Nvidia's $250B OpenAI Deal and the Capex Reckoning
Nvidia in talks to finance $250B OpenAI data center as hyperscalers eye $725B AI capex. What Wednesday's MSFT earnings mean for the cycle.
Key Points
- Nvidia dropped 4.99% Monday — the worst performance in the Magnificent Seven — as reports emerged of a $250 billion financing arrangement with OpenAI for an Ohio data center, raising circular-financing concerns.
- Global semiconductor sales doubled year over year to $110.5 billion in April 2026, and the SOXQ ETF has returned more than 58% in three months, but the sector faces a free-cash-flow reckoning as four hyperscalers prepare to spend $725 billion on AI infrastructure this year.
- Wednesday's Microsoft earnings call is the next pressure test: any softening on capex commitment or commentary on the active FTC antitrust probe could reprice the entire AI infrastructure trade.
Nvidia fell 4.99% on Monday — shedding more market cap in a single session than most S&P 500 companies are worth entirely — as reports surfaced that the company is in talks to provide $250 billion in financing for an OpenAI data center project in Ohio. That number is not a typo. It is also not straightforwardly bullish: analysts immediately flagged the circular logic of a chip supplier financing the customer that buys its chips, and the stock's reaction confirmed the market's discomfort. With Nvidia already trading at a forward P/E of approximately 43x and EV/Sales of 21.5x, any deal structure that complicates the clean revenue narrative gets punished immediately.
The $725 Billion Question
Goldman Sachs calculates that the four largest hyperscalers — Microsoft, Amazon, Alphabet, and Meta — will collectively spend $725 billion on AI infrastructure in 2026, a 77% increase year over year. Amazon alone has approximately $200 billion earmarked, primarily for data center buildout and custom silicon. These are not forecast numbers with wide confidence intervals; they are capital allocation commitments already reflected in balance sheets and forward guidance. The question the market is now asking — loudly, after Alphabet's capex disclosure spooked investors last week — is whether the free cash flow exists to sustain that spending without eventually pressuring dividends, buybacks, or earnings multiples.
The semiconductor industry data supports the demand side of the equation unambiguously. According to the Semiconductor Industry Association, global semiconductor sales rose 93.9% from April 2025 to April 2026, climbing from $56.9 billion to $110.5 billion. The SOXQ ETF has returned more than 58% in the past three months as Microsoft and Oracle placed massive orders for advanced chips. The revenue growth leaders are staggering in their magnitude: Nebius Group up 683% year over year, Micron up 346%, SanDisk up 251%, Credo Technology up 157%. These are not AI-hype stocks trading on narrative — they are companies reporting actual shipped silicon against actual purchase orders.
Nvidia's own reported numbers from its Q1 FY2027 on May 20 were similarly concrete: non-GAAP EPS of $1.87 against $1.77 expected, on revenue of $81.61 billion — up 85.2% year over year. The company's guidance targets $91 billion in Q2 revenue, a figure that would have seemed science fiction 24 months ago. Morgan Stanley maintained its Overweight rating and $288 price target after that print, citing four demand pillars: AI labs, hyperscalers, sovereign AI programs, and neocloud operators. Monday's 4.99% selloff does not invalidate any of those pillars. What it does is introduce a new variable — counterparty and balance-sheet risk — that the market had not previously priced into NVDA's multiple.
Microsoft Wednesday: Capex and the FTC Overhang
Wednesday's Microsoft earnings call is not just a revenue event. It is a referendum on the AI infrastructure thesis itself, and it carries a regulatory dimension that distinguishes it from any other print this week. The Trump administration's FTC has reportedly advanced a broad antitrust probe into Microsoft's cloud, AI, and software businesses — an investigation that originated under the Biden administration but has not been shelved. Microsoft's legal team will be precise in what the CEO and CFO say on Wednesday's call, but any acknowledgment of the probe's scope, or any disclosure of legal reserves, will move the stock.
On the numbers, Wall Street is watching two things above all else: capital expenditure trajectory and Azure revenue growth. If Microsoft's capex guidance for fiscal 2027 shows any deceleration from the current pace, it will be interpreted as a signal that the hyperscaler spending cycle is approaching an inflection — bad for Nvidia, bad for TSMC, bad for the entire AI infrastructure trade. If Azure growth holds above 30% and capex guidance is maintained or raised, it validates Goldman's $725 billion thesis and likely reverses some of Monday's chip-sector selling. MSFT closed up 1.94% Monday, suggesting the market is giving the company the benefit of the doubt ahead of Wednesday — but that optimism is priced into the setup.
Intel's position in this ecosystem is worth tracking separately. Under CEO Lip-Bu Tan, Intel is executing its IDM 2.0 initiative with the 18A process node, which entered volume production in late 2025. Intel fell 7.89% last week in the chip sector's broad selloff, the worst single-week performance in the group. The stock's fragility reflects how little margin for error exists in a turnaround story competing directly with TSMC while simultaneously trying to win foundry contracts from the same hyperscalers spending $725 billion. TSMC itself remains the structural beneficiary of the cycle: key metrics to track include 3-nanometer node revenue trends, gross margin versus the 63–65% guided range, N2 process ramp timing, and CoWoS advanced packaging capacity — the last of which is a direct bottleneck for Nvidia's ability to ship HBM-integrated GPUs at scale.
The Structural Level and the Week's Defining Trade
The week's central analytical question is not whether AI infrastructure spending is real — the semiconductor sales data and hyperscaler commitments confirm it is. The question is whether the market has already priced the cycle's peak into valuations that leave no room for execution risk, balance-sheet strain, or regulatory friction. At 43x forward earnings, Nvidia is priced for a continuation of 85%-plus revenue growth. At that multiple, the $250 billion OpenAI financing deal is not a benign headline — it is a reminder that the circular nature of AI investment creates concentration risk that does not show up cleanly in any single company's income statement.
The SOXQ ETF's 58% three-month return and the reported $725 billion hyperscaler capex commitment together define the bull case's upper boundary. For traders, the actionable framework through Thursday is straightforward: MSFT earnings Wednesday set the tone for capex conviction, Apple and Amazon Thursday provide the consumer and cloud data points, and Nvidia's $91 billion Q2 revenue guidance will be validated or challenged in real time by every data center commentary that crosses the tape between now and August. Watch NVDA's $175 level — a close below that on heavy volume before Thursday would signal the market is beginning to discount a capex deceleration, regardless of what any single company reports.
The Weekly Investor
Daily market analysis for active traders. Free.
