The leaked Anthropic (ANTH.PVT) IPO prospectus has once again reminded investors of the enormous cost of developing AI. The AI lab plans to spend $518 billion in the coming years on cloud, compute and infrastructure costs, according to Reuters, after posting a net loss of $42 billion in 2025.
Successful IPOs for Anthropic and OpenAI (OPAI.PVT) depend on optimism that their customers’ revenues will eventually exceed those expenses.
Two new analyzes call these calculations into question.
One of the latest comes from Torsten Sløk, chief economist at Apollo. In the future, customers of AI labs will need to earn more money to pay for AI labs. But analysts do not predict such an increase. (Disclosure: Yahoo is a holding company for funds managed by affiliates of Apollo Global Management.)
“Wall Street stock analysts work in sector silos, and when you add up their forecasts, the numbers are internally inconsistent,” Sløk wrote in Tuesday’s Daily Spark.
“Analysts covering technology expect the sector’s operating cash flow to more than double to around $2.4 trillion by 2028, an increase of more than $1.2 trillion. Meanwhile, analysts covering other S&P 500 sectors, which are technology’s customers, expect these companies to add significantly less operating cash flow,” Sløk wrote.
Read more: Apollo’s Sløk: is an ‘agent bank run’ coming?
“In other words,” he continued, “the technology silo is betting on a future in which demand for AI and technology services will explode, while the silos covering the companies that would pay for these services see much more modest prospects. Both cannot be right at the same time.”
“Ultimately, either technology customers will generate much more cash than their analysts predicted, or the technology sector’s cash flow forecasts are too optimistic, raising the question of who exactly will write all those checks to buy AI services.”
Likewise, a new report from Bain Capital predicts a significant funding gap between the spending needed to develop AI and the revenue these companies will generate.
“If we assume that capital spending represents approximately 25% of industry revenue (an ambitious but reasonable percentage based on trends among cloud providers), sustaining this level of investment would require an AI market approaching $6 trillion per year,” wrote the team led by Bain Silicon Valley partner David Crawford.
The problem is that the consumer and enterprise AI market will likely reach between $1.2 trillion and $1.8 trillion, according to the Bain report, leaving a gap of $4.2 trillion.
Gn bussni
