David George & Jack Altman on AI, Autonomy, and the Next $25 Trillion

Episode Summary

Executive Summary: David George argues AI is still in early diffusion: demand is outsized, supply remains constrained, and the winners will likely span frontier models, open source, apps, consumer, autonomy, and robotics. He sees token usage and productivity expanding across the economy, with huge venture and growth opportunities as AI, autonomy, and robotics reshape markets and investing.

Main Topics: AI is an 'and' market, not an either/or (Priority: 5/5): George repeatedly rejects binary debates in AI—frontier vs open source, models vs apps, infrastructure vs software—arguing that multiple layers of the stack can expand together as the market grows. AI diffusion is still early and concentrated (Priority: 5/5): Current revenue is concentrated in a small set of heavy users, especially coders, while broader knowledge work remains largely untapped. George says AI is under 5% diffused into the B2B economy and still early in enterprise and consumer adoption. Supply constraints and token demand (Priority: 5/5): He argues the supply side is bottlenecked across the data-center and compute supply chain, while demand remains insatiable. If more capacity came online, he believes it would largely be consumed. Frontier models, open source, and applications can all win (Priority: 4/5): George sees strong demand for frontier products among power users, but also massive room for cheaper N-1/open-source models and for application companies that deliver workflow-specific value and strong GTM. Consumer AI is underappreciated (Priority: 4/5): Despite a billion users, consumer AI is still mostly used like a better search engine. He expects the next wave to be proactive, multimodal, assistant-like products with subscriptions, ads, and new ad formats. Autonomy and robotics as next major waves (Priority: 4/5): Beyond AI text/workflows, George is highly bullish on self-driving, delivery/ride-hail, and robotics, seeing large markets, improving safety, and major productivity gains as these technologies diffuse. Capital cycles, growth investing, and narrative (Priority: 4/5): He argues growth investing is increasingly where venture-scale returns happen, especially in AI where more capital can directly improve products. Narrative, founder charisma, and vibes matter more than ever for customers, talent, and valuation.

Key Arguments: AI revenues are still concentrated among a small number of users, especially coders, so the current market underrepresents the eventual size of broad knowledge-work adoption. The answer to most AI questions is 'and': frontier and open source, models and applications, infrastructure and software can all succeed simultaneously. Demand is so strong that if more capacity were available, it would likely get used, though pricing may adjust. The biggest uncertainty is less about market demand than about supply-chain bottlenecks and product delivery/implementation friction. Frontier models retain a strong revealed preference among heavy users because of quality, first-party products, and tightly coupled model-product harnesses. N-1/open-source models will matter more as usage scales and cost becomes a larger concern for enterprises and application companies. Application companies can build durable businesses because last-mile workflow details, integrations, and go-to-market execution matter deeply. Consumer AI will likely evolve from reactive search replacement to proactive, multimodal personal assistants with action-taking capability. Autonomy can 10x markets by making travel safer and cheaper than human-driven ride-hail, while consumer car owners would pay for full autonomous capability. Robotics will eventually be larger than language because it spans both consumer and B2B use cases and can create feedback loops from real-world deployment. Growth-stage investing is more important because returns increasingly compound in later stages, and AI can absorb massive amounts of capital while improving with scale. Narrative and founder-led communication are a real asset: they influence fundraising, hiring, customer adoption, and valuation.

Data Points: AI infrastructure build-out as % of GDP: surpassed railroads - George cites infrastructure spending as an indicator of scale and argues AI build-out is already historically large as a share of GDP. Current infrastructure scale: 4 gigawatts - He references current live AI capacity when discussing whether additional capacity would be absorbed. Potential data-center availability: not until 2028 - He says supply-chain bottlenecks mean data-center capacity is constrained for years. Current AI revenue ecosystem: about $120 billion - Altman cites OpenAI, Anthropic, xAI, and Cursor together as generating roughly this amount of revenue. Revenue concentration: 95%+ of dollars - George says frontier labs capture essentially all market dollars today. Coders as revenue base: 30 million - He says coders are the main paying cohort driving most enterprise AI revenue. Knowledge workers: 1.5 billion - He contrasts the small current paying base with the much larger eventual addressable workforce. Current B2B diffusion: <5% - George estimates AI is still in very early diffusion across the B2B economy. Consumer users: 1+ billion - He notes AI already has massive consumer usage, but the product form remains primitive. Legal adoption lag: ~12 months behind coding - He characterizes legal as an early but lagging enterprise use case versus coding. Autonomy safety improvement: 10x to 14x safer - He cites Waymo performance relative to human drivers. Consumer car ownership cost: ~80 cents per mile - He compares fully loaded owned-car costs to ride-hail pricing. Uber/Lyft cost: ~$2+ per mile - Used to argue autonomous ride-hail has room to undercut current prices. Cars sold annually in the U.S.: 17 million - He uses this to frame the monetization opportunity for autonomous features. Existing U.S. car fleet: 250 million - He points to the installed base for autonomous feature monetization. Waymos in the U.S.: <10,000 - He emphasizes how small current deployment is relative to eventual market size. Market cap created by prior wave: 25 trillion - He compares the mobile/social/cloud era to the potential of AI, autonomy, robotics, and biohealth. Top public tech firms: 8 of top 10 - He notes most of the world’s most valuable companies are previously venture-backed U.S. tech firms. Private-market return split: 50/50 seed-to-B vs C+ - He cites a historical analysis showing half of returns arise from early stage and half from later stage. Market share estimate: ~20% - George says Andreessen Horowitz has roughly 20% share in growth land, by its own framing of the market.

Pivotal Quotes: "The answer in AI is probably and." — David George: Core thesis for how to think about frontier vs open source, labs vs apps, and other AI stack debates. "If you can drive in something that's 10 to 14 times safer, you would pay a massive premium." — David George: Explaining why autonomy can command strong consumer demand and expand the ride-hail market. "The answer to all this is and." — David George: His broader framing that multiple AI layers can succeed together rather than being mutually exclusive.

Implications: Listeners should expect AI adoption, autonomy, and robotics to expand in parallel, not competitively. For founders and investors, the opportunity is in riding massive product cycles, owning narrative, and building the last-mile products, distribution, and trust that labs won’t cover.