Nvidia's Post-Huang Era Tests Kress on China, Neoclouds, and Trust
Colette Kress faces her first earnings call as interim CEO amid a China scandal, declining stock, and questions over whether Nvidia can sustain its neocloud financing model without Jensen Huang's personal deals.
The Stock Has Dropped for Six Weeks Straight
Nvidia’s market cap has shed $1.7 trillion since Jensen Huang left last Thursday. That is the longest losing streak the company has posted since 2012 — a milestone no one inside the organization wanted to set. The numbers that come out Tuesday will matter, and analysts are tracking them closely. Consensus calls for roughly 91 percent revenue growth to $109.1 billion, driven largely by continued demand for the Blackwell GPU family. But those figures are not the real test for Colette Kress. They never were.
The test is whether investors believe Nvidia can operate without the one person who made it feel indispensable. Huang did not merely design products or set corporate strategy. He personally closed deals that no CFO could replicate on a scheduled conference call. He knew the founders of eleven frontier labs by their first names. He had dinner tables in three countries where GPU allocations were discussed before any contract was drafted. When he walked away, the machinery kept running — but nobody involved in that machinery has been able to convince Wall Street it will keep producing at the same rate.
Kress, who spent twenty-one years at Nvidia and served as CFO through every quarter of fiscal 2024 and 2025, knows the books inside out. What she does not have is the relationship capital that made those books look so clean. On Tuesday, the market will be watching not just what she says about revenue, but what she does not say.
China Sales Remain the Unanswered Question
Huang’s departure was triggered by a New York Times investigation that unfolded over six weeks in early 2026 and ultimately forced his resignation. The reporting centered on Alice Huang — a separate individual, no relation to Jensen — an executive at Megaspeed International whom U.S. intelligence was surveilling as a suspected Chinese asset. According to the investigation, Alice Huang used her position at Megaspeed to route Nvidia’s premium chips into China through a network of shell companies registered in Southeast Asia, most notably in Vietnam and Cambodia, violating federal export restrictions enacted in October 2023.
Alice Huang left Megaspeed in late 2025. The company says it launched an internal investigation shortly after her departure and has cooperated with the Commerce Department’s Bureau of Industry and Security. Nvidia has consistently reported essentially zero data center sales from China in its SEC filings — a figure that looks increasingly difficult to defend under scrutiny.
Culper Research, a short-selling firm based in Chicago, published a report in March 2026 estimating that 20 percent of Nvidia’s fiscal 2026 compute revenue was driven by China through pass-through buyers operating out of free-trade zones. Nvidia denies this categorically. Its CFO’s office issued a two-paragraph rebuttal that restated the company’s compliance posture without addressing the methodology behind Culper’s estimate. The contradiction itself — a formal denial versus a detailed alternative calculation — is what worries Washington.
Senators Ron Wyden and Richard Neal, both Democrats on the Senate Finance Committee, sent a letter to the Commerce Department in April requesting that all communications between Nvidia and Jensen Huang regarding chip sales approvals be preserved under document-retention rules. Treasury Secretary Scott Bessent raised the issue publicly at a Capitol Hill hearing in May, asking bluntly whether Nvidia knew its clients were shell companies or simply lacked the incentive to verify them. The question drew uncomfortable silence from Nvidia’s investor relations team when repeated at a subsequent conference.
Kress will not be able to deflect this line of questioning. She was CFO during every quarter in which these sales structures operated. Her signature appears on the financial statements that reported near-zero China exposure. On Tuesday’s call, the boardroom will watch closely to see if she confirms Nvidia’s denial or offers the first hint that the reality was messier than the filings suggested. Either path carries risk. Confirmation hardens the case against the company. Ambiguity makes investors nervous about governance standards going forward.
The Neocloud Financing Machine Could Stall
Huang did something unusual for a semiconductor executive. He built an entire category of competitors alongside Nvidia’s hyperscaler customers. He called the structure ACIE internally — AI Compute Infrastructure and Entrepreneurship — though the press settled on neoclouds. Companies like Megaspeed bought Nvidia compute power at wholesale, resold it to smaller AI labs and startups, and in some cases received direct equity investment from Nvidia’s venture arm. Nvidia did not just supply chips to these entities. It financed them. It invested in them. It used its balance sheet to secure land, power and GPU allocations that startups could never obtain on their own.
In August 2025, Huang announced a program pairing five frontier labs with six major financial firms to fund new data centers across the American South. Four of those sites broke ground by early 2026. No deal has fully closed yet — financing packages remain stuck in regulatory review, and two of the six financial firms have quietly withdrawn from the program, according to people familiar with the matter who spoke on condition of anonymity. Bloomberg Intelligence analyst Kunjan Sobhani noted plainly that those deals relied on Huang’s relationships and his willingness to leverage Nvidia’s capital as both guarantor and co-investor. Kress does not have the same access to those financial partners. Whether she wants to use it, or even whether she wants to continue the model, remains an open question inside the boardroom.
Morgan Stanley cut its 2027 and 2028 data center sales estimates for Nvidia by roughly 10 percent in the week following Huang’s resignation. The analysts wrote that the revision was deliberately conservative — a phrase that reads less like hedging and more like a warning to clients who might interpret the cut as a sign of temporary weakness rather than structural recalibration. JP Morgan followed suit three days later, lowering its conviction rating on the stock while maintaining its price target, a signal that the fundamental outlook had changed even if the valuation anchor had not.
The ripple effects extend beyond Nvidia’s P&L. Several neocloud operators that depended on Nvidia-facilitated financing have begun courting AMD and custom silicon providers as alternates. This is not yet a flood — most of those operators signed multi-year contracts with penalty clauses — but the direction of travel is visible. If the neocloud model shrinks, it removes a significant portion of Nvidia’s incremental revenue headroom, the very revenue that analysts have been using to justify growth multiples that already assume sustained double-digit expansion through 2029.
Demand Is Real. Trust Is Not.
Huang told investors in August that demand was accelerating. Kress repeated the phrase on the last earnings call, calling it a supply-constrained outlook with roughly 70 percent revenue expansion forecast for fiscal 2028. Those are the same words. The conviction behind them is harder to measure when the speaker has inherited a mandate rather than built it.
Nvidia has committed $92 billion in supply and capacity spending for the rest of this fiscal year, $87 billion for fiscal 2028 and $88 billion for fiscal 2029. Those figures show the company still expects demand to outstrip production by a comfortable margin. TSMC, Nvidia’s primary foundry partner, has reportedly reserved additional advanced-node capacity through 2027 specifically for Nvidia orders — a signal that the supply chain side of the equation has not lost confidence. But commitments are cheap when you have an ironclad relationship with Wall Street. They become expensive fast when that relationship frays.
The nuance that gets lost in headlines is that Nvidia’s backlog is not shrinking. Customers are not walking away. What is happening is slower, subtler, and potentially more damaging to the stock narrative. Startups that would have committed to Blackwell allocations eighteen months ago are now waiting to see what the interim leadership does about China compliance and neocloud financing before locking in their positions. That hesitation translates into forecast uncertainty, and forecast uncertainty is what sells off stocks faster than any single bad quarter ever could.
For thirty-three years, Nvidia’s story was simple enough to tell in a keynote: we build the chips, the world buys them, growth follows. Huang made the narrative feel inevitable because he made himself the bridge between every major customer and every major supplier. Kress inherited a business that works regardless of who sits in the CEO chair — the engineering culture, the Founders Day rituals, the relationships with TSMC and Samsung and the hyperscalers are institutional, not personal. But she inherited a narrative that no longer holds together on its own.
What Comes Next
The board has said it expects Kress to deliver on commitments to shareholders, customers and partners. Investors are waiting to hear whether that translates into action on two fronts: tighter controls on China-bound shipments and a clear signal on whether the neocloud financing model continues under her leadership. Both issues require answers that go beyond standard corporate language. Neither can be deferred to a permanent CEO search that may not conclude before fiscal 2027 earnings.
If Kress walks into Tuesday’s call and repeats Huang’s forecasts without addressing either issue directly, the market will read silence as weakness. Analysts have already sharpened their questions — at least four from different coverage teams are expected to press her on the Alice Huang situation and the status of the ACIE partnerships. If she offers specifics — even uncomfortable ones — she may reset expectations faster than waiting for a successor to arrive. The alternative is a prolonged period of ambient doubt, the kind that erodes multiples gradually and then all at once.
Huang founded this company in 1993. His absence is felt less in the product roadmap — the Blackwell refresh and the Rubin architecture are on schedule — than in the relationships that made the roadmap matter. Nvidia’s next chapter starts with whether those relationships survive him, and whether Colette Kress can convince the people who built them to keep believing in the company without its most visible architect.