As OpenAI would have it, interest in Generative AI is going to to continue to go up and up until they make so much revenue their annual revenue exceeds the annual revenues of companies like Ford, BMW, and Bank of America. On a path to make $13 billion last year, they are projecting $200 billion in revenue in 2030, according to a September 5 story at The Information.
Which they would need, because as the same story reports, training costs are immense. On computing costs alone, they are expecting to burn $150 billion between 2025 and 2030. They are burning money, billions a month; per the Information’s headline, “OpenAI Says Its Business Will Burn $115 Billion Through 2029”. And schoolchildren using ChatGPT to do their homework aren’t enough to cover that.
Needless to say, OpenAI can’t make its ambitious revenue numbers — 13x-ing current revenue in five years — unless revenue from paying big business customers keeps going up and up and up and up and up.
The Information was nice enough to capture that all optimism in a chart:
OpenAI’s revenue projectionss in one handy graph (left panel represents an earlier projection; the right panel represents the latest projections)
Will revenue really continue to go up and up and up at the pace they project?
Maybe yes, maybe no. Until recently things were looking rosy.
But new Census Bureau data, captured in a graph that Apollo Academy just shared, hints at a very different picture:
Slok elaborates in his newsletter:
Is that decline among big companies over the last two months just a blip? Or could the big companies have finally lost patience?
GenAI certainly sounds amazing – but it hasn’t always worked well enough to justify the costs.
Disappointment would certainly fit with the recent MIT study showing much disappointment at ROI.
The underwhelming GPT-5 results can’t help.
If things go up and up until they don’t, OpenAI is toast.