# AI capex — X 热门讨论 (2026-09-24 23:48 UTC)
## @GlobalMktObserv (Global Markets Investor) · 09-24 18:23 · ♥33 ↻17 💬7 🔴US Big Tech credit spreads are blowing out:
Hyperscaler 5-year CDS spreads have surged to new all-time wides, led by Oracle, $ORCL, SpaceX, $SPCX, Nvidia, $NVDA, and Meta, $META.
Oracle’s 5-year CDS surged above 220 basis points, surpassing its previous peak during the 2008 Global Financial Crisis, as investors demanded significantly more protection against rising credit risk.
The move comes as hyperscalers are taking on massive amounts of debt to fund the AI buildout, with Amazon, Alphabet, Meta and Oracle issuing ~$195 billion of bonds in H1 2026 alone, up ~80% from all of 2025.
Bond issuance from 5 hyperscalers is expected to reach $250 billion this year and $400 billion in 2027, versus just $16.7 billion in 2024 and $13.7 billion in 2023.
At the same time, hyperscaler free cash flow is collapsing as AI capex accelerates, raising questions over how quickly these investments can generate returns while debt continues to rise.
Importantly, wider CDS spreads do not necessarily mean investors expect imminent defaults, but rather reflect growing demand for protection against a deterioration in the credit quality of AI-related companies.
AI credit risk is skyrocketing. https://x.com/GlobalMktObserv/status/2103188468872380716
## @KyleReidhead (Kyle Reidhead | Milk Road) · 09-24 14:30 · ♥34 ↻13 💬8 AI lab revenue is catching up to AI capex, FAST
AI Lab revenue is accelerating, while Capex is decelerating
In 2024 the labs made $5B against $241B of hyperscaler capex (2 cents per dollar spent). Altimeter has it at $1T vs $1.3T by 2029 (77 cents)
Here's the only rule that matters for the AI trade: As long as revenue is growing faster than capex, there's no problem
And it's not even close right now:
2026: capex up ~87%, lab revenue up ~5x 2027: capex up ~38%, lab revenue up ~3x 2027 to 2029: capex up ~20%, lab revenue up another ~3x
Capex is decelerating while revenue compounds. The gap is closing in from both sides (which is exactly what you want to see)
"But they're borrowing to pay for it!"
Yes. Alphabet, Amazon, Meta and Oracle issued ~$223B of bonds this year as of August, more than double all of 2025
That's not a red flag, that's how every major infrastructure buildout starts. Railroads, the electric grid, telecom, all built on debt BEFORE the revenue showed up. Some debt is expected when you're building the rails for a new economy
The debt only becomes a problem if the revenue line stalls
So that's the thing to actually watch. Brad Gerstner said the labs need to hit a ~$180B run rate by year end to keep the AI trade intact. They were at roughly $100B in July, with Anthropic alone at $65B
My guess is they blow through it. Agents are just getting started and the labs still can't get enough compute to serve the demand they already have
Don't forget to give me a follow @kylereidhead for more insights on AI and markets https://x.com/KyleReidhead/status/2103129856250876039
## @MilkRoadAI (Milk Road AI) · 09-24 20:34 · ♥32 ↻8 💬7 Bill Ackman made the perfect bull case for Meta months before even Muse arrived.
At the time, the market looked at Meta’s massive AI spending and assumed Mark Zuckerberg was lighting money on fire.
However, @BillAckman argued that investors were asking the wrong question.
Investors should not focus only on how much Meta was spending but rather on why it was spending.
If Meta had doubled its capital expenditures simply to protect its existing business, then the stock deserved to fall.
But if the company was investing in AI infrastructure that could create new products and generate attractive returns, the spending was growth capex rather than maintenance capex.
But now Muse made that argument much easier to understand.
Meta used its AI infrastructure to launch a personal agent that could send emails, book travel, make purchases, and complete other tasks through its own app or WhatsApp.
Instead of building distribution from scratch, Meta could place Muse directly inside an ecosystem already used by billions of people.
Muse also introduced paid subscription tiers, giving Meta another potential revenue stream beyond advertising.
Muse may only be the beginning because if Meta can keep turning AI infrastructure into new products, that massive capex bill starts looking a lot more like an investment than an expense. https://x.com/MilkRoadAI/status/2103221542423511049
## @htsfhickey (fred hickey) · 09-24 18:36 · ♥36 ↻3 💬1 More examples of AI "Superintelligence" (slop) in today's WSJ:
When they can’t effectively capture patterns and nuance, models rely on rules-of-thumb understanding, such as "cooked shrimp appear roughly circular and have banded striations,” Schossau said. The result: “Shrimp that look like otherworldly fleshy rings.”
But Tuths’ disgust also stemmed from complicated feelings about AI. “It just feels like here’s a few good uses, here are 100 other uses that are maybe unnecessary and are maybe decreasing quality of life.”
Me: Yes, there are some good uses of AI (better search, faster computer code writing), but so far, not even close to the amount needed to justify the trillions of dollars of capex (mal)investment being spent by way too many model providers. https://t.co/D9WLHICSCB https://x.com/htsfhickey/status/2103191953353629914