Apollo Chief Economist: A Trillion-Dollar Math Contradiction at the Heart of the AI Bull Case
nashnova research
Apollo chief economist Torsten Slok argues that tech-sector analysts forecast operating cash flow doubling to $2.4 trillion by 2028 — but the analysts covering their customers project far less growth, exposing a trillion-dollar gap that calls the entire AI valuation framework into question.
What exactly is this "math contradiction"?
Analysts covering the tech sector expect operating cash flow to more than double by 2028, adding over $1.2 trillion to reach roughly $2.4 trillion.
But analysts covering the rest of the S&P 500 — the companies that would actually pay for AI services — project far more modest cash-flow growth.
This means → the sell side bets on explosive revenue, while the buy side's own analysts don't see customers generating the money to pay for it. In plain terms = one group forecasts "the product will sell at a premium," the other forecasts "the buyers can't afford it." Both cannot be right.
Slok traces the blind spot to Wall Street's silo structure: analysts work within their own sector and never reconcile the numbers across sectors.
Where will hyperscalers find the cash?
Hyperscale cloud providers — Amazon AWS, Microsoft Azure, Google Cloud and peers — have already tipped into negative free cash flow due to massive capital spending.
Analysts expect the group to generate over $120 billion in free cash flow by 2030. But Slok notes that maintenance capex on data centers alone will run to $920 billion over the coming years — before any other costs.
This means → meeting that forecast requires revenue growth that vastly exceeds a known, hard-dollar obligation — and that obligation is already enormous. In plain terms = just "paying the upkeep" on data centers already built costs $920 billion; where does profit come from?
Is pricing for AI inference getting worse?
Over the past three months, enterprises and individual users have shifted en masse to Chinese open-source models, and pricing for frontier-model compute has dropped sharply.
This reflects intensifying supply-side competition squeezing hyperscalers' monetization — heavy infrastructure investment meets falling per-unit revenue.
In plain terms = they spent heavily to build "AI factories," but the product price keeps falling, stretching the payback period.
Everyone talks about AI — why doesn't that equal real spending?
Slok argues that the ubiquity of AI discussion may itself be a bubble signal, not proof of fundamentals.
Companies talk about AI for three reasons: genuine adoption, investor hype, or plans to cut costs and headcount — but so far, none of this has translated into measurable productivity gains.
In plain terms = "everyone talking about AI" and "everyone spending on AI" are two different things, separated by a conversion gap from rhetoric to purchase orders.
What does this gap mean for markets?
Slok's framework exposes the core fragility of the AI bull narrative: between demand-side ability to pay and supply-side revenue expectations sits a trillion-dollar gap.
This means → the gap is not yet fully priced by markets — the validity of current AI valuations hinges on whether it can be closed.
Only two outcomes remain: either tech companies' customers generate cash flow far above current forecasts, or the tech sector's earnings projections face a major downward revision.
市场有风险，内容仅供研究参考，不构成投资建议。