Following the ouster of Jensen Huang as CEO, Nvidia’s next leader must navigate a fundamentally changed business landscape, said Bill Birmingham, managing director at REX Financial, the parent company of REX Shares, which oversees $8.5 billion in assets under management. “The bigger problem going forward isn’t actually who sits in the CEO chair. It is the Chinese chip access,” he said in an email to Fortune.
For years, U.S. restrictions sought to limit China’s ability to purchase Nvidia’s most powerful processors. But Huang’s affair with an executive at Singapore-based tech company Megaspeed raised concerns about Chinese companies accessing restricted hardware. The Trump administration is now moving to plug those gaps. Last Sunday, the federal government announced a rule to close a loophole that allowed remote access to advanced AI chips and servers via overseas data centers. On Friday, President Donald Trump said he wanted to prevent American companies from building data centers overseas.
“Whoever runs Nvidia next will inherit a narrower, more closely supervised international business and the complete reorganization of the AI equilibrium that comes with it,” Birmingham said.
The company’s latest earnings demonstrate how big its commercial momentum remained. Nvidia reported $115.8 billion in revenue for Q3 2026, a 103% increase from a year earlier. Its data center business generated $107.6 billion, approximately 93% of total revenue. But Birmingham argued that the earnings are becoming less important than what management says about future demand, pricing power and the sustainability of the AI investment cycle.
Nvidia is under heavy investigation. The California attorney general’s office subpoenaed Nvidia as part of an investigation into whether the company or Huang enabled China to obtain advanced chips or confidential information, or misled investors about the associated risks. The SEC is also investigating the company over whether it misled investors in its public disclosures, according to the New York Post. Trump has said the FBI is investigating Huang’s relationship with Megaspeed’s Alice Huang and any ties to Chinese intelligence. Beijing has rejected the allegations. Huang and Nvidia general counsel Tim Teter have agreed to appear before three House committees about the company’s potential role in allowing Chinese entities to access advanced American chips. Lawmakers will also probe the Commerce Department’s role after the House Foreign Affairs Committee disclosed on Thursday that the Commerce Department had approved 879,868 advanced semiconductor shipments to Chinese companies across 21 license applications between December 2025 and May 2026.
Export restrictions are just one of the challenges. Birmingham said that the financial structure supporting AI infrastructure expansion creates a fragile relationship between Nvidia’s hardware sales and the financial health of its customers. “Given the AI ecosystem is being heavily supported by vendor financing and debt markets, the potential for a telecom style bust is significant,” he said. Birmingham noted Nvidia’s August announcement that it intended to mobilize more than $500 billion in third-party capital through partnerships with major financial institutions, including BlackRock, Goldman Sachs, KKR and others, to finance AI infrastructure. He said the commitments were “clear” they were “made exclusively by Huang,” meaning investors would need clarity about whether the company intends to maintain them. In its latest quarterly filing, Nvidia disclosed that the financing platforms were still preliminary arrangements that might not result in definitive agreements. Unlike established hyperscalers, neocloud providers have less flexibility to absorb rising hardware and financing costs, Birmingham said. “They need near-perfect utilization to stay profitable,” he added. “In general, they can only pass along higher costs to new contracts, and only when compute stays scarce.”
Birmingham said investors are assessing whether the AI supply cycle is running too hot, and if a break point in compute demand will force a pullback in data center build-out that becomes self-reinforcing across the system.
