Nvidia’s leadership change puts $279 billion in commitments at test

By Tom Whitfield

Nvidia entered Jensen Huang’s resignation with $279 billion in supply and capacity commitments. Board chair Harvey Jones announced on Nov. 12 that interim CEO Colette Kress would preserve “operations, customer commitments and product roadmap to continue without interruption,” although no major customer had publicly changed its orders.

What does Colette Kress inherit?

Nvidia generated $96.2 billion in the fiscal second quarter ended July 26, 2026, including $89.0 billion from data centers. Hyperscalers contributed $48.7 billion, more than double a year earlier, while AI clouds, industrial and enterprise customers contributed $40.3 billion, up 138%. That group includes AI-native and sovereign buyers; edge computing supplied $7.2 billion.

Customer concentration remains the central operating risk:

• One direct customer represented 16% of quarterly revenue.
• In the first half, three direct customers represented 16%, 15% and 13%.
• Five direct customers each held at least 10% of receivables at quarter-end.

Nvidia leaves them unnamed. Its direct customers include server makers and distributors that build systems for end users, so supplier relationships do not map one-to-one with chip operators.

Commitments rose to $279 billion from $119 billion one quarter earlier. The 10-Q warns Nvidia may be unable to reduce them if demand differs from expectations, making demand forecasting a commitment-management issue.

LSEG data cited by CNBC estimate October-quarter revenue of $109.2 billion. Nvidia reports Nov. 17, its first results without Huang, and Kress has called a company meeting for Nov. 18.

In a staff memo, Jones said the board has full confidence in Kress “to deliver on our commitments to shareholders, customers, partners.” Nvidia did not answer CNBC’s request for comment; none of its largest U.S. customers has commented publicly.

Continuity holds on paper, but Nvidia has exposed both demand upside and contractual rigidity.

How closely is Nvidia tied to its biggest buyers?

Huang managed customer relationships personally. Executives at customers, partners and portfolio companies told CNBC that he often emailed or called CEOs to offer help.

Two people familiar with the discussions said he pressed Oracle to buy more Nvidia networking hardware to secure Grace Blackwell allocation; another source said his sign-off was needed for key project-finance deals. Michael Dell called him a “visionary of the AI age.”

The largest such commitment concerns OpenAI:

• In August, Nvidia agreed to guarantee up to $105 billion in lease payments for about 4.25 gigawatts at SB Energy’s Pike County, Ohio, campus.
• An OpenAI affiliate will tenant the site. Each of nine completed phases activates part of the guarantee, with the first expected in Nvidia’s fiscal 2029.
• The site will host Nvidia infrastructure exclusively, apart from limited exceptions, and the guarantees end if OpenAI earns a satisfactory credit rating.

Nvidia also carried $47.9 billion of private-equity holdings at July’s end, up from $3.8 billion a year earlier after $31.0 billion of net first-half additions. The AP said Kress has led Nvidia’s partnership and investment strategy.

It holds nearly 13% of CoreWeave and agreed in September to buy Hugging Face for about $11.9 billion, a pending transaction expected to close in the first half of 2027. Nvidia’s financing, infrastructure exclusivity and equity ownership now reach beyond ordinary chip purchases.

Which sovereign programs remain secure?

South Korea’s science minister, Bae Kyung-hoon, said on Nov. 13 that Nvidia confirmed its 260,000-GPU supply plan at corporate level. The pledge, announced during Huang’s October 2025 APEC visit, allocates:

• 50,000 Blackwell GPUs each to Samsung Electronics, SK Group and Hyundai Motor Group.
• 60,000 to Naver Cloud.
• 50,000 to the South Korean government.

A Nov. 13 review meeting with Nvidia agreed to continue existing projects without disruption. Seoul has secured about 21,000 of the 35,000 GPUs it buys directly; the rest are due by early 2027. South Korea was the first customer to speak publicly after the announcement.

The National reported that Humain expanded its Nvidia deal on Oct. 22, three weeks before Huang’s resignation, with plans for up to 84 megawatts and about 27,000 GPUs in two phases. The second, based on Nvidia’s Vera Rubin platform, remains subject to final agreements.

“We are moving quickly to bring the next generations of Nvidia-accelerated computing into Saudi Arabia at meaningful scale,” Humain CEO Tareq Amin said. Neither Humain nor the UAE’s G42 has issued a public statement since the resignation, and both depend on U.S. export approvals that the scandal has placed under congressional scrutiny.

Reuters reported that Chinese AI companies are trying to reserve Nvidia capacity at overseas data centers, fearing Washington will close the remote-access route. Nvidia said Hopper data-center shipments to China contributed less than 1% of data-center revenue in the July quarter.

Seoul has a direct commitment; Gulf and Chinese projects remain exposed to approvals and policy.

What does the market expect?

On Nov. 12, Nvidia fell 4.3%, while AMD rose 5.9% and Broadcom gained 5.5%. CoreWeave, which rents Nvidia chips and in which Nvidia holds nearly 13%, dropped 6.9%. The AP attributed the rival gains to expectations that Nvidia’s customers will diversify suppliers, although no major customer has announced an order change.

William Blair upgraded AMD to outperform and cut Nvidia to market perform. Analyst Sebastien Naji cited “increased scrutiny and a potential leadership vacuum” at Nvidia. Bernstein’s Stacy Rasgon, who downgraded Nvidia, wrote that “the fundamentals of the business (if not the leadership) are clearly still intact.”

Bret Greenstein, chief AI officer at consultancy West Monroe, told the AP: “This could turn what has effectively been a monopoly into a horse race again.” Carnegie Mellon finance professor Chester Spatt was more cautious, saying “their gains are going to be smaller than Nvidia’s losses.”

Separately, Cerebras said it signed a cloud deal that executives identified as Microsoft, with no stated link to Huang’s departure. Nvidia’s Nov. 17 results and call and the Nov. 18 company meeting will give Kress her first chance to address demand, allocations and the OpenAI and sovereign commitments.

Because Nvidia leaves its largest direct customers unnamed, a purchasing shift should appear first in concentration disclosures and hyperscaler capital-spending plans. Huang’s congressional testimony and Washington’s export-control debate will also determine how much sovereign and China business Nvidia can serve under either leader.

The market expects diversification before customers act; Nvidia’s next disclosure will test whether that becomes a measurable change in orders, capacity or supplier competition.