E173: Raoul Pal: The AI Race Will Make Crypto Holders Rich

Raoul Pal argues that the AI race is the most consequential capital allocation event in human history, driven by game theory that makes it impossible for any nation to stop. He contends only the US and China can afford to develop AGI, and that neither can be permitted to hold a monopoly, ensuring a two-pole race that cannot be abandoned. ARC research shows AI word output per year has already exceeded total annual human word output within three years of the technology's emergence, and Pal predicts that by 2028 AI will have produced more words than all of humanity has ever generated throughout history. He frames AI as the first real-world example of Reed's Law, which he describes as Metcalf's Law squared, meaning it follows the exponential of an exponential growth curve. Silicon-based AI processes intelligence at roughly one million times the speed of biological neurons, and Anthropic posted 80x revenue growth in its first quarter against a projection of only 10x.

Pal introduces the concept of an economic singularity, the point at which institutions and economic measurement systems can no longer keep pace with the speed of technology, which he estimates is approximately four years away. AI agents will become independent economic actors capable of instant capital formation and instant capital destruction, making the role of large corporations unclear. He argues AI agents represent an entirely new total addressable market for crypto, previously sized around human adoption at roughly 100 trillion dollars. Because AI agents will use crypto rails to build businesses and transact, the addressable market now extends to potentially trillions of agents rather than nine billion humans. He considers owning layer one blockchains the clearest way for retail investors to participate in the economic activity AI will generate, noting retail had no equivalent participation mechanism during the rise of the internet.

Pal identifies ETH, Solana, and Sui as the only three tokens that maintained economic density when the broader market fell 80 percent, and argues large infrastructure networks historically consolidate to three to five dominant players. He states ETH carries the largest amount of economic and intelligence density of any layer one, and that if Ethereum were shut down today, every layer two, all of DeFi, all real world assets, and all NFTs would go to zero, which in his view suggests ETH is undervalued. He draws a Microsoft analogy for Ethereum, saying it wins not on technical efficiency but on developer count, economic density, Lindy effects, and security. Solana is faster and cheaper with solid developer density. Sui operates at a different order of magnitude through programmable block times and speed to finality, though Pal cautions that Sui's TVL is currently equivalent to its stablecoin size and he wants to see at least twice that before treating it as mature. He discloses he sits on the Sui board. He views Hyperliquid as a strong niche product but believes it lacks a durable moat because Robinhood and eventually Coinbase will compete directly in the same space.

Pal characterizes Bitcoin's decline from 126,000 dollars to 60,000 dollars as a nasty correction within a bull market rather than a bear market. He added Zcash and more Sui during the recent correction, values Zcash as essentially Bitcoin with privacy, and estimates it could be worth 5 to 10 percent of Bitcoin's economic value, with an additional argument that it is also quantum resistant. He sees Bitcoin and crypto as approximately two standard deviations oversold relative to the Nasdaq on their long-term uptrend, calling this one of the cheapest entry points in the cycle. Stablecoins are already running at approximately 100 trillion dollars a year in transaction volume, and he expects the Clarity Act to be signed and open blockchain development to nearly everyone. He predicts both Solana and Ethereum could do a 13x from current levels.

Pal is dismissive of active trading as a strategy, noting that in 35 years he has not known anyone who can consistently buy bottoms and sell tops, and that figures like the Winklevoss twins and Dan Morehead made their fortunes by holding and doing nothing. His preferred approach is to buy when an asset is one to two standard deviations oversold on the log regression channel, which occurs roughly twice per four-year cycle, and otherwise do nothing. His long-run estimate is that the crypto market reaches 100 trillion dollars, at which point token holders effectively earn rent as institutions use those rails.

His bear case centers on unresolved Middle East conflict driving inflation, a renewed tightening cycle, and capital starvation slowing AI development. He assigns a 70 percent probability to a positive outcome and 30 percent to a negative one. He acknowledges uncertainty about whether 2022 marked a genuine AI acceleration or a bubble dressed in new-paradigm language, though he rejects the view that AI revenues are a Ponzi scheme, arguing all participants are genuinely profiting from real demand for intelligence, which he considers infinite.