The $400 Billion Gap | Warren Pies on Token Panic, a Trapped Fed and What Murders Bull Markets
TL;DR
Warren Pies explains why the Fed is trapped between runaway tech capex driving inflation and a deteriorating housing market, while arguing that AI's greatest existential risk is regulatory backlash rather than technological limits.
⚠️ AI's Regulatory Cliff 4 insights
Open-source panic lacks substance
Fears surrounding open-source models are overblown and not supported by deep data analytics, posing no immediate existential threat to major AI labs.
Political 'horseshoe effect' emerging
Both the political left and right are beginning to question data center buildouts, creating a bipartisan regulatory risk that could peak around the 2028 deadline.
Poor messaging from AI leadership
Lab leaders like Dario Amodei and Sam Altman project alarmist or untrustworthy vibes that fuel public skepticism rather than building consensus for the technology.
Model progress drives market momentum
Continuous model improvement remains the lifeblood of the current bull market, with the Mythos announcement in April catalyzing the recent semiconductor rally.
🏦 The Fed's Capex Trap 4 insights
Tech spending crowds out housing
Information processing investment ($1.5 trillion) has overtaken residential fixed investment ($1.1 trillion), diverting skilled trades like electricians and plumbers away from housing.
Tech directly fueling inflation
Information processing equipment is adding roughly 7 basis points monthly to core PCE, a significant contribution from previously negligible categories.
Monetary policy is a blunted instrument
The Fed cannot surgically slow hyperscaler capex with rate hikes without first crushing the more rate-sensitive housing market and residential construction employment.
Dual mandate conflict intensifies
Because inflation stems from a capital spending boom rather than wage pressure, tightening enough to curb tech inflation would likely violate the Fed's full employment mandate.
📉 K-Shaped Labor Reality 4 insights
Residential construction flashing warning
Residential construction payrolls have declined 2% for two consecutive months, a traditional yellow light that historically precedes recessions when hitting 8%.
Headline jobs mask underlying weakness
Excluding healthcare, education, and government, non-farm payrolls are essentially flat year-over-year despite seemingly strong headline reports.
Quality of jobs deteriorating
Job growth concentrates in sub-$40,000 service roles rather than high-income white-collar positions, limiting consumption power and economic velocity.
Wage stagnation at the top
College-educated workers face the weakest wage growth, with nearly 14% receiving zero raises, indicating slack in the traditionally strong segment of the labor market.
📊 Real-Time GPU Demand Signals 4 insights
Availability beats announcements
Real-time GPU availability via NeoClouds provides a cleaner read on actual demand than hyperscaler capex announcements or revenue figures which include non-AI business.
Hourly tracking since 2023
Researchers track GPU availability on an hourly basis across NeoCloud providers, watching for supply constraints on chips like H100 and Blackwell.
Rental prices confirm tightness
Cloud GPU rental prices typically rise approximately two months after availability metrics collapse, confirming genuine demand versus speculative inventory hoarding.
Data converted the skeptics
Originally collected from a posture of AI skepticism, the persistent supply constraints revealed by this data ultimately convinced researchers of the infrastructure boom's legitimacy.
Bottom Line
Monitor residential construction employment and NeoCloud GPU availability as leading indicators, while recognizing the Fed is structurally unable to tame tech-driven inflation without breaking the housing market first.
More from Excess Returns
View all
He Reverse-Engineered Why Amazon Won | Ritavan on Why the Moat Is the Wrong Test
The concept of economic "moats" is a static checklist that fails when business paradigms shift; investors should instead hunt for "system gambits," where companies sacrifice current performance to change the game entirely, building structural advantages that compound irreversibly in new systems.
Easy Bubbles. Hard 100 Baggers. Useless AI | 6 Things We Learned This Week
Ben Inker and Chris Mayer join the hosts to discuss how to navigate different types of market bubbles based on risk-return dynamics, why vague labels like 'AI' obscure true business analysis, and how accounting mechanics for massive capital investments can create temporary earnings peaks that mask underlying valuations.
He Wrote the Book on 100-Baggers | Chris Mayer on SpaceX, AI Reckoning, and Why Early Is Overrated
Investor Chris Mayer warns that SpaceX's 145x revenue valuation and the current AI boom echo past bubbles, arguing that investors should resist the urge to buy early during hype cycles because truly great companies endure severe drawdowns and offer multiple entry points over time.
We Asked GMO’s Head of Asset Allocation Why This Bubble is Easy — But Investors Will Get it Wrong
GMO's Ben Inker argues the current AI-driven US stock bubble is "easy" to navigate because investors can rotate to fairly priced international assets while maintaining normal risk levels, unlike the "hard" bubbles of 2007 and 2021 when all risk assets were simultaneously overvalued.