Valuations Are Falling for a Reason: AI Is Repricing the Future

Описание: In this week's video, I make the case that what's happening in software is not a rotation or a cycle — it is a structural regime shift that will define markets for the next decade. The S&P is flat year-to-date, yet 185 stocks have posted greater than 15% absolute moves in both directions, double the 81 names we saw on this same day last year. That level of dispersion has only been matched twice in 30 years: during the dot-com bust and the GFC. In both cases, credit spreads eventually widened. We're not there yet but the warning signs are multiplying.
The SaaS thesis is simple: you cannot forecast revenues three years out in a world where software can be created in seconds, where entrepreneurs are building bespoke solutions with AI rather than buying enterprise licenses, and where Chinese open-source models are already capturing 47% combined AI market share globally at 80–90% lower cost than U.S. frontier models. Anthropic's gross margins are forecast to fall to 40% even as it wins the inference race. Oracle faces a shareholder lawsuit for committing to $5 trillion in compute capex against revenue projections that may never materialize. This is not an AI bubble it is a deflationary spiral driven by hyper-competition, and it is eating the long-duration asset base from the bottom up.
Credit is the canary. Blue Owl halting redemptions, private equity down six weeks in a row, software OAS spreads diverging sharply from CDX high yield — this is where the next systemic fear comes from. If software debt losses spike, the knock-on to hyperscaler capex plans could be severe. Bitcoin remains correlated to SaaS for now, but I continue to believe it leads the next move higher when SaaS finally finds a floor it is the only growth asset AI cannot disrupt.

Timestamps
• (00:00–01:43) Macro outlook for the year: market finishes up but nowhere near consensus; multiple compression in software is the dominant theme; covariance matrix breakdown is a new regime, not a cycle
• (01:43–03:48) Dispersion explosion: 185 S&P stocks with more than 15% absolute moves YTD vs. 81 last year; structural shift in software PE; picking a SaaS bottom = AI bubbleista positioning
• (04:09–06:27) Supersonic tsunami recap: Phase 1 of AI (2022–2025) concentrated in digital brain builders; transition to physical infrastructure, commodities, and hardware is a multi-decade story
• (06:48–08:35) Regime framing: asset-light giving way to asset-heavy; SaaS is a value trap; Palantir preferred over Microsoft; Bitcoin leads when SaaS finds a bottom
• (08:55–10:22) Bitcoin phase analysis: deflation thesis vs. Raoul Pal/Andreas Steno; NASDAQ must underperform for Bitcoin to work; government cannot spend enough to offset exponential deflation
• (13:00–15:43) Valuation framework: Mauboussin CAP model applied to AI disruption; bifurcation in valuations = dispersion; concentration risk parallels dot-com and GFC
• (17:27–20:56) Credit risk building: tech sector OAS diverging from CDX; Blue Owl halts redemptions; private equity/credit/VC all at risk as long-duration assets face hyper-competitive disruption
• (22:00–24:53) Trade ideas: edge device analog basket poised to break out; durable goods/capital goods PMI headed to 60; memory demand flash wall approaching
• (31:30–33:39) Hyperscaler risk: Oracle lawsuit; Anthropic margin compression; Chinese models 80–90% cheaper creating deflationary spiral in inference pricing
• (35:29–39:42) China AI dominance: DeepSeek/Qwen from 1% to 15% global share in one year; 80% of open-source startup stacks running on Chinese models; OpenClaw accelerates adoption
• (40:23–46:19) Agentic world arriving: OpenAI hires OpenClaw founder; Grok agents live; Nvidia key to S&P stability through earnings; call to build AI mindset