Nvidia’s Colette Kress Signals the End of No-Strings A.I. Financing
Observer ·
On her first earnings call as Nvidia ’s interim CEO, Colette Kress delivered a clear message to the artificial intelligence industry: the era of Nvidia acting as an unconditional backstop for the A.I. infrastructure boom is coming to a close. Stepping onto the call in a dual role as interim CEO and longtime chief financial officer, Kress revealed that the chipmaker posted a blockbuster $115.8 billion quarter alongside a $130 billion fourth-quarter revenue guide. It was the company’s first financial report since Jensen Huang’s exit earlier this month.
Yet even as demand for Nvidia’s hardware continues to shatter records, Kress used her debut at the helm to enforce stricter financial discipline. Wall Street’s focus now shifts toward an existential question hanging over the technology sector: who is actually going to pay for the massive global build-out of A.I. data centers?
Throughout Huang’s tenure, Nvidia helped accelerate the AI boom not only by designing the industry’s dominant chips, but also by deploying its massive balance sheet to help customers buy them. That strategy—which critics occasionally labeled “circular dealmaking”—included investing tens of billions of dollars into AI startups and cloud providers that subsequently turned around to purchase Nvidia GPUs. Now, with the spreadsheets officially in charge, Kress is drawing sharper boundaries around how Nvidia deploys its corporate checkbook as the industry’s de facto financier.
During the earnings call Tuesday, Kress outlined stricter conditions governing Nvidia’s capital commitments. She emphasized that the company will put “suppliers first” rather than extending open-ended financial lifelines to emerging cloud operators. Nvidia has already committed roughly $390 billion to secure long-lead-time components such as memory, substrates, and advanced packaging from its own manufacturing partners, underscoring that safeguarding its supply chain takes precedence over bankrolling customer expansion.
When Nvidia does step in to assist partners, Kress made clear that future support will be far more selective and narrowly focused on “land, power, and shell”—the physical site, high-voltage electricity hookups, and core building structure that a data center needs before any chips arrive.
“The availability of land, power, shell is such a critical part of the whole structure in terms of building out those data centers,” Kress told analysts on the call. “So they may need the assistance. That’s a great asset for anyone to hold. So we will look at that as probably one of our top priorities.” At the same time, she stressed that Nvidia is “taking a very strong look in terms of the types of companies that we would want to help in terms of the financing,” noting that the company is rigorously evaluating all future investments, commitments, and guarantees.
The stakes of that scrutiny are most visible in the fast-growing neocloud sector: specialized cloud startups that build GPU clusters to rent out computing power to A.I. developers. While A.I. clouds, industrial, and enterprise customers formed Nvidia’s fastest-growing revenue segment in the quarter, many smaller operators rely heavily on complex debt and credit support to get their facilities off the ground.
Historically, Nvidia’s financial backing in this area has taken the form of lease and credit guarantees. In its July-quarter 10-Q filing, Nvidia disclosed that it had “entered into land, power, and shell guarantees for select A.I. cloud partners’ data center lease obligations in the event of their default.” In plain terms, if a neocloud partner ran out of cash and stopped paying rent to a data center developer, Nvidia was directly on the hook for the bill.
By the end of the October quarter, Nvidia’s maximum gross exposure across all infrastructure guarantees had swelled to $118 billion. That figure dominated by up to $105 billion backing a massive campus leased to OpenAI, alongside $9.5 billion in power and systems credit support and $3.5 billion specifically guaranteeing land, power, and shell obligations for neoclouds. In addition, Nvidia disclosed $32 billion in contingent future commitments tied to neocloud agreements.
CoreWeave exemplifies both the power of that model and the scrutiny around it. Nvidia backed CoreWeave through an arrangement to buy unsold compute capacity through April 2032, starting with a $6.3 billion order that paired a revenue-share mechanism with a demand floor. In January of this year, Nvidia announced plans to use its financial leverage to help CoreWeave secure land, power, and shell for new A.I. factories. CoreWeave has also accumulated roughly $35 billion in debt on expensive credit terms while remaining closely tied to Nvidia as both a customer and an equity partner—making it a bellwether for what happens when Nvidia ties future commitments strictly to independent financing.
Kress used the earnings report and accompanying CFO commentary to put strict guardrails around those obligations. Under the clarified terms, neocloud commitments only take effect if the operators independently secure their own financing and the computing capacity is actually deployed, made operational, and accepted by Nvidia. If a partner falls behind schedule or hits construction snags, Nvidia warned that those commitments could be reduced or eliminated altogether.
“Our guarantee doesn’t exist anymore,” Kress said bluntly when describing what happens if a customer fails to meet development milestones on time. Although she framed the disclosures as a clarification of existing contractual terms rather than an abrupt policy reversal, the public warning is a notable shift in tone from the dealmaking of the Huang era. Reporting by Reuters and the Wall Street Journal characterized the move as a pause on new revenue-sharing deals .
Even with $43.8 billion in quarterly free cash flow and a 20% increase in revenue compared to the previous quarter, Nvidia’s leadership is recognizing that its risk profile increasingly depends on electrical grid readiness and customer solvency, not just raw A.I. chip demand. By prioritizing supplier commitments and refusing to unconditionally underwrite delayed projects, Kress is signaling that Nvidia will no longer absorb the real estate and execution risks of every ambitious startup. Future financing for the physical foundations of the data-center boom will likely need to come from other sources.