Two Key Numbers

The stock market is entirely being driven by tech/AI names again. The chart below shows the performance of the S&P 500 overall vs. the S&P 500 excluding technology and other stocks that directly benefit from the AI buildout (non-AI stocks are roughly 55% of the stock market now).

What’s critical to understand about the AI industry is that it’s heavily dependent on the financial health of the two major AI labs, OpenAI (maker of ChatGPT) and Anthropic (Claude). More than a trillion dollars have already been spent by companies like Microsoft, Google, Amazon, and Oracle (known as the “hyperscalers”) building AI datacenters to mostly serve just these two companies, which, combined, are estimated to account for around 70% of total AI datacenter industry revenue (to be fair the hyperscalers also use these datacenters internally).

Right now the projections are that a mindboggling $6-$8 trillion more will be spent over the next five years. This spending is predicated on the assumption that the two major labs will actually have the money to pay for what’s being built on their behalf. Already they’ve committed to spending over a trillion dollars on future computing capacity. But to do so their revenue must keep growing at an extraordinary rate. These two numbers – OpenAI and Anthropic’s monthly revenue figures – are the two most critical data points in the financial markets right now.

The trouble is that since both companies are private, we really don’t know exactly how well they’re doing and analysts have had to cobble together estimates based on random management comments and figures leaked to the press. The following chart is based on these estimates:

So in the month of September 2026, OpenAI is estimated to have generated around $5.7 billion in revenue and Anthropic $6.3 billion – or $12 billion combined. If we assume that amount of revenue holds for an entire year, we’re looking at roughly $144 billion in total revenue annualized (reflected by the green line in the chart above). However, as noted above, the two companies have made financial commitments with the hyperscalers and chip companies to spend over a trillion dollars in the next 5-7 years. There are two major risks that they may not be able to fulfill these obligations:

1) They lose market share and pricing power faster than the industry is growing – ChatGPT and Claude are among the best AI models in the world. But cheaper open-source models like China’s DeepSeek are increasingly dominating the market (now about 75% of total usage).

OpenAI and Anthropic have continued to grow revenue despite losing market share, as the whole industry continues to grow rapidly but there is no guarantee this will continue. Moreover, the heightened competition is leading to a race to the bottom as far as token pricing goes (tokens are units of text that an AI generates, which AI models makers bill by). Token prices have been cut in half since June, for instance, as the open-source models have gained prominence.

2) AI usage growth overall slows – This is counterintuitive given the hoopla surrounding the technology, but so far anyway, AI has not had a measurable effect on businesses’ productivity.

If businesses struggle to find enough productive uses for the technology, we could see token usage growth slow, which would only add to the industry’s pricing pressure and make it even more difficult for OpenAI and Anthropic to fulfill their gargantuan financial obligations.

AI-related stocks will likely continue to work as long as OpenAI and Anthropic’s monthly revenue growth keeps surging. Should that growth falter, however, the AI trade currently driving the stock market is likely to be quickly unwound.