Early SpaceX Investor On AI Compute & Why Everyone Is MASSIVELY Underestimating SpaceX

Despite dire predictions of a lockup "bloodbath," SpaceX stock rose nearly 30% after insiders could sell, underscoring the host's broader argument that public markets misunderstand the company. The discussion highlights that SpaceX's fundamentals have improved since its IPO, with AI compute—not Grok or xAI—representing the massively underestimated opportunity. Because more compute instantly translates into more AI revenue, demand is near-infinite for decades, and SpaceX's ability to build cheap, fast compute could unlock hundreds of billions in annual revenue, dwarfing consensus estimates. External validation from top VC firm Benchmark backing a space-based data center startup supports the viability of this vision. Ultimately, the host concludes the market is underpricing SpaceX's AI compute-driven revenue trajectory, even as he admits his own model may still be too conservative on SpaceX's AI model success.

• ▶ 0:00 The host sarcastically sets up the "lockup expiry bloodbath," but SpaceX stock is actually up nearly 30% in the three days since insiders could sell, contradicting the narrative.
• ▶ 0:36 Context: the podcast was recorded about a week before the lockup expiry, so the discussion predates the "bloodbath" that never happened.
• ▶ 0:51 Gavin is skeptical that public markets truly understand SpaceX, arguing the fundamentals have improved since the IPO, citing Grok 4.5 and the Cursor acquisition.
• ▶ 1:10 Grok 4.5 is "Pareto dominant" — the most intelligence for the least cost — and is a "phenomenal grunt worker" rather than frontier-level AI; the big AI opportunity lies in cheap, near-Einstein-level tokens, and token usage tripled after the build went live post-IPO.
• ▶ 2:30 The AI model layer is the "biggest unknown unknown"; the 20-year valuation model deliberately does not require meaningful Grok success due to extreme competition, though the speaker remains quietly confident long-term.
• ▶ 3:39 A meta valuation exercise shows xAI/Grok represents only 2.2% of 2046 revenue and 2% of profits in the base case; the living model is available to supporters on Patreon at patreon.com/solvingthemoneyproblem.
• ▶ 4:37 AI compute has clearly accelerated meaningfully; SpaceX has shown it can bring more compute online faster and cheaper than anyone else over the last three years.
• ▶ 5:16 AI compute is the thing investors are massively underestimating—demand is astronomical, and cash-rich companies are bottlenecked by compute access because more compute directly means more revenue.
• ▶ 7:53 For any AI company, more compute translates almost immediately into large revenue gains via training new models and running inference, so demand for AI compute will stay near infinite for decades—making SpaceX’s cheap, fast compute buildout a huge long-term revenue engine.
• ▶ 9:35 SpaceX reportedly targeting 8 GW of compute at ~$50B per GW, implying ~$400B annual revenue potential that makes the $73B consensus look far too small.
• ▶ 11:00 Market skepticism is real — few players have surpassed 500 MW in a year, and a NY hedge fund short case expects compute prices to crash, though SpaceX's execution record ("making the impossible late") gives believers hope.
• ▶ 12:37 Investors aren't pricing in the AI compute opportunity: SpaceX remains under $2T, yet even a couple of GW online could add $100–200B+ ARR — but it all depends on execution.
• ▶ 14:24 External validation: Benchmark, a top VC firm outside the Elon ecosystem, funded orbital compute company StarCloud — which relies on SpaceX's Starlink laser tech — serving as a major "sanity check" that space-based data centers are viable.
• ▶ 16:02 The AI compute leasing business (on and off Earth) drives 79% of both revenues and profits in the 20-year Tesla and SpaceX model's base case for 2046, though the host notes this is just the model's estimate, not a prediction.
• ▶ 16:47 Bottom line: The market is underappreciating SpaceX's revenue growth trajectory and the magnitude of the AI compute business, while the host admits he's likely underestimating SpaceX's success with AI models themselves.

• ▶ 0:00 Intro and Market Context (0:00 - 1:10) - Day-three stock check and Gavin on how public markets are digesting SpaceX.
• ▶ 1:10 AI Models, Valuation, and Patreon (1:10 - 4:36) - Grock 4.5/Cursor, model uncertainty, a Grock valuation exercise, Patreon updates, and AI's impact on valuation.
• ▶ 4:36 AI Compute Demand and Revenue Thesis (4:36 - 9:35) - The underestimated AI compute business directly drives revenue through near-infinite demand for intelligence.
• ▶ 9:35 SpaceX's 8 GW Ambition and Market Skepticism (9:35 - 14:03) - Gavin's 8 GW compute thesis versus $73B consensus, market skepticism, and SpaceX's execution record.
• ▶ 14:03 Orbital Compute and Closing Takeaways (14:03 - 17:59) - Orbital compute credibility, model caveats, valuation bottom line, Patreon support, and AG1 sponsor message.