OpenAI and the AI bubble: revenue, commitments and funding

Why does OpenAI sit at the centre of the AI bubble debate?

OpenAI builds widely used AI models and sells access to them to consumers and businesses. Many other companies plan their spending around its success: cloud providers that sell it capacity, chip makers that supply those providers, and software firms that build on its models. A change in its fortunes would therefore travel along a long chain.

It is also private. There is no daily share price, so the market signal arrives through funding rounds, contract announcements and the figures the company chooses to share. The tests below treat each of these as a dated claim and ask what supports it. This habit keeps the reading calm, because each claim gets weighed instead of repeated.

How should you read the revenue figures of a private model builder?

Treat each figure as a claim with a date and a source. Annualised revenue takes a recent month and multiplies it, so it can overstate the full year when growth is uneven. Prefer figures that state the period, the definition and whether they come from the company or from a third party.

Then split revenue by type: consumer subscriptions, business contracts and developer usage. Business contracts renew on a schedule and give visibility. Consumer subscriptions can move faster in both directions. A mix that shifts toward renewable business contracts supports the case for stable growth. Ask also how many customers pay at the top end, since a broad base of paying users matters more than a headline user count.

What do computing commitments mean for the AI bubble question?

A model builder needs computing to train and run its models, and it signs multi-year agreements to secure it. These commitments are fixed obligations, and they sit against revenue that depends on growth in usage. If usage grows as planned, the commitments look sensible. If it lags, they become a heavy weight.

Compare the total of announced commitments with the revenue run rate and with cash raised. The ratio shows how many years of current revenue the commitments represent. A large ratio is not a failure on its own, and it does show how much future growth the plan already assumes. Revisit the ratio each time a new agreement is announced, because it can change fast.

How is OpenAI funded, and why does funding matter for the bubble?

Private model builders raise money in rounds from investors, and each round sets a valuation. A valuation is the price agreed between the company and its buyers, which makes it a negotiated number. Higher valuations help raise more money, which helps pay for more computing.

This loop works while investor confidence lasts. The EQOS Asia AI study records valuations up ×4 in 18 months across the AI market (April 2026). Watch the size and frequency of rounds, who invests, and whether strategic investors are also suppliers, since that links funding to the orders it creates. A round led by outside investors with no commercial tie gives a cleaner signal than one led by a supplier.

Which companies rise or fall with OpenAI’s success?

Cloud partners that sell capacity, chip makers behind them, data centre builders and power suppliers all book revenue linked to model builders. Each of them lists large customers in its filings, so you can see the dependence in their own accounts.

Circular arrangements deserve a check. If a supplier invests in a customer and the customer then buys from that supplier, demand looks bigger than the customer’s own revenue supports. Ask for the net flow of money: how much comes from outside buyers who pay with their own income. That figure is the cleanest measure of real demand, and it is the one to follow over time.

Which five tests show whether OpenAI’s economics work?

Test 1: revenue growth against commitment growth. Test 2: the share of revenue that renews under contract. Test 3: cash raised against commitments due in the next 2 years. Test 4: the cost of serving each user against what the user pays. Test 5: the share of demand that comes from outside investors versus paying customers.

Score each test as green, amber or red on the date you read it. Three reds in one quarter is the signal to treat the bubble risk as high for this link of the chain. Keep the dated scores, so you can show how the reading moved. The EQOS Asia method page explains how such readings turn into a single score.

How can a business manage its reliance on OpenAI as a model provider?

If your products or processes rely on one model provider, treat it as a supplier risk. Keep your prompts and data in portable formats, test a second provider on a sample of real tasks, and avoid long commitments that grow faster than your own revenue. A contract with an exit clause and a usage-based price keeps your costs in line with your results.

This page is general information, not investment advice. It frames risk and leaves valuations and dates open. For a larger AI purchase in Singapore, ASEAN or Hong Kong, the Λ simulator gives a first score, and the 30-minute pre-diagnostic with EQOS Asia applies it to your case.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile