Nvidia's Quiet Transformation: From Chip Supplier to AI's Financial Backstop

Nvidia's Q2 revenue may beat estimates, but its $105B OpenAI financing and $500B compute platforms shift risk from sales to construction execution.

The numbers due on August 26 will capture the headlines, but the more consequential story may be unfolding in Nvidia's rapidly expanding role as financier, guarantor, and equity partner to the very industry it supplies.

Jefferies analyst Blayne Curtis expects the chipmaker to report second-quarter revenue of $95 billion for fiscal 2027 — comfortably ahead of the $92.07 billion consensus — with the third quarter potentially reaching $108 billion versus the $104 billion analysts are modeling. The company itself had guided to roughly $91 billion. Underpinning those projections is relentless demand for AI infrastructure and the production ramp of the Vera Rubin architecture, which Nvidia confirms is now in full manufacturing, with initial systems slated for partners including Microsoft Azure and CoreWeave in the second half of 2026.

Yet for all the optimism baked into those estimates, the stock has been treading water. Shares changed hands at €185.78 in the primary source, barely moved on the day, while the secondary source cited €186.60, up 0.5 percent. Both sources agree the equity is down roughly 4 percent on the week — 4.6 percent per one account, 4.2 percent per the other — and sits about 8 percent below its 52-week high of €202.50. The stock remains up 16 percent year to date and holds a modest 2.7 to 3 percent cushion above its 50-day moving average.

A Balance Sheet That Now Backstops the Buildout

The most consequential development in recent weeks has little to do with silicon. Nvidia has been assembling a sprawling network of financing vehicles and direct investments that effectively makes it a creditor to its own customer base.

The centerpiece is a commitment of up to $105 billion to support OpenAI's massive data center project in Ohio. The credit facility initially secures 4.25 gigawatts of compute capacity, with an option for an additional 3.75 gigawatts, and the infrastructure is expected to come online in stages beginning in 2028. Nvidia is also investing $1.5 billion in SB Energy, which it announced on August 17, to secure land, power, and building capacity at the PORTS-Pike technology campus — the same site where OpenAI intends to lease eight gigawatts over 20 years.

Elon Musk was quick to pour cold water on the timeline, posting on X that the project will take "much longer to come online than the participants think." That skepticism touches on a genuine shift in risk profile: when a chipmaker becomes the financier of record for its customers' infrastructure, it trades pure sales risk for construction and execution risk largely outside its control.

The Ohio deal is hardly an isolated move. Nvidia recently announced a $5 billion investment in Ilya Sutskever's Safe Superintelligence startup. And alongside Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, the company is building compute-financing platforms designed to mobilize more than $500 billion in third-party capital for AI infrastructure — a structure enabled by the SEC's recent easing of securitization rules for data center investments.

The Portfolio Within the Portfolio

Nvidia's own 13F filing reveals how deep the entanglement runs. The company held 122.76 million SpaceX shares at the end of the second quarter of 2026, a position stemming from SpaceX's acquisition of xAI and valued at roughly $21 billion — making it Nvidia's second-largest holding behind its Intel stake, which was worth about $22 billion at quarter's end, down from $30 billion at the start of the period.

S&P Global Ratings has weighed in on the shifting risk picture, affirming Nvidia's "AA" issuer rating with a stable outlook. The agency noted that the new credit guarantees for AI infrastructure increase debt-like exposure, though the effect on credit metrics remains moderate.

On the regulatory front, Nvidia has received approval to ship advanced H200 processors to Chinese customers including ByteDance and Tencent — a move that media reports suggest could generate up to $30 billion in additional revenue. The company also pushed back on a report that it was developing a China-specific "LPU" using licensed Groq technology to comply with US export rules, stating there are no LPU sales in the Chinese market and no such product in development. Unconfirmed reports indicate Nvidia is testing Rubin Ultra accelerator variants with reduced memory — HBM4 instead of HBM4E — owing to supply constraints at SK Hynix and Micron.

The Test That Matters

None of this — not the $80 billion buyback authorized in May, not the new quarterly dividend of 25 cents per share — resolves the central question hanging over the stock. The August 26 earnings release will show whether the analysts' elevated revenue expectations hold up. But the longer-term verdict depends on something else entirely: whether the data centers Nvidia is now helping to finance actually come online on schedule and generate the returns that justify the balance-sheet exposure.

For a company that built its franchise on selling the picks and shovels of the AI gold rush, the shift to underwriting the mines themselves represents a profound change in what it means to own the stock. The near-term math is favorable — every financed project secures future chip demand. The longer-term calculation is more demanding, and only time — and the construction crews in Ohio — will supply the answer.