At the SALT London conference — hosted by Anthony Scaramucci's SkyBridge Capital at a Mayfair hotel charging up to $3,000 a night — the investor class spent Wednesday arguing about when, not whether, the AI trade unravels. The three most revealing moments came from speakers who each represent a different faction of capital trying to position itself for whatever comes next. Jack Selby, a managing director at Thiel Capital and managing partner at Copper Sky Capital, delivered the bluntest assessment: AI is a bubble approaching its pop. His evidence is structural. The capital intensity is "absurdly high," the chip-and-datacenter cost stack is brutal, and AI executives have resorted to "outrageous and fantastical stories" to keep the fundraising machine running. Selby singled out Anthropic, whose revised S-1 filing this week showed $2.4 billion in interest income over nine months against roughly $1.3 billion in operating losses, sitting atop approximately $96 billion in cash and equivalents. "They have more money than they know what to do with, so why are they rushing to go public?" he asked. His bet: the "euphoric" momentum market ends before year-end 2026. Selby also flagged defense tech as the second-largest bubble behind AI, predicting that the Democratic wave in this week's midterm elections would create headwinds for firms like Anduril — though he framed it as a healthy "glass of cold water" for an overheated sector. Nick Finegold of Curation Capital offered the counter-narrative, channeling Elon Musk's abundance thesis. He predicted that a potential SpaceX-Tesla merger would trigger "the biggest job creation program in the history of the world," and that the narrative must shift from AI stealing jobs to AI creating them. His firm is already monetizing this thesis through Curation Connect, a platform that uses AI video tools to convert regulatory disclosures into explainer videos at a fraction of traditional media-agency costs. He estimated that content-creation costs are collapsing so fast that a $50,000 budget will soon match $10 million in production value. Hightower Advisors' chief investment strategist Stephanie Link offered the most pragmatic framing. She argued cybersecurity "will be bigger than AI" due to the security challenges AI itself creates, and that robotics hasn't even entered the game yet. Citing Amazon CEO Andy Jassy's shareholder letter noting over one million robots operating in fulfillment centers — still at an "early stage" — she estimated active robots need to grow roughly 30-fold to over 150 million units by 2035 to meet demand. The through-line across all three speakers is revealing: nobody at SALT is debating whether AI transforms the economy. The fight is over who captures the value and when the current financing structure breaks. Selby sees the debt markets as the last prop holding up unsustainable capital deployment. Finegold sees a narrative pivot that converts Musk's techno-optimism into investable job-creation stories. Link sees the real money in the picks-and-shovels layer — datacenters, grid upgrades, cybersecurity, robotics — rather than the AI models themselves. What's absent from the SALT stage is as telling as what's on it. No speaker addressed wage effects on displaced workers, the concentration of AI ownership among a handful of hyperscalers, or the public-infrastructure costs of the datacenter and grid buildout that Link described. The job-creation thesis remained assertion, not analysis. The investor class is talking to itself about rotation timing, not about who bears the adjustment costs.
