S&P 500 Snubs SpaceX, Leaving AI IPOs in the Lurch

S&P 500 blocks SpaceX entry

SpaceX’s request to fast‑track into the S&P 500 was denied on June 4.^12345678 The index’s manager, S&P Dow Jones Indices, kept the standard eligibility criteria intact.^1234567 That decision bars SpaceX, OpenAI and Anthropic from the automatic $14 billion, $8 billion and $4.6 billion passive‑fund inflows that Bloomberg Intelligence estimated would follow a fast‑track inclusion.⁶

The rule‑change proposal would have cut the seasoning period for new IPOs from twelve months to six, waived the investable‑weight‑factor (IWF) requirement, and dropped the profitability screen for MegaCap firms.³⁴⁵ SpaceX’s plan to offer only about 3 percent of its shares to the public and its $29 billion debt load made the request unusual. The index’s final statement: “no changes will be made to the eligibility criteria including financial viability screens, seasoning period, or minimum IWF.”⁷

Why the index matters

Passive funds that track the S&P 500 control roughly $7.5 trillion.⁷ When a company joins the index, those funds automatically buy its shares in proportion to the company’s weight. For SpaceX, the projected inflow was $14 billion.⁶ OpenAI and Anthropic would have seen similar windfalls—$8 billion and $4.6 billion respectively.⁶

Those dollars are not just a vanity metric. They represent a steady, low‑cost source of capital for companies that still need to fund massive AI‑compute projects. The decision keeps retirement‑savings plans and other passive portfolios insulated from SpaceX’s speculative AI and orbital‑data‑center bets.

AI firms feel the funding pinch

Both OpenAI and Anthropic have recently overhauled enterprise pricing. Anthropic shifted its “Claude seats” model to $20 per seat per month in November 2025, a change reported by The Information on April 14 2026. OpenAI aligned its Codex and ChatGPT Enterprise plans with API token pricing in early April 2026. The new rates match raw API costs, eliminating the deep discounts that early adopters once enjoyed.

The pricing moves coincide with rumors that Anthropic is on the cusp of its first profitable quarter. Users on HN note that heavy‑use plans now cost $100 per month for a modest token allowance, while a $2,180 token bill over 30 days translates to $200 in subscription fees—a “fantastic deal” for coding agents, according to a subscriber. The shift suggests both firms have found product‑market fit in developer‑focused agents and are now monetizing at scale.

The hidden cost of inference

Running frontier models at scale burns cash faster than most software services. A back‑of‑the‑envelope calculation posted on HN assumes a 72‑GPU cluster of Nvidia H100s at $2 per hour each, totaling $144 per hour. With 37 billion active parameters and a batch size of 32, the setup can ingest about 46.8 billion input tokens per hour at roughly $0.003 per million tokens.

Output generation is far more expensive. The same cluster produces only 46.7 million output tokens per hour, costing $3.08 per million tokens—a thousand‑fold difference. Those numbers illustrate why AI firms cling to high‑margin enterprise contracts: the raw compute bill for serving output can dwarf the revenue from token‑based pricing.

Industry context and future stakes

The S&P 500’s stance signals that traditional market gatekeepers remain wary of AI‑heavy, unprofitable megacaps. Even if SpaceX, OpenAI or Anthropic eventually meet profitability screens, they will still need to survive a full year of public trading before qualifying for the index. In the meantime, they must rely on private funding rounds, strategic partnerships, and the kind of enterprise pricing that just went live.

AMD’s recent Linux patch for the Steam Deck shows that performance gains can still be extracted from existing hardware. The patch lifts “1 % low” frame rates by roughly 32 percent in EPP mode, a reminder that software optimizations continue to deliver tangible value without new silicon.

What to watch

Track the next S&P Dow Jones Indices consultation on MegaCap eligibility—any softening could reopen the door for SpaceX, OpenAI or Anthropic. Also monitor the quarterly earnings reports of Anthropic and OpenAI; a profitable quarter would validate the new pricing strategy and could reignite calls for index inclusion. Finally, keep an eye on compute‑cost benchmarks from hyperscalers, as any shift in H100 pricing or MoE efficiency will directly affect the economics outlined above.