September 19, 2026 · BriMindInvest Research Team · 12 min read
A small group of companies — Nvidia, OpenAI, Microsoft, Oracle, and CoreWeave — now sit at the center of a web of investments and compute commitments that has become one of 2026's most debated risks in AI infrastructure investing. Here is the actual structure of these deals, the bull and bear case for what it means, and how to read backlog numbers with appropriate skepticism.
Mapping the Web: Who Owes Whom, and For What
Each link in this chain is individually a normal commercial arrangement. What makes it notable is how the chain closes back on itself — money and compute commitments flow in a loop that touches the same handful of companies repeatedly.
Nvidia → OpenAI
Nvidia committed up to $100B in staged investment tied to OpenAI deploying multiple gigawatts of Nvidia systems — investment and revenue flowing in the same direction, between two closely linked parties.
Microsoft → OpenAI
Microsoft's multi-year, multi-billion-dollar Azure compute commitment to OpenAI is repaid largely in Azure consumption — cash converts into cloud revenue that shows up on Microsoft's own income statement.
OpenAI → Oracle
OpenAI's Stargate-linked compute commitments to Oracle run into the hundreds of billions over the contract's life, a backlog Oracle has publicly cited as transformative for its cloud infrastructure business.
Oracle → Nvidia
Oracle, in turn, spends heavily on Nvidia GPUs to build out the capacity it has promised to OpenAI and other AI customers — closing a loop where Nvidia's customer's customer is effectively paying for Nvidia's own chips.
Nvidia → CoreWeave
Nvidia has taken an equity stake in CoreWeave and reportedly backstops portions of CoreWeave's GPU capacity, while CoreWeave is one of Nvidia's largest GPU customers and an AI-cloud provider to OpenAI and others.
The Bull Case: This Is How Infrastructure Buildouts Work
Real, contracted revenue and real compute delivered: Unlike the dot-com era's vendor financing schemes that often involved little real economic activity, these deals result in physical data centers being built, chips being manufactured and shipped, and cloud services actually being consumed.
Each party has independent demand drivers: Microsoft needs Azure capacity regardless of OpenAI; Oracle needs cloud revenue growth regardless of Stargate; Nvidia needs distribution regardless of any single customer. The relationships overlap, but none of these companies is purely dependent on the others for its own survival.
Circular financing isn't inherently fraudulent: Vendor financing arrangements — where a supplier helps finance a customer's purchases — are a long-standing, legitimate business practice used across industries from semiconductors to telecom, provided the underlying demand and cash flows are real.
The Bear Case: Concentration Risk Disguised as Demand
It inflates the appearance of organic demand: When Nvidia's investment dollars flow to OpenAI and back to Nvidia (via Oracle, via CoreWeave, via direct purchases), it becomes difficult for outside investors to distinguish genuine end-market AI demand from demand manufactured by the deal structure itself.
Concentration risk is compounding, not diversifying: If OpenAI's revenue growth disappoints or its path to profitability stalls, the shock doesn't stay contained to OpenAI — it cascades through Microsoft's Azure guidance, Oracle's backlog assumptions, CoreWeave's contract book, and ultimately Nvidia's order pipeline, because they are all counterparties to each other.
It echoes prior tech-cycle financing patterns: The telecom buildout of the late 1990s featured a similar pattern of vendor financing between equipment makers and network operators, where genuine infrastructure got built but valuations collapsed once financing dried up and the mismatch between committed capacity and realized demand became clear.
Accounting treatment can obscure the real economics: Multi-year compute commitments, equity stakes, and staged investment tranches all involve judgment calls about revenue recognition and timing that make headline backlog and revenue figures harder to compare across companies than a simple cash sale would be.
How to Read AI Backlog Numbers With Appropriate Skepticism
Ask who the counterparty actually is
A $50B compute backlog is a very different risk if it's spread across hundreds of independent enterprise customers versus concentrated in one or two AI labs that are themselves dependent on continued external funding.
Separate cash revenue from committed-but-unbilled backlog
Multi-year contract value announced in a press release is not the same as revenue recognized on an income statement — watch actual quarterly revenue and cash flow conversion, not just headline backlog figures.
Track whether investment dollars and purchase commitments move in the same direction
When a company both invests in a customer and sells to that same customer, or when a supplier and a customer take equity stakes in each other, look closely at whether the transaction sizes are proportionate and whether independent demand exists outside the relationship.
Watch OpenAI's public financial disclosures closely, where they exist
As the private hub through which much of this financing web routes, any signal about OpenAI's revenue growth, cash burn, or path to profitability has outsized implications for its public counterparties' AI-related guidance.
Bottom Line: Not a Fraud, But a Concentration Risk Worth Pricing In
The circular deals structure among Nvidia, OpenAI, Microsoft, Oracle, and CoreWeave doesn't mean AI infrastructure spending is fake — real data centers, real chips, and real compute are being delivered and consumed. But it does mean that a meaningful share of the AI infrastructure story's growth is concentrated among a small number of interdependent counterparties, which is a real risk that headline backlog numbers alone don't capture.
The practical takeaway: continue to hold AI infrastructure exposure if you believe in the multi-year buildout thesis, but weight positions toward companies with diversified, independent customer bases over those whose growth is most concentrated in this specific deal web, and watch OpenAI-linked disclosures as a leading indicator for the group.
Data sources & disclosures: Financial data and metrics cited in this article are sourced from company SEC filings, earnings releases, and investor relations materials. Market prices and fundamental data are provided by financial market data providers. Market size estimates and industry projections are sourced from industry research and analyst reports. Figures reflect information available at the time of writing and may have changed. AI scores and price targets are proprietary estimates — see our Methodology. This article is for informational and educational purposes only and does not constitute financial advice or a recommendation to buy or sell any security. Investing involves risk, including the possible loss of principal. Please read our full Disclaimer and consult a licensed financial adviser before making investment decisions.
We use cookies to keep you signed in and to understand how people use our site (Google Analytics). You can accept all cookies or limit them to what's strictly necessary. Privacy policy