Don't Fear the Demise: Raoul Pal's Playbook for the Exponential Age

Raoul Pal organizes his entire investment thesis around one principle: markets and the universe solve for maximum intelligence output per unit of energy. He argues AI is the greatest discovery in human history because it creates intelligence by running electricity through silicon rather than carbon, and silicon substrates process information roughly a million times faster than biological ones. Because AI is built on top of the internet, which already scales by Metcalfe's law, the combined effect produces what he calls Reed's law, or Metcalfe's law squared, a double exponential with no precedent in prior technology or biology.

Adoption metrics support his claim that the buildout is real and accelerating. AI reached 100 million users within a week of launch, and OpenAI and Gemini have each reached one billion users. Anthropic's revenue is projected at roughly 100 billion dollars this year from effectively zero eighteen months ago, which Pal characterizes as the fastest trajectory of any company in history by an order of magnitude. Crypto scaled faster than the internet, he argues, because its investable network value created a built-in incentive mechanism the early internet lacked.

On valuation concerns, Pal acknowledges the CAPE ratio sits near 40 times earnings, approaching dot-com levels, and that Buffett has cut his Apple position while holding historic cash. His counter is structural: currency debasement since 2008 has mechanically pushed equity prices higher in nominal terms without a corresponding deterioration in real value, inflating the CAPE ratio in ways that require adjustment for the pace of money printing. He adds that current markets have not yet reached the collective insanity he witnessed in 1999 and 2000, and that the cycle has further to run.

Pal draws a sharp distinction between today's AI infrastructure buildout and the 2001 fiber optic collapse. Fiber optic companies had zero revenue, enormous leverage, and negative free cash flow. Hyperscalers today carry debt of roughly 4 percent of market cap, generate large cash flows relative to interest obligations, and went cash flow negative deliberately to build compute rather than to fund unproductive spending. GPU prices and rental rates are rising, compute leasing margins run at 30 to 35 percent or higher, and hyperscaler margins near 55 percent are holding simultaneously across all players with no competitor losing share to demand growth. The arrival of cheaper Chinese models has not observably reduced margins at Anthropic or OpenAI, which he takes as evidence of a growing rather than fixed pie. He acknowledges that special purpose vehicles such as CoreWeave carry real distress risk, but argues that failed compute assets would be purchased almost instantly, accelerating intelligence growth rather than interrupting it. The one genuine throttle he identifies is electricity permitting, which has left the buildout roughly 30 percent below where it should be this year.

Pal applies Jevons paradox to compute: greater abundance of intelligence generates greater demand rather than saturation, because AI agents themselves consume compute and expand the total addressable market beyond what human demand alone could support. Amazon is his clearest current example, with its robot-to-human ratio approaching one-to-one, hiring flat while robot deployment has grown vertically, and the company simultaneously operating self-driving vehicles, delivery drones, and in-house robotics.

On the macroeconomic implications, Pal defines GDP growth as population growth plus productivity growth plus debt growth, then notes that labor force participation is declining in the US and population growth is negative across most of the Western world. He argues AI and robotics fill that gap and that GDP growth of 10 to 20 percent is plausible beyond 2030, though he is explicit this is not yet visible in current data. At that level of productivity growth, he contends it becomes very difficult to generate inflation or debase currency, and debt as a percentage of GDP falls in a dynamic similar to the financial repression of the 1950s, when US debt fell from over 100 percent of GDP after World War II to roughly 15 percent by the end of the 1960s by keeping interest rates below GDP growth rather than inflating debt away.

Pal contends that Trump, Besant, and Kevin Warsh have all explicitly cited the Greenspan late-1990s period as their policy playbook, with Warsh's role being to manage liquidity, limit rate increases, print money to service debt, and deregulate technology companies. He notes core CPI is currently not rising because AI is already deflationary and productivity is picking up, and predicts GDP growth may peak around 4 percent in the current cycle before the larger AI productivity dividend arrives after 2030. On portfolio construction, he argues leverage is the primary mechanism by which investors lose crypto holdings, and that unleveraged participation in exponential technology trends is the correct approach, sizing positions so that volatility is tolerable and adding on drawdowns rather than attempting precise market timing.