What the money shows
The biggest AI companies invest in their own customers, and the customers spend it back.
NVIDIA puts equity into AI labs and clouds that then buy NVIDIA chips. OpenAI raises from Amazon, NVIDIA, SoftBank and Microsoft, then commits hundreds of billions to Oracle, Microsoft, CoreWeave, AMD and NVIDIA. The same pattern reaches startups in India and Europe. All of it is disclosed; the open question is how much real end-customer demand sits underneath.
Main findings
What this evidence does not show
- No sign of money laundering. Laundering means hiding where illegal money came from. Every deal here was announced by the companies or reported from filings and named sources.
- Circular financing is legal. Vendor financing is common in hardware booms. The risk is that it makes demand look stronger than paying end customers justify, as it did for telecom suppliers in 2001.
- "Up to" is not cash paid. Many figures are ceilings or multi-year commitments that depend on milestones.
- Reported is not confirmed. Items marked reported or single source come from outlets citing unnamed people or one publication. Check those before relying on them.
How to read it
Views
Network
Everyone connected to everyone
Each line is one sourced relationship. Hover a dot to see its links, click it for the evidence. Turn on the founders layer to see where the people came from. Pinch or use the buttons to zoom; hold Ctrl or ⌘ to zoom with the wheel.
Money loops
Where the money comes back
Each diagram is one loop: a company invests in a customer, and the customer spends with the investor. Blue arrows are equity, orange arrows are spending or contracts, green dotted lines are partnerships.
Flows
OpenAI has committed far more than it has raised
Investments on the left, the labs in the middle, the clouds and chips they have committed to on the right, in billions of US dollars. NVIDIA appears twice: as an investor and as a seller. Widths mix one-time equity with multi-year commitments, so read them as orders of magnitude, not one year of cash.
Funding rounds
Valuations climbing in months, not years
Every priced round we found for the tracked startups and superintelligence labs. Height is the post-money valuation on a log scale, bubble area is the amount raised. Blue rounds had NVIDIA in them. OpenAI ($122B at $852B, March 2026) is left off because it would flatten everything else.
Founders
Where the founders came from
Public professional background only: schools, employers and companies, each with a source. A filled dot means they worked there, a ring means they studied there, a square means their company was incubated there. Shaded columns are big tech companies and OpenAI.
Profiles
Bubble debate
Is it a bubble?
The strongest evidence on both sides, from the last few weeks where possible. The spending is not in dispute; the argument is whether revenue can catch up before investors lose patience.
2026 AI capex guidance
Billions of US dollars, low to high end of company guidance. Oracle is fiscal 2027. Source: Goldman Sachs via Tech Times, 28 Sep 2026.
Spending vs payback
Goldman Sachs estimate for the five largest US hyperscalers, billions per year.
Signs of a bubble
Signs of real demand
Where the evidence lands
Both sides are partly right. Demand at the chip and cloud layer is real and growing fast. The fragile parts are private valuations far ahead of revenue (Instinct, the superintelligence labs), demand that the sellers finance themselves, and dependence on two customers, OpenAI and Anthropic, that still lose money. A May 2026 study reached the same split verdict: a real technological revolution with localized bubbles.
What to watch next
Superintelligence
Billions raised before a product
A new set of labs says it is building superintelligence, AI that would exceed people at most intellectual work. They are led by researchers from OpenAI, Google DeepMind, Meta and Salesforce, and they raise seed rounds above $1B. NVIDIA is reported as an investor in three of the five.
Who funds them
Meta Superintelligence Labs is funded from Meta's own balance sheet.
Latest valuations
None of these labs has disclosed revenue.
Evidence log
Every claim, with its source
All links, rounds and market claims in one list. Filter by type or confidence, or search for a name.