PodcastDavid Lin ReportThomas Hayes

Great Rotation Happening Now: Why Chip Stocks Are Falling First | Thomas Hayes

Hayes sees a rotation taking shape as semiconductor stocks weaken and market breadth remains poor beneath indexes near record highs. He points to AI’s potential to lower inference costs and improve productivity, while warning that some AI-linked bond projections depend on ambitious cash-flow targets. He expects smaller, value, and international stocks, as well as bonds, to attract interest; he also sees longer-term housing support and potential recovery in consumer discretionary.

SemiconductorsMarket BreadthTechnology StocksAI InfrastructureMarket ValuationsEquities

Preview · 13 more points in the app

Read the Full Summary

Key Takeaways

Hayes argues that AI’s larger gains may accrue to companies using it to improve productivity and margins, rather than only to AI infrastructure providers: he cites GXO Logistics’ warehouse automation and says suitable tasks could move to open-weight models at half the token cost while maintaining 99% efficacy. He also says CPUs can handle nearly the same inference work as GPUs at much lower cost.

Hayes sees semiconductor cyclicality as a risk to the chip trade: high memory-chip margins could draw Chinese, Korean, and Japanese competitors into adding capacity just as demand falls, repeating a shortage-and-oversupply cycle. He says the short-term semiconductor trade is over after sharp July declines, and expects a rotation toward international, small-cap, and value stocks like the one he compares with 2000–2006.

Hayes contrasts indexes near all-time highs, supported by a handful of AI stocks, with broad weakness since July 31: 89% of S&P 500 stocks were down more than 10%, and 54% were down more than 20%. Against that backdrop, he expects a potential rally beneath the indexes over the next six months and says $8 trillion in money-market funds could help fuel a broader rally, including in small caps if yields slow or decline.

TECHNOLOGY

Inference Economics — Hayes says CPUs can handle nearly as much AI inference work as GPUs at much lower cost. He also expects businesses to route suitable tasks to open-weight models at half the token cost while maintaining 99% efficacy, potentially doubling productivity and margins.

“I mean look when this game started at the deepseek moment what was the whole narrative for the first year you can't run you cannot run AI on CPUs it…”—

ECONOMY

Starter-Home Demand — Hayes says a 4 million-home deficit and millennials reaching age 35 support housing demand. If rates ease, he says builders may need to address affordability with smaller homes, such as 1,500-to-2,000-square-foot starter homes, or by building two houses on a lot.

“You still have a 4 million house, deficit…”—

Yield Divergence — Hayes says yields are rising globally, but for different reasons: US yields are rising alongside economic growth, while European yields are rising without growth. He cites nominal US growth of 8% but does not expect that pace to continue.

“So number one, yields are blowing out globally…”—

MARKETS

Chip Supply Cycle — Hayes says high memory-chip margins will attract Chinese, Korean, and Japanese competitors, adding capacity and potentially flooding the market with cheaper memory and semiconductor chips. He describes a recurring cycle: shortages prompt capacity expansion just as demand falls, putting pressure on stocks.

“High prices cure high prices. Did you think the Chinese, the Koreans, and the Japanese were going to sit back and say, "Oh, Micron, look at those margins. Congratulations…”—

AI Bond Exposure — Hayes says roughly half a trillion dollars of bonds depend on two frontier language models meeting their income projections. He argues that OpenAI and Anthropic would need free cash flow equal to twice the combined free cash flow of the economy’s other nine sectors to meet those projections.

“The problem is, David, is that you've got a half a trillion dollars of bonds that are predicated on two frontier large language models meeting their income projections.”—

Semiconductor Pullback — Hayes says the short-term semiconductor trade is over after a July decline of 35% to 60% by name, a dead-cat bounce, and renewed weakness. He says this does not mean AI productivity gains or broader market gains are over. Semiconductors corrected by at least 50% three times during the 1995–2000 technology rally, he notes.

“Okay. So in July the semiconductor trade was down depending on the name 35 to 60%…”—

FORECASTS

Rotation Beyond Growth — Hayes expects international equities to outperform US equities, small caps to outperform large caps, value to outperform growth, and bonds to attract bids. He compares the potential rotation with 2000–2006, when he says the equivalent of the “unmagnificent 493” dramatically outperformed as chips rolled over.

“That's going to change…”—

Broader Rally — Hayes expects a potential rally beneath the indexes over the next six months. Lower-weighted S&P 500 stocks could potentially double or triple while the indexes perform more subduedly than people expect. He contrasts this outlook with the preceding period of broad stock weakness beneath resilient indexes.

“And yet the indices held up by five or 10 AI stocks were at or near, you know, within the stones throw of all-time highs…”—

Long-Term Housing — Hayes says the housing thesis may not work over the next one to three quarters but expects it to outperform over the long term. He cites a 4 million-home deficit and millennials needing homes as they form families. If rates ease, builders may need to build smaller homes.

“and, and that's setting him up, you know, with a 3, 5, 10 year view…”—

Intro.

Hayes argues that memory and commodity semiconductor shortages naturally invite competition and added capacity, so chipmakers such as Micron face a familiar boom-and-bust cycle.

Semiconductor Selloff

Hayes says Chinese firms are already challenging US AI leaders, while open-weight models help companies cut costs; he cautions that ambitious cash-flow forecasts underpin large AI-related bond issuance.

Is China's AI Threat Growing?

Hayes says his fund has exited tech and sees the commodity side of the AI trade as overextended, while maintaining exposure through natural gas and Alibaba.

Why Hayes Is Exiting Tech

Hayes expects market breadth to improve even if major indexes lag, as out-of-favor stocks and defensive holdings gain; he highlights Crown Castle, healthcare, staples, and AI adopters.

A New Stock Market Rotation

Hayes sees rising demand for utilities, staples, healthcare, and bonds as investors seek defensive income, with Treasury yields increasingly competing against equity earnings yields.

Defensive Stocks and Bonds Rally

Hayes rejects comparisons between Freddie Mac’s decline and the 2008 housing collapse, and favors building-supply roll-up opportunities led by operator Brad Jacobs.

Housing: Another 2008?

Hayes attributes higher global bond yields to a mix of growth, inflation, oil-related pressures, and capital demand, but says an end to the Iran conflict could pull Treasury yields lower.

Why Bond Yields Are Rising

Hayes expects small caps to benefit from favorable manufacturing data, substantial money-market cash, easing yield pressure, and their ability to adopt AI quickly.

Small Caps Poised to Rebound

Hayes says weak consumer sentiment reflects inflation pressures even as employment and spending remain resilient, and sees business optimism as a possible leading positive signal.

AI Bubble: Is This 1999 Again?

Hayes sees semiconductor froth as distinct from AI’s longer-term productivity potential, favoring companies that apply AI to improve margins rather than firms financing or supplying the buildout.

Related Episodes

Important Disclaimers

Not Investment Advice: For education only — not financial, tax or legal advice.

AI-Generated: Made by AI from source transcripts and may contain errors.

No Guarantees: Past performance does not predict future results; investing carries risk.

Our Commentary · Request Removal