S&P 500 Refuses SpaceX, OpenAI and Anthropic

In today's issue

01 · Business

S&P 500 refuses to bend rules for SpaceX, blocking OpenAI and Anthropic too

S&P Dow Jones Indices rejected SpaceX's request to skip its standard rules for fast entry into the S&P 500 on June 4, blocking a precedent that would have also shut OpenAI and Anthropic out of the index at IPO. The ruling denies SpaceX $14 billion in automatic passive fund buying and protects $7.5 trillion in S&P 500-tracking assets from absorbing unprofitable, thinly-held mega-cap debuts.

The details:

S&P rejected three requested waivers: the 12-month seasoning period, the 10% public float requirement, and the profitability screen that MegaCap IPOs must meet.
Bloomberg Intelligence pegged passive buying foregone at $8 billion for OpenAI and $4.6 billion for Anthropic if they later sought S&P 500 inclusion.
Nasdaq-100 will admit SpaceX in 15 trading days and FTSE Russell's Russell Top 500 in 5 trading days—far faster alternatives that still trigger meaningful but smaller passive inflows.

Why it matters: S&P's refusal to bend is a rare win for index discipline over capital flows. The decision forces SpaceX, OpenAI, and Anthropic into a two-tier outcome: they'll get rich from other indexes and private wealth, but lose the $7.5 trillion halo of S&P 500 inclusion. Expect massive pressure to mount if any of these companies hit trillion-dollar valuations—but for now, index governance held the line against mega-cap exceptionalism.

Read the full story →

02 · Business

Google to pay SpaceX $920M a month to rent xAI compute capacity

Google will pay SpaceX $920 million per month for access to 110,000 Nvidia GPUs through June 2029, a 32-month deal disclosed as SpaceX prepares for a $1.75 trillion IPO next week. The contract lets Google secure massive compute capacity while SpaceX converts billions in AI infrastructure spending into locked-in revenue ahead of its public offering.

The details:

The deal runs October 2026 through June 2029 and covers GPUs plus CPUs and memory components in SpaceX-operated data centers, with capacity ramping through September 2026.
This is SpaceX's second mega compute lease; Anthropic signed a $1.25 billion-per-month agreement last month for the entire Colossus 1 Memphis facility.
Google's parent Alphabet faces $180–190 billion in 2026 capex and announced an $85 billion stock sale this week, signaling it expects even larger AI infrastructure costs in 2027.

Why it matters: SpaceX is monetizing AI infrastructure before it goes public, and Google is subsidizing a competitor's growth to avoid building all its own chips. This deal is a band-aid: Google's capex guidance keeps climbing, and SpaceX's AI unit is losing $2.5 billion a quarter even with $2.07 billion in monthly contracted revenue locked in. Expect more such deals as hyperscalers realize they can't build fast enough alone—and as startups leverage balance-sheet scale to become infrastructure providers.

Read the full story →

03 · Business

Enterprises blow past 2026 AI token budgets by 3x, scramble for controls

Enterprises are spending 3x their full-year 2026 AI token budgets by April, with Uber burning through its entire coding budget in four months and one company facing a $500M Claude bill after failing to set usage limits. The explosion stems from agentic AI models released in November that consume tokens at rates per-token discounts cannot offset, forcing companies to scramble for cost controls.

The details:

Priceline's Cursor contract renewal cost 4-5x more than expected; per-developer token consumption rose 18.6x in nine months across tracked companies.
Top token users were 2x more productive than light users but burned 10x more tokens to achieve those gains, per Jellyfish research.
The Linux Foundation launched the Tokenomics Foundation this week to standardize AI cost metrics, with 180 vendors already participating in FinOps Foundation efforts.

Why it matters: The token bill has arrived, and enterprises are learning the hard way that cheap-per-token pricing doesn't matter when agentic AI multiplies consumption by 18x. The infrastructure for tracking trillions of monthly token rows doesn't exist yet—most cloud cost systems collapse under the load—which means billing disputes and audit nightmares are coming. Expect a new vendor category in cost-tracking software and stricter token guardrails to replace 2025's "go fast and use the best model" ethos.

Read the full story →

AI Box

Every AI model. One chat.

The latest models from ChatGPT, Claude, Gemini, Sora, ElevenLabs — 80+ models in a single chat. Compare answers side by side. Pick the best one every time.

● ChatGPT, Claude, Gemini, Grok, DeepSeek — in one chat

● Generate images & video with Sora, Veo, Ideogram

● Compare any two models side by side

● From $8.99/mo · 80+ models, all included

Try AI Box →

aibox.ai

Trusted by 3,000+ teams

04 · Business

Airbnb CEO Brian Chesky is starting his own AI lab

Airbnb CEO Brian Chesky is founding a new AI lab focused on user interaction and design while keeping his CEO role, directly competing with OpenAI—a company he has advised for years. Chesky believes existing AI tools lack the rich interfaces needed for travel and e-commerce, creating both a philosophical rift with Sam Altman and a $1 billion revenue opportunity for Airbnb.

The details:

Chesky will not run the lab himself; another executive will serve as CEO while Chesky acts as founding chair, mirroring his hands-on "founder mode" approach at Airbnb.
Airbnb has rejected ChatGPT plug-in partnerships because underlying tools aren't robust enough, per Chesky, setting the rationale for his competing lab.
Brett Adcock's Hark, launched late last year, represents similar bet that incumbent model labs miss the UX frontier—but Chesky's design-school background differentiates his angle.

Why it matters: Chesky is betting that OpenAI and Anthropic have optimized for generalist chat when the real value is in domain-specific interfaces tailored to travel, commerce, and complex user flows. His move signals a fracture in founder-mentor relationships under AI pressure—loyalty to Altman evaporates when the product vision diverges. Expect more operator-founders to fork their own model labs as they realize off-the-shelf LLMs are table stakes, not differentiation.

Read the full story →

05 · Business

About 90 investors hold stakes in both OpenAI and Anthropic

About 90 venture firms and money managers have invested in both OpenAI and Anthropic, with 42% of OpenAI's backers also holding Anthropic stakes. The overlap is unprecedented for direct rivals and signals that top investors see no clear winner in the AI race—they're hedging by owning both.

The details:

Sequoia Capital, Greylock, Founders Fund, and Redpoint Ventures all sit on both cap tables; at least 13 of the 31 investors named in Anthropic's latest raise already own OpenAI shares.
Each company has raised over $100 billion at $1 trillion valuations ahead of planned IPOs this year, making individual stakes small enough that classic conflict-of-interest and governance concerns largely disappear.
Roughly 30 of the overlapping investors are hedge funds, private equity, or wealth managers that routinely spread bets across competitors; traditional venture firms are now adopting the same posture.

Why it matters: The venture world's shift from picking winners to owning the entire category is a tacit admission that foundation models are a commodity-in-formation, not a winner-take-most market. Smart money is betting on the IPO lottery—whichever AI lab reaches public markets first and pops will fund the others' eventual losses. Watch for this dual-exposure pattern to spread to xAI and other deep-pocketed rivals.

Read the full story →