AI Infrastructure at a Crossroads: What Nvidia’s Scandal, Anthropic’s IPO, and Power Constraints Mean for Investors
The AI infrastructure trade is facing its most consequential week of 2026. Between Nvidia’s CEO scandal, Anthropic’s massive IPO filing, and soaring power costs for data centers, investors are being forced to reassess every assumption about the AI build-out. Here is what the data says, where the risks concentrate, and which companies are positioned to benefit.
Is Nvidia’s Leadership Crisis a Buying Opportunity or a Red Flag?
Nvidia shares fell 5.52% to $168.89 on November 10, wiping roughly $260 billion off market capitalization after reports that CEO Jensen Huang’s relationship with Alice Huang—a woman U.S. intelligence believes is a Chinese spy—may have facilitated the movement of advanced chips to Chinese entities. The board has launched an independent investigation, the Senate Banking Committee has threatened a subpoena, and House committees have summoned Huang for a closed-door briefing ahead of Nvidia’s November 17 earnings report.
The short answer: the sell-off is company-specific, not a broad reassessment of AI chip demand. AMD rose 1.30% to $515.21 and Broadcom gained 3.91% to $331.98 on the same day, indicating investors are rotating rather than exiting the sector.
The evidence supports this read. Nvidia still commands up to 95% of the data center GPU market, trades at 11x next year’s expected earnings, and maintains a 74.67% gross margin. Wall Street expects 91% revenue growth this year, with management guiding to approximately 90% Q3 sales growth and 70% next-year growth—all described as supply-constrained. However, if Huang steps down, Nvidia lacks a clear successor. He runs the company with a flat structure and 30-60 direct reports, and his departure could cause the core business to stumble. Even if he remains, his influence on U.S. AI policy is expected to diminish, potentially opening the door for stricter export controls and licensing rules.
For investors, the risk calculus is clear: Nvidia’s fundamentals remain intact, but governance and geopolitical risk have materially increased. The November 17 earnings report will be the first real test of whether hyperscaler AI infrastructure spending holds.
Can Anthropic Justify a $1.4 Trillion Valuation?
Anthropic plans a November 18 Nasdaq IPO under ticker ANT, with a share price range of $70-$78 implying a valuation of $1.26-$1.41 trillion—below the $2 trillion market expectation and trailing SpaceX’s $1.77 trillion IPO. The company expects to raise up to $94 billion to fund roughly $1 trillion of compute-related commitments.
The numbers are stark. Anthropic reported Q3 2026 revenue of $17 billion, a 1,100% year-over-year increase, but a 47% quarter-over-quarter slowdown from 144% sequential growth in Q2. Operating expenses surged 292% to $17.5 billion, swinging from a $974 million operating profit in Q2 to a $567 million operating loss in Q3. The company has $895 billion in aggregate commitments, of which approximately $70 billion is due within 12 months following September 30.
Conservative IPO pricing is a deliberate tactic. Analysts view the lower valuation range as a way to manage investor sentiment and secure capital amid political headwinds affecting data-center projects in Texas, Georgia, and Ohio. The reduced downside creates upside potential, but the operating loss and decelerating top-line momentum signal that revenue growth alone will not fund the compute build-out.
Anthropic’s “Anthropic Direct” strategy—leasing chips directly and placing them in dedicated data centers—separates data center capacity commitments from compute infrastructure commitments. Current suppliers include Amazon Web Services, Google Cloud, Nvidia, and AMD. The company also committed $3.4 billion to acquire 1.7 gigawatts of heavy-duty natural gas generators from OnePWR Solutions, the largest direct investment in power generation equipment by an AI company.
The bottom line: Anthropic’s IPO is a capital-intensive scaling play. Investors must weigh soaring revenue growth against massive computing capex commitments and widening operating losses.
Why Is Power Becoming the Binding Constraint for AI Data Centers?
AI data centers are running into a power wall, and the industry’s response is reshaping energy infrastructure investment. Anthropic’s $3.4 billion natural gas generator commitment is just one piece of a broader trend.
Industry peers are securing power at scale. Alphabet acquired Intersect Power for $4.75 billion in December 2025. Meta is pursuing roughly $1.6 billion in on-site gas generation in Ohio with Williams. Microsoft signed a 20-year, $16 billion agreement with Constellation Energy for Three Mile Island nuclear power. Amazon has a Talen Energy nuclear-powered campus agreement and a $20 billion state investment. xAI uses natural gas turbines for Colossus data centers in Tennessee and Mississippi.
Anthropic’s S-1 filing warns that reliance on natural gas exposes its data center buildout to volatile gas prices, competition for gas supply and pipeline capacity, and long-lead-time procurement risks for transformers, turbines, generators, and substations—equipment that can take years to procure amid higher prices, tariffs, and limited production capacity. The company may need to buy power generation above market rates.
For investors, this means the AI trade is increasingly an energy trade. Companies that control power generation, transmission, and distribution infrastructure are becoming essential partners in the AI build-out. Eaton, for example, manufactures switchgear, circuit breakers, and power distribution equipment for data centers. The company reported Q3 2026 revenue of $8.9 billion, up 28% year-over-year, with rolling 12-month organic orders up 39% and backlog up 33%. Six major hyperscalers are targeting more than $1.3 trillion in capex in 2027 versus $870 billion in 2026 and $470 billion in 2025.
How Are Inflation and Interest Rates Affecting AI Infrastructure Financing?
October CPI rose 0.4% month-over-month, above the 0.3% economist consensus, with no year-over-year figure available because the October 2025 government shutdown prevented data collection. Employers added 165,000 jobs and unemployment fell to 4.1%. The stronger-than-expected data increased the likelihood of a half-point Federal Reserve rate hike in December, with FedWatch odds spiking 11.1% on the day and the 10-year U.S. Treasury yield rising more than 13 basis points to near 5%.
Higher borrowing costs directly threaten AI infrastructure financing. CoreWeave, an AI cloud infrastructure provider, reported Q3 revenue of $3.66 billion, up 168% year-over-year, with contracted revenue backlog reaching $132.5 billion. However, the company took on tens of billions of dollars of debt to fund data-center expansion and chip purchases. Debt reached $47 billion, free cash flow was a $7.7 billion loss in the quarter, and interest expense jumped to $846 million from $310 million a year earlier. CoreWeave reported a GAAP net loss of $709 million, or $1.27 per share, and guided fourth-quarter interest expense of $1.05 billion to $1.15 billion.
The company plans to collect $9 billion in customer prepayments and deposits in the fourth quarter and said it would avoid unsecured debt at current interest rates. Risks include a current ratio below 0.5, interest expense above 20% of revenue, and expectations that the Fed will raise rates in December after a hot CPI report.
For AI infrastructure companies, the financing environment is tightening. Year-to-date U.S. corporate debt issuance reached $2.2 trillion, with borrowing by the top five hyperscalers rising from 2% of investment-grade debt in 2023-24 to an estimated 11% in 2026. JPMorgan projects total hyperscaler issuance of $279 billion by year-end, up 156% year-over-year. Memory-chip prices surged 400% this year, squeezing margins for data-center and downstream manufacturers.
Is AMD the Beneficiary of Nvidia’s Troubles?
Investors are rotating into AMD as Nvidia’s scandal unfolds. AMD closed at $515.21, up 1.30%, with a market cap of $834.12 billion and gross margin of 51.19%. The company will invest $2.5 billion of non-voting Class N common stock in Anthropic as part of a $26 billion private placement with major technology companies, signaling a large AI chip supply agreement and revenue visibility.
AMD’s Q3 data center sales rose 97% year-over-year, driven by EPYC server CPUs and the MI450 GPU launch. The company forecasts the server CPU market growing from approximately $26 billion in 2025 to $220 billion in 2030, with AMD’s share at approximately 40% in 2025. The Instinct GPU market is seen expanding from approximately $200 billion to $1.4 trillion by 2030, representing about 64% of AMD’s total revenue in Q3.
Key customers running AI workloads on AMD solutions include Meta, Oracle, OpenAI, Microsoft, and Anthropic. AMD raised server revenue growth outlook for next year to double, and its MI455X GPUs have started shipping. Next-gen MI500 interest is described as “very high.” AMD targets revenue growth above 35% annually and adjusted EPS above $20 within 3-5 years, with management expecting to “significantly exceed” that EPS goal.
The stock trades at 28x 2027 earnings estimates against FactSet’s expected 67% annualized earnings growth, leaving room for outperformance. Near-term risk: a potential slowdown in AI data-center spending could delay deployments. But for investors concerned about Nvidia’s governance and export-control risks, AMD offers a direct alternative with strong customer relationships and a clear product roadmap.
What Do CoreWeave, GE Vernova, and Tidewater Reveal About AI Capex Visibility?
Corporate earnings across the AI infrastructure stack confirm that capital spending remains robust, even as financing costs rise.
CoreWeave’s $132.5 billion backlog—up from $104 billion at the end of Q2—demonstrates contracted future revenue visibility. The company raised full-year revenue guidance to $13 billion to $13.3 billion and guided fourth-quarter revenue of $4.7 billion to $5 billion, implying revenue more than tripled in the fourth quarter. Despite balance-sheet concerns, customer prepayments and deposits signal strong demand for AI computing capacity.
GE Vernova ended Q3 2026 with a total backlog of $197 billion, likely exceeding $200 billion, comprising $104 billion equipment RPO (97% expected recognized within five years) and $92.6 billion services RPO (54% expected within five years). The midpoint 2026 revenue guidance is $47 billion, less than a quarter of the backlog, indicating high revenue visibility. Wall Street consensus forecasts 2030 revenue of $79.3 billion and FCF of $13.1 billion, implying a forward P/FCF of 14.9x. About a third of equipment RPO is recognized annually, and the backlog is driven by early-cycle demand, notably power for data centers.
Tidewater Inc., an offshore support vessel operator, reported Q3 2026 revenue of $360.4 million, up 5.3% sequentially on higher day rates and one month of Wilson Sons acquisition contribution. Net income was $17.5 million versus a $0.8 million loss in Q3 2025. Adjusted EBITDA rose to $140.1 million from $135.5 million, with gross margin of 46.8%. Leading edge day rates increased for the third consecutive quarter to $27,373, and management expects fleet-wide rates to rise about $2,700 per day in 2027. Backlog stands at $1.1 billion for 2027, providing revenue visibility. While not an AI pure-play, Tidewater’s results reflect broader energy infrastructure demand that supports data center power generation.
These companies illustrate that AI-driven capex is translating into contracted revenue, but the financing environment is becoming more challenging as rates rise.
What Risks Does the Litigation-Finance Market Pose to AI Companies?
The U.S. litigation-finance market—valued at $20.6 billion in 2025 and projected to more than double by 2035—presents emerging risks for AI infrastructure companies. U.S. commercial funders manage roughly $16 billion in assets, and 32% of funding commitments in 2024 were for patent cases, where about 30% of suits are now funded.
Heightened opacity and national-security risks stem from undisclosed third-party funders. A Shenzhen-based TCL-linked funder financed Samsung patent suits, and foreign sovereign-wealth-fund involvement—such as Mubadala’s majority stake in Fortress—raises geopolitical concerns. Delaware’s Court of Chancery has not adopted the 2022 federal disclosure order that cut patent filings by 41% in that district, while state reforms in Ohio, North Carolina, and New Jersey, along with pending federal legislation to bar sovereign-wealth-fund financing, could reshape the sector.
For AI companies, this means potential undisclosed funding of patent litigation against their technology, with foreign sovereign wealth funds potentially influencing outcomes. Investors should monitor disclosure developments and assess litigation exposure when evaluating AI infrastructure holdings.
Investment Implications: Where to Position Now
The AI infrastructure trade is entering a more complex phase. Nvidia’s dominance remains intact, but governance risk and export-control uncertainty have increased. Anthropic’s IPO offers exposure to a high-growth AI company with massive compute commitments, though operating losses and decelerating revenue growth warrant caution. Power infrastructure is becoming the binding constraint, creating opportunities for energy equipment and utility companies. AMD is capturing investor rotation away from Nvidia, with strong data center growth and a clear product roadmap.
Macro headwinds from inflation and rising interest rates are compressing financing flexibility for AI infrastructure companies. CoreWeave’s balance-sheet stress demonstrates that even high-growth AI cloud providers face refinancing risk. Companies with contracted backlog, fee-based revenue, and strong cash generation—such as GE Vernova and Eaton—offer more predictable paths through the volatility.
Investors should focus on three criteria: contracted revenue visibility, power infrastructure exposure, and balance-sheet strength. The AI build-out is not stopping, but the winners will be those that can finance their expansion without succumbing to rising rates, regulatory scrutiny, or leadership uncertainty.
FAQ: Quick Answers for Investors
Should I sell Nvidia on the CEO scandal? Not necessarily. The sell-off appears company-specific, with AMD and Broadcom rising on the same day. Nvidia’s fundamentals—95% data center GPU market share, 74.67% gross margin, 91% expected revenue growth—remain intact. However, governance risk and potential export controls warrant position sizing caution until the November 17 earnings report provides clarity.
Is Anthropic’s IPO valuation justified? The $1.26-$1.41 trillion valuation is conservative relative to expectations, reflecting the company’s $17 billion Q3 revenue (up 1,100% YoY) but also its $567 million Q3 operating loss and $895 billion in aggregate commitments. The lower pricing leaves upside room but signals that investors should expect continued losses as compute spending scales.
Which companies benefit from AI data center power constraints? Eaton (switchgear and power distribution), GE Vernova (power equipment with $197 billion backlog), and energy companies securing gas or nuclear power for data centers—including Alphabet, Meta, Microsoft, and Amazon—are directly exposed to the power infrastructure build-out.
How do rising interest rates affect AI infrastructure stocks? Higher rates increase financing costs for debt-heavy AI infrastructure companies. CoreWeave’s $47 billion debt load and $846 million quarterly interest expense illustrate the risk. Companies with strong cash flow and contracted revenue are better positioned to withstand tighter monetary policy.
Is AMD a viable alternative to Nvidia? Yes. AMD’s Q3 data center sales rose 97% YoY, with major customers including Meta, Oracle, OpenAI, Microsoft, and Anthropic. The $2.5 billion Anthropic investment signals supply agreement visibility. AMD trades at 28x 2027 earnings with 67% expected annualized earnings growth, offering a direct alternative for investors rotating out of Nvidia.