Sources told the media on Thursday that NVIDIA (NVDA.US), the AI chip titan that just delivered a blockbuster earnings report, has paused a portion of transactions under its new financing initiative. That program was designed to offer credit support to cloud computing firms focused on AI infrastructure services, in exchange for a share of their revenue. According to those familiar with the matter, the chip giant pulled back on parts of the plan last week, while adding that NVIDIA may later redesign the initiative or fold it into another program.

What NVIDIA has paused is only a segment of the revenue-sharing deals under its "AI Compute Partnership" program, not a termination of the entire initiative, and certainly not a sign that demand for its AI GPU clusters such as Vera Rubin is softening. The company has made clear that this business model, aimed at expanding compute accessibility, remains in place and will continue to be adjusted in response to robust demand. What has actually been halted is a portion of the most controversial combination of deals — those in which NVIDIA both sells GPU clusters and backstops customers' unsold capacity, assists with their financing, restricts which tenants they can lease to, and takes a 50% cut of revenue above a certain threshold.

NVIDIA's second-quarter fiscal 2027 revenue surged 106% year over year to $96.2 billion, with data center revenue climbing 117% to $89 billion, and the company guided for roughly 70% revenue growth in the next fiscal year. Meanwhile, Anthropic and Nscale signed a six-year, $45 billion compute procurement agreement to deploy the latest Vera Rubin platform. These figures signal that global demand for AI infrastructure resources and next-generation compute clusters remains exceptionally strong — NVIDIA's pause on some deals is not a cooling of compute orders, but a risk contraction in its credit support, compute leaseback, and revenue-sharing model, actively pushing the most contentious compute transactions toward "de-circularization."

Financing Boundaries Tighten! NVIDIA Halts Some Revenue-Share Deals, Triggering Compliance Reassessment of a $500 Billion Credit Expansion

More precisely, NVIDIA is erecting firewalls around the "circular financing" controversy, antitrust scrutiny, and concerns over excessive control of customer operations: the program can both help smaller AI cloud service providers buy NVIDIA chips, and potentially turn a hardware supplier into a financier, a backstop lessee, and a revenue-sharer all at once. As Jensen Huang, Sam Altman, Elon Musk, and other tech leaders are set to appear together at a G20 technology meeting, this adjustment shows the AI compute super-cycle is shifting from purely competing on capital deployment toward a new phase that equally emphasizes financing quality, regulatory boundaries, and genuine end demand.

By pausing part of the deals that combine "GPU sales + credit support + leaseback guarantees for unsold capacity + 50% of excess cloud revenue," NVIDIA is personally driving the "de-circularization" of compute deals. This effectively reduces the feedback loop of "financing customers to buy your own chips, then guaranteeing their revenue," forcing orders to face greater scrutiny from independent end demand and real cash flows, while sidestepping risks tied to antitrust rules, over-control of customer operations, and "AI circular financing" that could threaten credit ratings.

A NVIDIA spokesperson responded to the latest media reports, stating: "This new business model for opening access to compute for the rapidly growing AI ecosystem... is still in operation and continues to evolve due to strong AI compute demand." According to media reports, NVIDIA's action comes less than two months after announcing the program, which was designed to meet the strong financing needs and AI compute resource demands of smaller AI cloud companies. Previously, NVIDIA had planned to reverse-lease compute capacity from cloud customers if they could not rent it all out, providing a safe buyer with backstop support and making it easier for these startup AI cloud firms to borrow the massive capital needed to purchase NVIDIA AI GPU clusters. Under this model, NVIDIA first generates sales figures by selling the hardware itself, then also takes a share of the cloud revenue customers generate from using NVIDIA compute.

During this week's earnings call, NVIDIA said that over the medium to long term, this virtuous operating model could generate revenue on the order of billions of dollars. But in recent months, as NVIDIA has continually plowed capital back into the AI ecosystem, investor scrutiny and caution have intensified, fueling serious market concerns about its participation in so-called "AI circular financing and trading." Worries that NVIDIA's dominant AI GPU influence could be artificially inflating demand have led to market chatter about an "AI credit bubble burst," triggering short-term selling pressure on both the share prices and credit bond trading prices of NVIDIA and other AI compute supply-chain leaders.

The company this month helped its customers secure $500 billion in financing from major U.S. financial institutions and agreed to provide up to $105 billion in backstop guarantees to help OpenAI lease a large data center. Media reports citing sources familiar with the matter said some senior NVIDIA employees have expressed concerns to existing and potential customers that the program could attract antitrust review. Cloud customers also have sensitive worries about how much control NVIDIA can exert over their operations, especially those of hyperscale cloud giants. According to reports, in the initial weeks of the program, the degree of control NVIDIA sought to impose had already angered some potential large partners. Reports added that NVIDIA has told some cloud service providers they could only rent out NVIDIA AI GPUs to customers approved by the company; NVIDIA also stated it prefers allocating compute resources across multiple smaller enterprises or early-stage AI cloud leasing firms rather than concentrating capacity with a single hyperscale cloud customer. Sources said that under some proposed deal terms, once a cloud provider's revenue from using NVIDIA AI GPUs exceeds a certain operating threshold in the agreement, NVIDIA would take 50% of the revenue above that threshold — but these preliminary terms have led some large cloud companies to raise questions.

The AI Bull Market Is Not Over — The Financing Flywheel Is Simply Being Trimmed! NVIDIA's AI GPU Chain Enters a "De-Circularization" Era

NVIDIA's latest move resembles a "proactive brake" on regulatory and credit risk: it may reduce potential revenue sharing in the near term, but it helps alleviate concerns over circular financing, antitrust reviews, and customer operational autonomy, enhancing the verifiability of future revenue. For NVIDIA's strong fundamental outlook, this is likely a neutral-to-slightly-positive governance repair overall. Demand has not simultaneously turned a corner; it continues to show exponential expansion. NVIDIA's second-quarter fiscal 2027 revenue grew 106% year over year to $96.221 billion, with data center revenue surging 117% to $89 billion. Third-quarter revenue guidance came in at $108 billion, plus or minus 2%, with Vera Rubin demand strong and full-scale production already underway. Management expects revenue growth of about 70% in fiscal 2028, emphasizing that this figure is already constrained by memory/storage chip supplies and TSMC's advanced process manufacturing capacity.

Simon Leopold, a veteran analyst at the well-known Wall Street investment firm Raymond James, sharply raised NVIDIA's price target from $352 to $550 in a Thursday report. Leopold's analysis suggests this target implies roughly a 130% upside from current levels, which would put NVIDIA's market value at around $13 trillion. The chip giant's current market cap is approximately $5.2 trillion. Meanwhile, Anthropic and Nscale signed a six-year, $45 billion compute procurement agreement, locking in roughly 460 megawatts of power capacity and planning to deploy NVIDIA's Vera Rubin platform. These signals indicate that the global AI infrastructure race is extending from chip orders further into data center core infrastructure bottlenecks, power resources, and long-term compute contracts, with capital deployment intensity still in an acceleration phase.

Market descriptions of NVIDIA as an "AI central bank" or offering "Balance-Sheet-as-a-Service" aptly capture the trend of NVIDIA expanding from chip supplier to ecosystem financing intermediary. However, treating the entire $530.5 billion in off-balance-sheet commitments and guarantees on NVIDIA's balance sheet as already-formed debt is inaccurate. NVIDIA's official filings show this figure comprises $366 billion in future commitments, $56 billion in additional AI cloud and third-party data center commitments, and a $108.5 billion maximum notional guarantee exposure. Of this, $279 billion is supply commitments toward memory and manufacturing capacity, some of which can be cancelled, rescheduled, or adjusted, while the $105 billion OpenAI-related guarantees only begin to take effect gradually from fiscal 2029 as nine construction phases come online.

What truly warrants vigilance is cash conversion: accounts receivable rose from $38.466 billion to $63.059 billion, with the top five customers accounting for 70% of that. Second-quarter operating cash flow was approximately $24.077 billion, notably below the $59.688 billion GAAP net income, though the gap is also affected by equity investment gains, inventory increases, and extended payment terms from investment-grade customers, so it cannot be directly equated with deteriorating customer credit. For the entire AI compute supply chain, NVIDIA's latest adjustments to compute revenue-sharing agreements and credit guarantees do not mark a turning point in the cycle's health, but rather a potential watershed in financing discipline that could ease AI circular-financing and antitrust anxieties. Orders driven by hyperscale cloud providers and high-credit customers — including TSMC-led NVIDIA capacity chains, HBM and high-performance NAND-dominated memory, advanced packaging systems, high-speed Ethernet network infrastructure, optical interconnect, data center power chains, and liquid cooling — still carry high certainty. What will truly face pressure are small and mid-sized emerging cloud providers that rely on NVIDIA's credit backing but lack long-term offtake contracts and stable cash flows, as well as highly leveraged financing chains built around speculative data center campuses.