Half the S&P's Earnings Gains This Year Are Just Cloud Giants' AI Spending
Goldman Sachs calculates that roughly half of the S&P 500's earnings growth this year comes directly from cloud giants' AI capital spending; the market is betting not on today's sales, but on whether this money converts to revenue in three years.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Breadth collapse mirrors the internet bust, but does not reliably predict tops
The S&P 500 is only 0.5% from all-time highs, yet just 51.2% of its constituents trade above their 200-day moving average—the last time the market was this close to records with so few stocks participating was the day after the internet bubble burst. But Luke counters with precedent: in 1998, breadth was equally dire, yet the S&P went on to gain for two more years. Breadth alone does not determine tops.
— Luke KawaConsumer stocks are now the proxy for betting against AI exuberance
Nike crashed from $180 in 2021 to $35; McDonald's, Walmart, and Costco face similar valuation compression. Luke's theory: these stocks became proxies for shorting software companies expected to be disrupted by AI. Walmart's steady monthly throughput resembles subscription ARR; Costco's membership fee is essentially recurring revenue. When AI stocks surge, the market needs something to punish.
— Luke KawaHalf of S&P earnings growth this year is cloud capex itself, not its monetization
Goldman Sachs estimates that roughly half of the S&P's earnings growth this year comes directly from cloud giants' AI capex spending. Valuations now hinge not on next year's sales expectations but on three-year-ahead sales. Economist Hanno Lustig goes further: to fully justify this spending, a handful of companies must eventually capture 9% of GDP in revenue—both Luke and Joe call this absurd. Nvidia has begun separately listing sovereign AI and robotics revenue streams in its earnings to argue the addressable market is broader than cloud capex alone.
— Luke KawaRising bond yields reflect surging growth, not uncontrolled inflation
Luke attributes the 10-year Treasury yield breaking 5% primarily to strong global growth pushing real rates higher, not inflation expectations spiraling. Breakeven inflation rates remain stable. This surprises him: the Fed has failed to keep inflation near its target for years, so you'd expect markets to discount central-bank credibility. Yet breakeven rates show no such skepticism.
— Luke KawaMarkets have moved too fast for shorts to build effective hedges
Luke sums up last year as TINA (There Is No Alternative); this year, it's prices and earnings rising too fast to hedge effectively. AI is compressing the timeline of economic events; add AI agents trading while humans sleep, and the market operates at an unprecedented pace. Robinhood retail net-buy data show peaks in June (around Micron earnings) and October last year (around Palantir, bitcoin, quantum compute hype), with September recovering to 89% of June's peak.
— Luke KawaNegative free cash flow at giants demands increasingly creative financing tactics
Luke argues that once cloud giants' free cash flow turns negative, two paths remain: slow AI buildout, or get creative with financing. The latter is happening. He links this to Minsky moments: as long as the debt looks safe (cloud balance sheets, federal deficits), valuations hold. But that's exactly how "safe" tips into "unsafe."
— Luke KawaThis AI expansion is built on debt, not on AI-generated cash
Luke borrows from Wall Street Journal's Greg Ip: past internet giants funded expansion with their own cash machines—Facebook's ads, Google's search. This time is different. Oracle has no pure cash cow; it must borrow. Credit default swaps keep rising. Nvidia even steps in to boost credit for peers, trying to make all AI borrowers look equally safe.
— Luke KawaAI has replaced real estate as the driver of economic cycles
Luke notes that wealth gains from equities now exceed those from real estate, a watershed moment. Real estate defined the business cycle because HELOCs and wealth effects flowed into consumption; the same logic now applies to AI—capex surges and wealth-effect surges emanate from here. He calls this abnormal: ultimately, downstream consumers must buy or the cycle breaks.
— Luke KawaIn their own words · checked verbatim
The last time we had that proximity to a high with that few stocks above the 200 day was literally the day after the dot-com bubble peak. So that gets a lot of people mad, but then you have to point out, okay, wait, breadth was also that bad in 1998. The market ripped for another two years after that.
Luke Kawa4:01
Nike is the classic example. That's crazy. That was $ 180 stock in 2021. It's 35 now.
like a handful of companies will have to get revenue that's like 9% of GDP in order to justify it, which is just like unbelievable.
Joe Weisenthal13:12
This is too fast to fade both prices and earnings, just moving too fast to fade. And speed is something that I think has completely defined 2026.
Luke Kawa20:18
Either the AI boom slows or the financing starts to get more creative. The financing has gotten more creative.
Luke Kawa29:30
No, I mean, nothing against Oracle. It's a good, successful company, but it does not have that pure money spigot the way the other hyperscalers had.
Luke Kawa32:00
If AI kills us all by setting off a nuclear weapon, our portfolios will be on fire. And you'll feel stupid for not having gone out to eat. If AI doesn't kill us all, then our portfolios will also be on fire.
Luke Kawa40:41
Figures
| S&P 500 year-to-date gain | up 12% or more | 3:00 |
| S&P 500 distance from all-time high | 0.5% | 4:01 |
| S&P constituents above 200-day moving average | 51.2% | 4:01 |
| Major cloud giants' investment-grade debt issuance: September | 0 | 6:04 |
| Nike stock price | $180 in 2021 to $35 | 8:07 |
| Portion of S&P earnings growth driven by cloud capex | approximately half (Goldman Sachs estimate) | 9:08 |
| Revenue as share of GDP needed to justify AI capex | 9% | 13:12 |
| 10-year Treasury yield | broke above 5% | 13:12 |
| Robinhood September retail net stock buys | 89% of June peak | 22:19 |
| Semiconductor stocks hitting 52-week highs in Q3 | only 2 (AMD, Skyworks) | 31:34 |
Glossary
- TINA (There Is No Alternative)
- Investment stance that stocks are the only viable asset class with acceptable returns
- Breadth / Market breadth
- Metric measuring how many individual stocks participate in an overall market rise
- Negative FCF (Free cash flow)
- Operating cash flow minus capex is negative; growth relies on external financing
- CDS (Credit default swap)
- Derivative contract that prices the risk of corporate bond default
- Minsky moment
- Sudden reversal of debt-fueled boom, triggering a wave of selling and credit unwinding
- Credit wrap
- Strong-credit party guarantees debt of weaker party to reduce perceived default risk
How to listen
Investors and buy-side analysts watching U.S. equity valuations and AI capex sustainability, plus anyone trying to understand why consumer stocks keep getting sold.
32:36–34:38: discussion of Manchester City financial violations and football league tangent unrelated to market mechanics; can skip.