The Certainty Premium: What’s Driving the Tech Sector Divergence? | Jeff Keller

Episode Summary

Executive Summary: Jeff Keller argues the AI trade is still in a secular uptrend but is moving from euphoric, retail-driven momentum into a more sober phase where fundamentals, valuation, and short-term certainty matter more. He sees software dispersion, strength in cyber/infra, ongoing hyperscaler capex, and rising retail participation as the key dynamics to watch, while cautioning that overbought, highly cyclical AI-adjacent names could suffer if adoption, compute pricing, or funding conditions deteriorate.

Main Topics: AI adoption remains an open-ended secular trend: Keller believes AI is still early in an S-curve, with adoption, capex, and usage continuing to rise. He thinks the recent volatility is normal for a large thematic trade rather than evidence that the trend has ended. Software rerating and sector dispersion: He splits software into application software versus infrastructure/cyber. Application software has rerated lower as growth expectations normalized, while infrastructure and cyber benefit from usage-based upside and short-term certainty. Hyperscaler capex as existential option value: He argues hyperscalers are not just making ROI-driven investments; they are buying optionality to avoid strategic defeat. That helps explain why capex can continue even if near-term free cash flow weakens. Retail participation and market psychology: Keller warns that heavy retail involvement often signals a late-stage trade, but he does not think AI has fully reached that point. He sees more sobriety than in 2021 and expects longer, slower cycles. Valuation, cyclicality, and short-term certainty: He emphasizes that many AI-linked names trade on very different multiples despite sharing the same drivers. In a market craving certainty, names with visible near-term growth command premiums, but that also raises risk. Long-biased, optional long/short portfolio construction: Keller describes his fund as long-biased rather than market-neutral. He prefers flexibility, low leverage, and the ability to add risk during dislocations instead of being forced to de-gross at the wrong time. Data center politics, compute pricing, and future risks: He is less worried about political or labor resistance to data centers than the market seems to be. The bigger risks are overbuild, falling compute prices, and whether adoption or lab ARR slows.

Key Arguments: AI is still in an S-curve; the recent drawdown is more likely a pause/rotation than a top. Software should be viewed in two buckets: application software faces disruption questions, while infra/cyber benefit from AI-driven usage growth. Short-term certainty is being rewarded disproportionately in the current market, which is why cyber and infra have outperformed. High retail participation often marks a trade nearing maturity, so increased enthusiasm is a caution signal even if the trend remains intact. Hyperscaler spending is better understood as buying strategic optionality than optimizing near-term ROIC. The market may be underestimating how much AI buildout can continue even with higher long-end yields because the strategic stakes are existential. The best AI outcomes may occur in lower-multiple names tied directly to the buildout rather than the most creatively marketed 'hidden winners'. Long-short investors should focus on one or two major opportunities a year, preserve optionality, and avoid being forced to trade short-term noise. AI-driven narrative cycles and AI-generated content may make markets less efficient by amplifying overconfidence and echo chambers. A public listing of Anthropic should reduce uncertainty by providing real metrics, even if it also attracts capital and could rotate flows within the AI complex.

Data Points: Software and services performance since Q3 start: up 20% - Used by the host to frame the recent divergence within tech sectors. Technology, hardware, and equipment since Q3 start: basically flat - Host noted this as part of the sector rotation. Semiconductors and semiconductor equipment since Q3 start: down 7.5% - Host cited this as the laggard within tech. Q1 AI-related stock gains: 100% gain in stocks - Keller referenced the early-year surge in AI-related names. Large language model / AI lab growth example: double digits for 10 to 15 years - Bull case Keller cited for Apple Services, Google Search, Meta, and AI labs. Cyber valuation: 25-30x revenue - Keller said large cyber companies have rerated to these levels. Cloud provider concentration: 3 vendors to 9 or 10 scaled providers - He described the cloud/compute market as becoming more competitive. AWS growth example: 40% - Keller said people used to pray for AWS to grow 18%, and now it is growing 40%. Hyperscaler / AI infrastructure raises year to date: about $500 billion - He cited major labs, SpaceX, and Google raising roughly this amount, with markets still functioning. Google adjusted free cash flow estimate for 2027: negative $18.8 billion - Host referenced estimates showing negative FCF before a later recovery. Google adjusted free cash flow estimate for 2028: positive $28.5 billion - Host used this to question certainty of AI capex returns. Keller's time horizon for valuation: 2 to 3 years - He said he generally underwrites through-cycle earnings over this horizon. Highly volatile sector examples: EWI 73 vol; SOX 55 vol - Host cited one-month realized volatility to illustrate market turbulence. Peak realized volatility examples: around 185 - Host mentioned prior peak volatility levels in the sectors.

Pivotal Quotes: "Anything with short-term certainty is trading at a huge premium." — Jeff Keller: Keller summarized the market's current preference for visible near-term growth and durability. "The real AI winners are CrowdStrike and some like that, rather than maybe some of the semis." — Jeff Keller: He argued that some cyber/security and infrastructure names may capture more of the AI value chain than the most obvious semiconductor names. "I think of it as you got to put $1,000 into the ground because if you don't, you might lose your whole business." — Jeff Keller: He explained why hyperscalers continue heavy capex even without clear near-term ROI.

Implications: Listeners should expect continued AI-capex spending, but with more scrutiny, dispersion, and volatility. The easiest gains may be behind us; winners will likely be those with real adoption, pricing power, and strategic optionality.