June was one of the ugliest months in years for the mega-cap AI leaders and billionaire investor Bill Ackman chose exactly that moment to plant his flag on the bull side.
On Tuesday, Ackman amplified a widely shared post on X by analyst Rihard Jarc, who writes the UncoverAlpha newsletter, adding that it was “an important read on the hyperscalers and why we are bullish.”
The call came as Microsoft Corp. (MSFT) logged its worst month since December 2000, falling 17.2% in June and Oracle Corp. (ORCL) fared even worse, plunging 35% and notching the stock’s worst month since September 1990.
Other famous hyperscalers such as Amazon.com Inc. (AMZN), Meta Platforms Inc. (META) and Alphabet Inc. (GOOGL) fell 11.9%, 10.6% and 6.1% respectively.
Why Hyperscalers Sold Off
The trigger was one word: capex.
The five largest hyperscalers — Amazon, Microsoft, Alphabet, Meta and Oracle — are projected to spend over $700 billion in 2026 on AI infrastructure.
Investors have decided that spending, not revenue, now drives the stocks.
In a recent note, Bank of America highlighted hyperscaler capex has climbed from about 70% of operating cash flow in 2025 to nearly 100% in 2026 — leaving little left over for buybacks and dividends.
Bank of America’s data shows the group’s forward free cash flow as a share of net income collapsing from well above 100% toward near zero, even as the rest of the S&P 500 holds around 90%.
The market has decided the AI buildout is eating the cash that once funded buybacks and dividends — and it sold the stocks accordingly.
The result: the Magnificent Seven shed roughly $2.3 trillion in market value in June.
Why Ackman Is Bullish Now
The selloff was, at its core, a verdict on spending and Ackman leaned directly into that fear.
The billionaire investor is backing a thesis from analyst Rihard Jarc, who argues the market is beating up the wrong part of the AI stack.
The hyperscalers are the tollbooth, and they collect the same fee on every token that crosses, no matter which model produced it.
For 18 months, enterprises practiced “token maxing,” routing every workload to the single best, most expensive model.
Now the bills have come due, with some firms burning through a full year’s AI budget in a quarter, and the behavior is changing. The cheap, easy 80% of requests get routed to open-weight models, and only the hardest jobs go to the frontier.
That downgrade crushes what the model providers can charge. It also sends usage through the roof.
When inference gets cheap, nobody rations it: agents run in loops and chew through millions of tokens on tasks that used to cost pennies to think about.
And almost all of that traffic still lands on someone’s cloud — Amazon’s Bedrock, Microsoft’s Azure AI Foundry, Google’s Vertex — where the hyperscaler takes its cut regardless of which model did the work.
AWS has historically run a 35–38% operating margin, and Google Cloud cleared 33% in the first quarter of 2026. A dollar of that margin looks the same whether the token came from a $50-per-million frontier model or a $1-per-million open-weight one.
More tokens, cheaper each, running on data centers the hyperscalers already paid for: Jarc’s bet is that it eventually pushes the three stocks higher, not lower.
That fits how Ackman has been talking for months. He has said investors are piling into AI chip and energy names while overlooking cash machines like Amazon, Meta and Microsoft, the same way the market wrote off Berkshire Hathaway as “old economy” in 2000.
Ackman started a new Microsoft position after the stock dropped following February’s earnings.
