an individual described as an operative of a foreign intelligence service. Huang later acknowledged in an internal message to staff that he had been unfaithful to his wife; the company's formal announcement addressed neither claim. Senate Majority Leader Thune said Congress would conduct oversight of the matter regardless of the resignation.
For AI infrastructure teams, the product and roadmap implications are modest in the short term. Chip development is managed by deep engineering teams that report to Huang but do not depend on him for day-to-day operations. The gap is strategic. Huang negotiated long-term co-design agreements with hyperscalers, shaped the CUDA ecosystem that locked in the developer base, and possessed the institutional knowledge to anticipate demand curves one or two generations ahead. Colette Kress is a disciplined financial operator; she is not, by background, an infrastructure architect. Axios noted as recently as 9 November that Nvidia had no disclosed succession plan. Nvidia reports quarterly earnings on 17 November. That call will be the first measure of how the board intends to manage the transition.
Anthropic fixes its IPO price at $70-$78
Anthropic updated its S-1 filing to set an expected offering price of $70 to $78 per Class A share. The offering covers 825 million shares, with 801 million newly issued and roughly 24 million from existing holders; Anthropic retains all proceeds from the new shares. At the midpoint, the raise would total approximately $61 billion, valuing the company at around $1.2 trillion to $1.45 trillion. Pricing is expected on 18 November; trading under the ticker "ANT" on Nasdaq begins 19 November.
Alongside the public offer, Anthropic disclosed a $26 billion private placement at the IPO price to seven strategic investors: AMD, Broadcom, Google, Micron, Microsoft, Nvidia, and TSMC. The placement is for nonvoting shares. Its composition reflects that every significant chip and cloud platform is now a formal stakeholder on the demand side of the AI market.
The governance shift merits attention. Anthropic as a public company will face quarterly reporting obligations, analyst pressure on margins, and the possibility of activist shareholders. None of those pressures existed under a private benefit-corporation structure. Its product and pricing decisions will now be partly shaped by public-market expectations. Teams that have built on Claude should monitor whether the IPO's accountability structure accelerates or constrains Anthropic's investment in safety research versus revenue-generating capabilities.
Anthropic blocks access from China, Hong Kong, Macau, and Russia
Anthropic updated its Supported Regions Policy around 10 November, with the restrictions taking effect on 12 November. The prohibition bars use of Claude from mainland China, Hong Kong, Macau, and Russia, and extends to companies controlled by entities registered or operating from those regions, regardless of where the end-user terminal is located. All Anthropic surfaces are in scope: Claude.ai, Claude Code, Cowork, the API, and models accessed through third-party integrations authorised by Anthropic.
Enforcement carries real teeth. Patterns consistent with unauthorised resale or use from restricted regions will trigger identity checks requiring a government-issued document and a selfie. Failed checks can result in account termination.
The compliance obligation for enterprise buyers is upstream, not just at the terminal. If your organisation accesses Claude through a reseller, value-added reseller, or a third-party integration, verify that the intermediate party is not based in or controlled from one of the restricted regions. That obligation now rests with the customer.
Three developments, three structural shifts. The AI supply chain lost its most consequential individual on Wednesday. The lab most operators depend on will carry public-market accountability within a week. And the first formal access perimeter in the enterprise AI market is live. Each item is manageable on its own; together they define a quarter where operational assumptions about AI infrastructure deserve a careful audit.