Analysis
Three of the largest listings in the pipeline share a structural feature that has not been priced: their anchor investor is also their supplier, their customer, or both.
Nvidia has committed $1.5 billion to buy non-voting Class N shares of SB Energy at the IPO price, at a company whose campuses will be filled with Nvidia systems and where Nvidia separately acts as residual value guarantor on an Ohio site, per CNBC. It is in talks to put roughly $2 billion into Nscale ahead of that company's New York listing, according to Yahoo Finance. It already holds a position in CoreWeave, the public bellwether for the category. On the model side, Anthropic goes to market with Amazon owning roughly 21 percent and Alphabet about 15 percent, both of which sell it compute.
None of this is hidden -- it is in the filings, which is what filings are for. The problem is analytical, not legal. When a chip vendor funds the buyer of its chips, revenue at both companies grows, and an outside investor cannot easily separate demand from financing. The same dollar can appear as Nvidia's revenue, as a customer's capital expenditure, and as backlog at a third company, and each appearance is legitimate accounting.
“On the model side, Anthropic goes to market with Amazon owning roughly 21 percent and Alphabet about 15 percent, both of which sell it compute.”
The Historical Rhyme
The historical rhyme is telecom vendor financing in 1999 and 2000. Lucent, Nortel and Motorola lent money to carriers who used it to buy their equipment; revenue looked strong until the carriers stopped being creditworthy, at which point the vendors took the losses twice -- once on receivables and once on demand. The differences today are real and worth stating: Nvidia is funding equity rather than extending trade credit, its customers have contracted backlog from creditworthy counterparties, and Nvidia generates enough free cash flow that these positions are small relative to its balance sheet. This is not 1999. But the analytical problem -- circular demand signal -- is the same one.
Three Disclosures That Matter
For public-market investors, three disclosures determine whether the structure is benign. First, what share of a listing candidate's contracted backlog comes from entities in which its anchor investor also holds a stake. Second, whether any purchase commitments are conditional on continued financing. Third, the residual value guarantees -- Nvidia's Ohio arrangement with SB Energy is exactly the kind of off-balance-sheet support that becomes material if utilization disappoints.
For founders raising strategic money, the read is more practical. Strategic capital at the top of the market comes with implicit purchase expectations that show up as concentration risk in your own filing later. That is an acceptable trade when the alternative is not building at all, and Nscale, CoreWeave and SB Energy all made it deliberately. Just know that the same check that gets a campus built narrows the set of investors willing to underwrite the exit.
There is a second-order effect on private valuations that founders should understand. When a strategic investor with a supply relationship sets the price of a late-stage round, that mark becomes the reference point for every subsequent investor -- but it was set by a party whose return comes partly from product sales, not just equity appreciation. A financial investor pricing the same company off cash flows would generally arrive somewhere lower, and the difference is what gets tested when the company faces a public market that only pays for the equity.
The number that will eventually settle the argument: utilization rates at these campuses in 2027, disclosed quarterly, against the take-or-pay terms nobody outside the counterparties has read.
