For Subscriber Information Visit ai.22vresearch.com Contact Mark Whaling: mwhaling@22vresearch.com AI Is Changing Time: Those Who Can't Compete, Compute This week I focus on the most important story in AI investing: the insatiable demand for compute, and why almost every traditional framework investors use to read this cycle is broken. The S&P finished up 40bps, IWM and the Qs up 1%, the AI thematic portfolio up 3%, with new all-time highs across most major global indices. Earnings growth is 51%, sales up 15% y/y, global EPS up roughly 47%. That is not an isolated U.S. tech bubble. That is a worldwide breakout. The reason: AI is compressing time. Agents do not sleep or stop working. "Year-over-year" is a human measurement applied to a non-human workforce. Earnings compound at 30%+ while hiring is basically zero, aggregate weekly payrolls post their weakest six-month rate of change since 2012, and participation keeps sliding. Every prior bubble came with over-hiring. This one comes with productivity. The next constraint may not be physical. It may be financial. Consumer agents need payment rails, settlement, stablecoins and tokenization, which is why crypto becomes a much bigger part of this conversation over the next 12-24 months. Timestamps (00:00) Personal notes, Maine, and independent thinking Why Maine matters, and why I will never reach out on WhatsApp, X or DMs offering trades. My research philosophy: I care about everyone's opinions, not anyone's opinions. (03:50) The core thesis: compute demand is insatiable Compute demand is structural, like food demand. Billions of agents are coming online and they all need tokens. 2027 becomes the year of consumer agents. (06:14) Weekly recap: global earnings are breaking out Earnings growth 51%, sales up 15% y/y, global EPS up roughly 47%. This is not a few U.S. tech names. It is a global earnings breakout. (07:51) Earnings up, hiring zero The productivity boom shows up in margins instead of payrolls, and enterprise agents are not even fully deployed. What took a year may soon take a quarter. (10:35) The compute step function Chatbots needed one level of compute. Coding agents need 10-100x more. Enterprise and consumer agents, autonomous driving, world models and humanoids need another step up. (12:31) Capital is flooding into AI infrastructure Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR are building $500B+ financing platforms. Morgan Stanley has a $1.5T initiative. Memory remains the real bottleneck. (14:38) CoreWeave and scarcity value A100s from 2020 are contracted through 2029 and sold out. Useful life expectations are moving from 2-3 years to 6-9 years. Physical compute now has scarcity value. (16:00) Google's AI regime change Google may be shifting from frontier model prestige toward infrastructure economics. Capex at a documented 30% ROIC is very different from high-variance model development. Google may become the OPEC of compute. (19:54) SpaceX and orbital compute SpaceX may add 6-8 gigawatts of incremental data centers in 2027, possibly above 10, much of it earmarked for space-based capacity. (23:32) Why the Street can't cover this AI is not a sector. It cuts across industrials, materials, chemicals, energy, power and semis. YTD: Nasdaq +20%, the 100-name AI portfolio +46%, the 25-name +74%, the 10-name +94%, average forward PEG near 0.8. (30:10) The conviction prompt and tactical risk How to use weekly transcripts to measure structural conviction, tactical outlook and risk appetite over time. Structurally bullish compute, tactically cautious into the parabolic exhaustion, constructive near capitulation. (34:54) VisserLabs and the AI-native workflow Visserlabs.com was built in about an hour with prompts. The point is not the website. It is that AI-native workflows let small teams do what used to require far larger organizations. (36:13) Crypto rails for AI agents Figure's consumer loan marketplace volume was $4.3B, up 132% y/y, with its tokenized on-chain marketplace approaching 70% of volume. Stablecoin card spending rose another 16%. (39:05) The yen, Warsh, and the academic Fed Bonds will not rally despite softer inflation, weak retail sales and a jobs miss. The academic Fed is still anchored to old frameworks while AI changes labor, productivity and time. (43:32) The labor and inflation data academics ignore Aggregate weekly payrolls at the weakest six-month rate of change since 2012. Job creation is concentrated in healthcare and leisure. PCE core is the inflation outlier. (48:20) The debasement trade and the compute dollar Dollar at new lows, gold and silver have moved, Bitcoin has not yet. China added 20 tons of gold in July, its largest since October 2023. Korea resumed buying after 13 years.