The Impact of S&P Regulations on IPOs
This podcast episode covers the S&P 500's refusal to bend its rules for SpaceX's IPO inclusion, which will also affect future OpenAI and Anthropic IPOs. Additional topics include Google's $920M/month compute deal with SpaceX's XAI, enterprises massively overshooting AI token budgets, Airbnb CEO Brian Chesky launching an AI lab, and the unusual pattern of ~90 investors holding stakes in both OpenAI and Anthropic.
Summary
The episode opens with a discussion of the S&P 500's decision to reject SpaceX's request for expedited index inclusion. The S&P cited three standing requirements: a 12-month post-IPO seasoning period, a 10% float requirement, and a profitability screen. SpaceX would have gained approximately $14 billion in automatic passive fund buying had it been included, while OpenAI and Anthropic stand to miss out on $8 billion and $4.6 billion respectively when they eventually IPO. The NASDAQ 100 and FTSE Russell's Russell Top 500 are moving much faster to include SpaceX, in 15 and 5 trading days respectively, though neither carries the same financial weight as the S&P 500.
The host then covers Google's agreement to pay SpaceX $920 million per month to rent XAI compute capacity, specifically accessing 110,000 NVIDIA GPUs through June 2029 in a 32-month deal. This is SpaceX's second major compute rental agreement after Anthropic signed a $1.25 billion per month deal for the entire Colossus One Memphis facility. The host notes that SpaceX's AI unit is losing $2.5 billion per quarter but now has $2 billion in monthly contracted compute revenue, suggesting the overbuilding of infrastructure may have been strategically advantageous.
Enterprise AI token consumption is explored next, with companies blowing past their 2026 budgets by 3x. Examples include Uber exhausting its entire annual coding budget in four months, and one unnamed company facing a $500 million Claude bill after failing to set developer usage limits. Priceline's Cursor contract renewal cost 4-5x more than expected, and per-developer token consumption rose 18.6x in nine months across tracked companies. Research from Jellyfish found top token users were 2x more productive but burned 10x more tokens. The host advocates for subsidized max subscription plans over direct API usage as a cost-control measure.
Airbnb CEO Brian Chesky is launching a new AI lab focused on user interaction and design, while maintaining his CEO role at Airbnb. The lab will be run by another executive, with Chesky serving as founding chair. He has previously rejected ChatGPT plugin partnerships, citing insufficient robustness of underlying tools, and believes existing AI lacks the rich interfaces needed for travel and e-commerce use cases.
Finally, the episode examines how approximately 90 investors hold stakes in both OpenAI and Anthropic simultaneously — representing 42% of OpenAI's backers. Notable dual investors include Sequoia Capital, Greylock, Founders Fund, and Redpoint Ventures. The host attributes this unusual cross-competitor investing to the sheer scale of fundraising (each company has raised over $100 billion near trillion-dollar valuations), the involvement of hedge funds and private equity accustomed to spreading bets, and traditional VC firms adopting similar hedging strategies ahead of a major IPO season.
Key Insights
- The S&P 500's refusal to waive rules for SpaceX sets a precedent that will also block OpenAI and Anthropic from fast-tracked inclusion post-IPO, costing those companies $8B and $4.6B respectively in automatic passive fund inflows, according to Bloomberg Intelligence.
- SpaceX's apparent overbuilding of AI compute infrastructure has turned into a revenue advantage, with $2 billion in monthly contracted compute rental revenue from Anthropic and Google offsetting its AI unit's $2.5 billion quarterly losses.
- Jellyfish Research found that top AI token users at enterprises were 2x more productive than light users but consumed 10x more tokens, revealing a steep cost-productivity tradeoff that is causing companies to blow past annual AI budgets by 3x.
- Brian Chesky argues that existing AI tools lack the rich interfaces needed for sectors like travel and e-commerce, and has rejected ChatGPT plugin partnerships on those grounds — positioning his new lab as a direct competitor to tools he previously advised on.
- The host argues that the traditional Silicon Valley norm against investing in direct competitors is breaking down in AI, with 42% of OpenAI's backers also holding Anthropic stakes, driven by the scale of fundraising and a rush to capture pre-IPO equity across the entire sector.
Topics
Full transcript available for MurmurCast members
Sign Up to Access