OpenAI와 SB Energy는 계약 체결 시 임차인에게 금융 인센티브를 제공하는 임대 계약을 체결하여 전통적인 부동산 금융 모델을 재편했습니다.
According to Wall Street Journal reporting dated around August 31, 2026—citing internal or filing figures that remain unconfirmed by OpenAI and SoftBank—SB Energy, the SoftBank-majority-owned power developer, granted OpenAI warrants valued at approximately $3.6 billion at issuance in January 2026. By the end of June, those warrants had appreciated to a reported paper mark of roughly $5.5 billion. These figures represent WSJ’s account of unaudited, unrealized positions; the SB Energy S-1 prospectus, which has not yet publicly disclosed these numbers, remains the document that would formally confirm them. Concurrently, OpenAI reportedly contributed approximately $500 million in cash to SB Energy and is expected to hold a single-digit equity stake following the company’s planned listing.
The inducement logic is direct: SB Energy compensated OpenAI in warrants rather than cash to secure its commitment as an anchor data-center tenant. SB Energy is not a shell entity or pre-revenue operation. It is an established solar developer with more than three gigawatts of utility-scale solar operating under long-term power purchase agreements—including contracts with Google—and roughly five gigawatts either operating or under construction, backed by over $9 billion in raised project capital. What it lacks is any operating AI data center. Instead, it has signed a reported contracted backlog exceeding $400 billion across approximately eight to nine gigawatts of data-center leases, anchored by OpenAI. The company is targeting an IPO as soon as September 2026, aiming to raise $5 to $7 billion at a reported valuation of approximately $50 billion. Neither the timing nor the valuation is confirmed, and both should be treated as preliminary until reflected in official filings. Separately, Nvidia is reportedly in discussions to commit roughly $3 billion, with approximately half tied directly to the IPO itself. Those talks remain ongoing and unconfirmed. Every figure cited here derives from WSJ reporting or SB Energy’s own disclosed metrics, not audited financial statements, and should be evaluated accordingly.
The critical insight lies in the transaction’s architecture. Per WSJ’s reporting, OpenAI did not invest in SB Energy and receive warrants as a bonus; rather, the warrants constituted the price SB Energy paid to secure the lease. A frontier AI lab’s willingness to consume power has been converted into a financial instrument worth billions on paper, marked as an unrealized gain before SB Energy has delivered a single megawatt of contracted data-center capacity. The demand signal itself has become the asset.
From a structural perspective, this arrangement operates on a clear mechanism: the tenant is being paid to be a tenant. This is neither a metaphor nor rhetorical flourish. SB Energy required an anchor tenant to underwrite its project financing. OpenAI’s lease signature held greater value as a financial instrument than SB Energy could afford to leave unmonetized, so it converted that signature into equity-linked paper and transferred it back to OpenAI. The demand signal is effectively monetized twice: first as the anchor tenancy that enables SB Energy to raise project debt and equity, and second as a warrant position that grants OpenAI a reported multibillion-dollar unrealized gain simply for providing the commitment.
This double-monetization mirrors the vendor-financed compute model previously tracked at the chip and cloud layers. In that paradigm, a supplier helps fund the demand it subsequently books as revenue, obscuring the true demand curve from external observers. The SB Energy structure applies identical logic to the power layer, with relabeled components: the supplier is now a power developer, the demand signal is a lease, and the financing instrument is a warrant rather than a loan. The underlying architecture remains unchanged. As analyzed in prior coverage of circular AI capital expenditure structures, when a supplier finances its own demand, the resulting revenue figures describe a closed loop as much as they describe an open market.
Viewed through this lens, SB Energy’s prospective IPO substantially sells public investors a securitization of OpenAI’s credit. While the company possesses an established solar business—real gigawatts, real power purchase agreements, and real project capital—the AI data-center backlog justifying a reported ~$50 billion valuation rests entirely on one anchor customer’s commitments across leases that dwarf SB Energy’s current revenue base. That anchor customer owns a slice of the entity it rents from. Nvidia is reportedly positioned to cornerstone the IPO. Every component of the structure references the same counterparty’s forward commitments as its load-bearing pillar. This pattern aligns with previous analyses of Lambda and neocloud pre-IPO financing, where the infrastructure layer increasingly bootstraps itself against the credit of its largest tenants rather than against independent operating cash flows.
“The AI build-out’s financing is becoming self-referential across every layer simultaneously: the chip vendor helps finance the compute, the compute tenant helps finance the power, and each books the other’s commitment as its own forward revenue. From the outside, enormous momentum and a closed loop look identical — until something outside the loop has to pay cash.”
The deeper structural reality, developed in prior work on Beyond Nvidia’s Moat, is that the AI infrastructure stack is growing self-referential at every tier concurrently. Chip vendors extend financing to compute buyers. Compute tenants extend their demand-signal credit to power developers. Power developers securitize that credit into an IPO and invite the chip vendor to cornerstone it. The names on both sides of each transaction keep recurring. This is not evidence of fraud; warrants-for-tenancy is a legal, and in supply-constrained infrastructure markets, a rational inducement structure. The question it raises is analytical rather than accusatory: how much of this ecosystem’s stated demand is backed by cash from entities genuinely outside the loop, versus commitments the participants have written to each other? The answer to that question will determine whether the $400-billion-plus backlog describes a functioning market or merely describes the loop itself.
**Implication 1 — IPO Investors Are Pricing OpenAI’s Credit, Not Just SB Energy’s Solar Assets**
A reported ~$50 billion target valuation for an established solar developer with zero operating data centers represents substantially a bet on OpenAI’s ability to honor multi-decade lease commitments at a scale that dwarfs its current revenue. Public investors evaluating the IPO must underwrite OpenAI’s forward demand durability as the primary risk variable, rather than focusing solely on SB Energy’s project execution—a more conventional infrastructure metric that the prospectus will likely foreground.
**Implication 2 — The Warrant Structure Redefines What “Infrastructure Customer” Means at Frontier Scale**
When a tenant’s willingness to sign a lease commands billions in warrants before a single megawatt is delivered, frontier AI labs’ demand signals have effectively become a distinct asset class. In infrastructure financing terms, these signals are scarcer than physical capacity itself. This inverts the traditional landlord-tenant power dynamic, granting frontier labs implicit leverage over the terms of every infrastructure relationship they enter, with compounding effects on who can build at scale and under what conditions.
**Implication 3 — The Tell Is Outside Cash, And It Is Not Yet Visible**
The circular-financing pattern leaves a critical blind spot: the absence of independent, third-party capital backing the projected demand. Until audited disclosures or actual cash inflows from non-affiliated counterparties materialize, the distinction between genuine market demand and internally generated financing loops will remain obscured. Investors and analysts must treat the reported backlog and valuation targets as forward-looking constructs dependent on continuous rollover of intercompany commitments, rather than settled commercial realities.