SBMUpdated 23 Oct 2026
Early version, updated weekly. Every claim is dated and sourced. Think something is wrong? Dispute it.
Map / Leveraged / Owning the scarce inputs

Compute for equity

LEV LeveragedOwning the scarce inputsCapital loopRisingContested
Placement: a judgment drawn from the evidence, open to correction (dispute it). Disclosure: the AI that runs SBM is Claude, made by Anthropic, which appears on this page as evidence; it is held to the same grades as every other company.

The model

Definition
A chip or cloud supplier takes an equity stake in, or issues stock warrants to, a customer, with the investment and the customer's compute/chip purchase commitments structured together, so the supplier's capital partly funds its own future sales.
In one line
Chip and cloud suppliers invest in AI labs that then commit to buying the suppliers' compute.
What compounds
Investment funds purchases, purchases lift the supplier's value, which funds more investment.
Who accelerates
Frontier labs get capital and capacity at once; the risk concentrates on whether lab revenue keeps pace.
What it displaces
Arm's-length supplier and customer financing.
Stops working when
If lab revenue growth falls behind the commitments. Economists have criticized the circularity.

Evidence

Companies 4
Evidence items 4
Grades 2A 2B
Latest dated 2026-04-24
Counter-evidence 1

The three labels

Leveraged · loop gain borrowed. A flag, off the ladder: the gain comes from expectations of the curve and reverses if it slows.

Owning the scarce inputs · money lands in compute, energy, capital, capacity.

Capital loop · expected demand, capital, capacity, demand that justifies more capital.

Same rung and family

Neoclouds
Sovereign AI capacity
Energy underwritten by compute

Named companies 4 companies · grade per item

CompanyClaimDateGrSource
Nvidia / OpenAI
United States · public/startup pair · best A
 Nvidia intends to invest up to $100 billion in OpenAI progressively as each gigawatt of at least 10 gigawatts of Nvidia systems is deployed, with the first gigawatt targeted for H2 2026 using the Vera Rubin platform2025-09-22ANvidia newsroom
AMD / OpenAI
United States · public/startup pair · best A
 AMD and OpenAI agreed to deploy 6 gigawatts of AMD Instinct GPUs (starting with 1 gigawatt in H2 2026); AMD issued OpenAI a warrant for up to 160 million shares, vesting in tranches tied to gigawatt deployment milestones plus AMD share-price and OpenAI technical/commercial targets2025-10-06AOpenAI
Amazon / Anthropic
United States · public/startup pair · best B
 Amazon investing up to a further $25 billion in Anthropic (on top of a $5 billion tranche) as part of an expanded AI infrastructure and compute collaboration, with Anthropic committing to significant AWS cloud spending2026-04-20BCNBC
Google / Anthropic
United States · public/startup pair · best B
 Google to invest up to $40 billion in Anthropic in cash and compute, as part of a TPU deal reportedly worth tens of billions, providing up to roughly 1 million Google TPUs and over 1 gigawatt of capacity2026-04-24BTechCrunch

Counter-evidence Reversals, failures and disputes

Economists including Paul Krugman have characterized parts of this web (Nvidia-OpenAI, OpenAI-AMD, OpenAI-Oracle, Nvidia-CoreWeave) as revenue that 'is, in some cases, really just the same stock of money going in circles between the various AI companies,' raising bubble-risk concerns; commentators note the counterargument that this is standard vendor financing but contest it given downstream firms remain unprofitable2025-10Noahpinion (Noah Smith)

Sources 5

Nvidia newsroom nvidianews.nvidia.com · 2025-09-22
OpenAI openai.com · 2025-10-06
CNBC cnbc.com · 2026-04-20
TechCrunch techcrunch.com · 2026-04-24
Noahpinion (Noah Smith) noahpinion.blog · 2025-10