As of late 2025, the most valuable private company in history has reportedly walked away from an IPO and attached a $1 trillion valuation target to its next milestone. Run the arithmetic before you accept the narrative. Public reporting puts 2024 revenue near $3.7 billion, 2025 guidance around $13 billion, and 2026 projections in the $29–30 billion range. A $1 trillion price tag against roughly $30 billion of forward revenue implies a 30x to 80x price-to-sales multiple, depending on which forward year you credit. No public-market allocator signs that cheque without a defensible cash-flow story. The decision to stay private is not a retreat from ambition. It is a decision about who gets to price the risk — and who is left holding it when the music stops.
Map the capital structure, because it explains the decision better than any public statement. In October 2024 the company raised $6.6 billion at a $157 billion valuation. By spring 2025 the figure had roughly doubled to $300 billion on a $40 billion round led largely by SoftBank. By late 2025, secondary sales of employee shares were reportedly clearing near $500 billion. The stated pre-IPO milestone is now $1 trillion — a sixfold expansion from where the cycle began, achieved without a single full public reporting year behind it.
Behind that valuation sits a commitment schedule that dwarfs the revenue line. Stargate, announced in January 2025, carries an initial $100 billion price tag and a $500 billion ambition. Multi-year cloud contracts with Oracle are reportedly sized in the hundreds of billions. Procurement agreements with NVIDIA, AMD, and Broadcom lock in compute capacity years ahead of demonstrated demand. The 2025 restructuring into a public benefit corporation, with a nonprofit foundation retaining control through special voting rights, added another layer that any IPO diligence team would have to unwind line by line — especially as the legality of that structure remains contested across several US jurisdictions.
The regulatory picture compounds it. The EU AI Act imposes transparency and copyright-disclosure obligations on general-purpose models. US frontier-model reporting requirements and procurement eligibility rules widen the disclosure surface. Multiple copyright suits from publishers, authors, and rights holders remain unresolved. For a company preparing an S-1, unresolved litigation and an untested governance structure are not footnotes. They are pricing inputs.
Here is the structural point most coverage misses. The $1 trillion figure is almost certainly not a founder's aspiration — it is a financing covenant. SoftBank's capital has been deployed in tranches, and tranche-based capital almost always carries valuation gates. If the milestone functions as a trigger, then "waiting for $1 trillion" is not patience. It is an obligation. That distinction reframes the story from ambition to constraint, and it changes what you should be monitoring.
Now watch how the cash flows loop. NVIDIA invests in OpenAI. OpenAI commits to buying NVIDIA silicon. Oracle books OpenAI's cloud orders, and that backlog supports Oracle's equity, which raises capital for more infrastructure, which flows back as capacity for OpenAI. Chip vendors take warrants in exchange for purchase commitments. Every node in the graph is simultaneously a customer, a supplier, and a shareholder of the adjacent node. In private markets, the loop is self-reinforcing because nobody is required to mark the correlation. In a public filing, related-party disclosure would strip the loop to net exposure — and the net is dramatically smaller than the headline.
There is a second-order effect most coverage ignores. When the same vendors that supply compute also hold equity in the buyer — through warrants, convertible instruments, or strategic stakes — the vendor's own revenue recognition depends on the buyer's ability to pay. That is not diversification. It is correlation. In 2021, DeFi protocols discovered that when your collateral, your borrower, and your oracle are the same entity, the liquidation cascade is not a risk scenario. It is a scheduled event.
I have audited this exact shape before. In 2021, as a junior researcher at a Melbourne Series A, I documented that 70% of a protocol's user liquidity sat trapped in its own illiquid governance token. The token was the collateral, the incentive, and the reward simultaneously — a closed loop that rendered beautifully on a dashboard and behaved like a bank run the instant redemptions started. I proposed a pivot toward real-world asset tokenization. Leadership rejected it. Six months later the model failed precisely along the lines in my memo. Compute commitments are the governance tokens of this cycle. If they carry take-or-pay minimums — and multi-year procurement agreements usually do — they are, in credit terms, debt: senior claims on future cash flow, invisible on any balance sheet not forced to consolidate them.
Then layer the maturity mismatch. Revenue arrives in billions, spread across years. Commitments are drawn in hundreds of billions, up front. The gap must be rolled continuously through equity, debt, and structured vehicles — joint ventures, SPVs, equipment leases. That is not a business model. It is a funding stack that requires the private window to stay open indefinitely. When the window narrows, the commitment schedule stops being a moat and becomes a guillotine.
One more cash-flow line rarely gets modelled: people. An IPO is the single largest liquidity event available to a research workforce. Deferring it defers every engineer's exit, which means the company must keep buying back employee shares at rising marks to prevent attrition to competitors or to self-funded startups. Each tender offer sets a fresh, higher reference price — and that price becomes the anchor for the next private round. Employee retention is being financed with valuation inflation, and the inflation is now a fixed cost.
The valuation math is where this gets uncomfortable. At 30x forward sales, $1 trillion requires roughly $33 billion in annual revenue. At a more conservative 10x — a multiple befitting a capital-intensive utility with commodity risk — it requires $100 billion. The 2026 estimate of roughly $30 billion clears the first threshold only if you accept the most generous multiple in the range. Every step of optimism you remove from the model adds a multiple of revenue the company must eventually produce. That is the entire bet: the $1 trillion is not a price for today's business. It is a prepayment on a business not yet built, financed by parties who are also the counterparties.
The consensus read is that OpenAI postponed its IPO because it is afraid. I think the opposite is closer to the truth. The company is choosing the venue where valuation is set by narrative rather than cash flow — and it is choosing rationally. Public markets would price this as a capital-intensive utility facing commodity risk. Private markets are still pricing it as an AGI monopoly. Those are not the same company, and only one of those prices clears $1 trillion.
Here is the part that should unsettle anyone holding crypto. The AI sector is running the exact playbook DeFi ran in 2020–2021 — promises first, cash flow later, valuations set by reflexive insider loops — and it is being rewarded for opacity with record private capital. Crypto ran the same playbook, was punished for its transparency, and is still trading below prior highs. The difference is not virtue. It is that public ledgers force the loop to be marked to reality in real time. Opacity is currently the most valuable asset in technology finance, and it is not listed on any exchange.
This connects directly to where I spend my time. If autonomous agents become primary liquidity providers in DeFi by 2026 — and every agentic roadmap I review points that direction — the settlement rails those agents use will be on-chain. The capital funding those agents is being formed right now, in private AI rounds. AI's capital stack is not a Silicon Valley curiosity. It is the upstream plumbing of crypto's next liquidity cycle.
Watch the tranche, not the tweet. The next tender offer, the size and valuation of the next private round, and whether any compute contract gets delayed or renegotiated will tell you more than any model launch. Track the Oracle and CoreWeave earnings calls specifically — when backlog converts slower than promised, the loop leaks. If the private window stays open, $1 trillion is a scheduling problem. If it closes, the same commitments that built the moat become the margin call — and crypto, which already ran this experiment and paid for the lesson, will be the first place that pain surfaces as liquidity.