Walking The Thin Line: How AI Is Compressing The Tails
Why the intelligence age is functioning the same way negative real rates did in 2021 and how the obsolescence tax compounds faster than inflation
Today, I went deep on how AI is functionally retooling the production function of the entire economy. This is the same setup as 2021, with one substitution: the constraint is no longer monetary; it is technological. AI is forcing capital into equities the same way negative real rates did four years ago. Alfie Kerswell joined the second half to walk through the equity valuation identity and how cross-border flows are transmitted through the rate complex. Today’s IGV up 200bps while SMH down 100bps was the cleanest expression of the equity long short positioning unwind in real time.
The paid subscriber section is at the bottom of this report, where I cover the proprietary research.
LIVESTREAM RECORDING FROM TODAY:
Today’s Livestream: Main Talking Points
1. The world is short vol. Non-participation is the greatest risk in a system where change compounds faster than capital can react. In 2021, holding cash meant losing 22 percent of purchasing power as real rates went negative. The 2026 analog is the obsolescence tax from AI. If you are not integrating AI into your business, portfolio, or capital allocation framework, you are functionally short the system. The compounding rate of AI capability is faster than the rate at which holding cash got punished four years ago.
2. The intelligence age is the new negative real interest rate. A six month software build now takes three weeks. That changes how much capital companies need, how fast they can deploy it, and how quickly they take market share from competitors who do not adapt. Same mechanism as 2021, different input. In 2021 it was monetary. In 2026 it is technological. The output is the same: capital is forced into equities because non-participation has a compounding negative return.
3. AI is no longer a tool inside a production function. It is the new production function. The gap between Mag 7 and the broad S&P has blown out since the ChatGPT launch because the market is rewarding companies rewriting their entire production stack around AI. Palantir is the cleanest example as the one stop shop for non technical companies to integrate AI into their decision systems. Q4 2025 commercial bookings up 103 percent. Palantir is the picks and shovels for the production layer rewriting.
4. The 600 billion dollar question is whether the AI capex prepayment ever pays off. NVIDIA reported quarterly data center revenue of 47 billion. Cumulative capex commitment across the hyperscalers is closer to a trillion. Actual revenue from AI services is roughly 50 billion. If the run rate from OpenAI, Anthropic, and the rest never catches up, the entire valuation premium across the sector unwinds. CoreWeave is the cleanest expression of this trade because the stock is the wedge between forward GPU lease commitments and ultimate AI revenue.
5. The obsolescence tax is faster than inflation. Companies and individuals not adapting to AI bleed at a higher rate than holding cash bled in 2021. The Goldman Sachs AI beneficiaries versus AI at risk index is the cleanest visualization. The spread is at extremes since 2025 because the market is pricing the devaluation of capital not exposed to the AI cycle. If you are not on sides, you are losing money in real terms even with positive nominal real rates.
6. AI is no longer a financial cycle. It is economic statecraft. Every country needs its own AI production stack, which transmits directly into geopolitical risk. The semiconductor index at highs while the Shanghai composite is at lows is the cleanest expression of the AI arms race priced into asset markets. China actively wants to close that spread, which raises the probability of more aggressive geopolitical action across Taiwan and the broader supply chain. The TSMC chokehold transmits directly into hyperliquid, perpetuals, and cross border flows.
7. The credit cycle melt up always sows the seeds of its own demise. We are walking the thin line right now. Real rates near zero, AI capex build out, equity valuations at the highest level in US history on price to sales, S&P dividend yield at lows last seen in 2000. If AI capex run rate disappoints, valuations get crushed and the IPO window closes. If real rates turn negative on top of the current cycle, you get a 2021 style melt up that compounds further. The two outcomes are diverging fast.
8. IGV up 200bps and SMH down 100bps today is the equity long short positioning unwind in real time. Hedge funds are positioned long SMH and short IGV. Today’s geopolitical shock from crude moving up should have hit the entire tech complex evenly. Instead it hit hardware and lifted software. That divergence is mechanical, not fundamental. Microsoft holding flat as the institutional anchor for OpenAI and Azure tells you the unwind has more room to run as IGV reprices higher and SMH reprices lower.
Slide Decks From Livestream:
Here is the report I referenced on the timing for WHEN AI begins to unravel the labor market:
You can follow all of Alfie Kerswell work here:
Tomorrow’s Livestream: Cross Asset Positioning and Equity Long Short Positioning Inside the AI Compression Trade
Tomorrow, I will map cross-asset positioning and equity long-short positioning as they directly connect to the AI compression of the tails framework I laid out today.
TOMORROW’S LIVESTREAM: LINK
Paid Subscriber Section: Proprietary Macro Research
Keep reading with a 7-day free trial
Subscribe to Capital Flows to keep reading this post and get 7 days of free access to the full post archives.