SpaceX IPO, Anthropic Fable 5 & AI CapEx
Brad Gerstner hosts Gavin Baker (Atreides), Andrew Fox, and Clark Tang to break down the upcoming SpaceX IPO at a ~$1.77T valuation. The core thesis is that SpaceX has rapidly morphed from a launch/Starlink story into a major AI compute player, leapfrogging into the #4 hyperscaler slot in roughly 30 days after striking multibillion-dollar deals with Anthropic and Google. The bullish case rests on Starlink's <1% household penetration, Elon's proven ability to stand up data centers at unmatched speed (Colossus in 122 days), and the optionality of orbital compute, which could cut the non-GPU half of data center CapEx from ~$25B/gigawatt to ~$5B/gigawatt once Starship achieves rapid two-stage reusability. On the model side, the group discusses Anthropic's Fable 5 launch and the broader shift toward long-running, agentic tasks that consume vastly more tokens. They argue Noam Brown's point that we may never fully know how smart a given model is because we can't run it long enough before the next generation arrives, and frame this as a major tailwind for compute demand. The consensus view is that frontier labs (Anthropic, OpenAI, xAI/SpaceX, Google) are capturing the overwhelming majority of AI revenue, even as open-source models handle the majority of tokens, and that this is bullish for compute providers even if it pressures frontier margins. The conversation turns to AI CapEx, with 2027 forecasts likely closer to $1.5T (versus Morgan Stanley's $1.1T) as Anthropic's revenue inflected decisively this year. They argue the CapEx-to-revenue math works if inference revenue scales toward $300B+ and gross margins hold at 50-70%. Finally, the panel does a market check, noting extreme dispersion (semis ripping, internet down 16%, software down 8%) and expects a period of consolidation after a steep run, while remaining structurally bullish on AI-exposed names.
Original sources
Podcast BG2Pod with Brad Gerstner and Bill Gurley
YouTube BG2Pod
Video:The SpaceX IPO, Fable 5, AI Capex Update & Market Check w/ Gavin Baker, Andrew Fox & Clark Tang ↗
AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.