THE AI CAPEX BOOM 💥
THE NUMBER
$182 Billion
The Combined Long-Term Debt Of Alphabet And Meta Platforms, Two Major Hyperscalers, Was $182 Billion As Of June 30.
At the end of Q2 2022, before OpenAI released ChatGPT, Alphabet’s long-term debt totaled $14.7 billion, with Meta Platforms carrying zero debt. The figure has grown by 1,134% during that time, compared to a 124% rise in their consolidated operating cash flow. Shareholders are now accepting greater financial risk as the hyperscalers increasingly borrow to spend relentlessly on AI infrastructure.
THE VISUAL
Note: Historical data is shown for each company’s most recent fiscal quarter (Alphabet: Q2, Micron: Q4) as of October 14 that year. Estimates are for each company’s next four fiscal quarters combined.
Source: The Blind Spot / Data: Company Filings, Fiscal.AI
Alphabet’s capital expenditures, projected to reach $200 billion in 2026, now exceed operating cash flow. On the other side of that spending, Micron is benefiting from the memory shortage, supporting pricing power. The question for allocators is durability: Alphabet is investing ahead of returns that are negatively impacting its FCF, while Micron’s windfall rests on a supply squeeze in a historically cyclical industry.
THE QUOTE
“If we end up misspending a couple of hundred billion dollars, I think that that is going to be very unfortunate, obviously. But what I’d say is I actually think the risk is higher on the other side.”
The CEOs of Amazon and Alphabet echoed this perspective that falling behind what could be a generational technological platform shift is unacceptable, so money is no object when trying to become established as a leader in AI. The outlays are exploding, with Vanguard calculating top-five hyperscaler capital expenditures approaching $800 billion this year and $1.1 trillion in 2027 (up from $416 billion in 2025). Even if AI ultimately delivers, a multi-year lag between spending and monetization could test investor patience.
THE MENTAL MODEL
Second-Order Thinking
Applied To The AI CapEx Boom
Because of how massive the numbers are, the market’s attention goes to the amount of money being spent today. But it’s critical to think about the downstream impact that the capital expenditure supercycle will have on the income statement. According to BCA Research, the top-five hyperscalers will incur $581 billion in depreciation and amortization expense in 2029, almost six times higher than the total from 2022, introducing a major earnings headwind.
THE QUESTION
How will the AI infrastructure supercycle play out in the next five years?
(My take italicized in bold)
Keeps accelerating (revenue grows into the build)
Governments take over (sovereign AI becomes the marginal buyer)
Growth simply stalls (spending plateaus as replacement capex): Railroads, utilities, and telecom networks never stopped spending; they each moved from building out to replacing worn assets. GPUs can become obsolete within a few years, so the hyperscalers will have to keep spending heavily just to stand still. Most infrastructure booms suffered a bust along the way, but even then, spending settled into a steady replacement cycle rather than disappearing.
Credit runs out (lenders pull back first)
Returns disappoint (spend outruns payoff)
Reply with your answer, and why
Until next week,
Neil Patel
One person in your network needs to read this. Forward it to them.
