Key insights
- An analyst highlights a divergence between net income growth and declining free cash flow among big tech companies, driven by massive AI-related capital expenditures and off-balance-sheet data center leases. The analyst questions the sustainability of the AI buildout, suggesting potential renegotiation risks in lease contracts and a potentially low ROI compared to other investments like MLPs, signaling a bearish outlook for the sector.
The Economist just ran this on "big tech" free cash flows: https://www.economist.com/business/2026/05/13/big-tech-is-sacrificing-its-cashflows-to-prop-up-the-ai-boom
I love this line: "America's biggest companies have gone from printing money to burning it."
Because I have my trusty SEC/Edgar database, I can go more into depth on the math:
Across 43 big tech companies in 2025: — Net income grew $157B — True free cash flow shrank $10B — CapEx grew $170B
The market seems to be reacting to the income growth, but the cash flow isn't following.
The Economist also found something my screener doesn't have (still struggling to get quarterly data): $820B in off-balance-sheet data center lease commitments, up from $270B a year ago. A banker told them: "When we ask our lawyers to find ways a hyperscaler might wriggle away from a lease contract, often they come back with a very long list."
In plain English: the financing underpinning the entire AI buildout may be more renegotiable than the bond buyers funding it realize.
Back of napkin on the $800B 2026 CapEx projection: if 10% of Americans pay $1,000/year for AI and every Fortune 500 spends $100M/year, that's $80B in revenue on $1.1T in cumulative AI investment. 7.3% ROI assuming it all flows back to the hyperscalers, which it won't. I can already earn 6-10% on MLP energy investments. Real, tangible money.
Full piece here: https://cavemanscreener.substack.com/p/bridges-to-nowhere-part-ii-the-economist