Key Takeaways:
Wall Street is about to get a crash course in the token economy, the usage-based metric that will underpin the $180 billion wave of AI IPOs from OpenAI, Anthropic and SpaceX.
A token — roughly three-quarters of one word — is the atomic unit of the AI economy, and investors will need to understand it as OpenAI, Anthropic and SpaceX prepare to go public. OpenAI filed confidential IPO paperwork on Monday, a week after Anthropic did the same. SpaceX is slated to debut as early as Friday.
"It's a work in progress for all of us navigating this new terrain," Gil Luria, tech analyst at D.A. Davidson, said.
OpenAI charges $5 per 1 million tokens of input and $30 per 1 million tokens of output for its most powerful model, GPT-5.5. Anthropic's Claude Opus 4.8 charges $25 per 1 million output tokens. SpaceX's IPO filing mentions tokens 62 times, defining them as "the basic units of text or images processed and generated by an AI model, used to measure AI."
The three companies could raise at least $180 billion combined, according to PitchBook, making an understanding of tokens crucial for a market that has traditionally valued companies on revenue and profit — not computing usage. The SpaceX IPO is already two times oversubscribed, with $150 billion in capital competing for $75 billion of stock.
The Token Pricing Puzzle
Unlike cloud computing, which shifted enterprise software from licenses to subscriptions, AI billing is granular. Every ChatGPT query, every image generation, every coding session consumes a specific number of tokens. Google's models now process more than 16 billion tokens per minute via direct API use, up from 10 billion last quarter, CEO Sundar Pichai said on the company's April earnings call. Over the past year, 330 cloud clients processed more than 1 trillion tokens, while 35 reached the 10 trillion token milestone.
The pricing math is complex. OpenAI's GPT-5.5 costs six times more for output tokens than input tokens, reflecting the computational intensity of generating responses. Cerebras, the chipmaker that went public in May, mentioned tokens 23 times in its prospectus, describing them as "how AI converts compute into intelligence."
"Token volume is a useful directional metric, but businesses ultimately care about impact and ROI," Scott Breitenother, co-founder and CEO of AI startup Kilo Code, said.
The Cost Reckoning Arrives
The shift from flat-rate to usage-based pricing is forcing companies to rethink their AI budgets. Coinbase has instituted weekly price caps ranging from $500 to $5,000 per employee based on job level and role, according to Rob Witoff, the crypto exchange's infrastructure executive. A March and April survey of 200 executives by Wakefield Research found that 79% were concerned their AI budgets would be cut because spending wasn't tied to new revenue.
Salesforce CTO Parker Harris said the company is spending "far more" than planned on tokens for its 2026 fiscal year. "We can't tell our investors, 'Yeah, sorry, we gave half of our upside this year to Anthropic so they can go public,'" he said.
The cost pressures are reshaping competitive dynamics. Some companies are offloading basic work to cheaper models from Chinese firms like Deepseek and MiniMax. Ahmad Awais, founder of coding agent startup Command Code, said his company gained 10,000 customers in a recent 30-day stretch, largely driven by demand for cheaper models.
For OpenAI and Anthropic, the timing is critical. Both companies are racing toward IPOs while facing margin pressure from rising infrastructure costs. Anthropic committed to paying SpaceX $1.25 billion a month for three years for data center capacity. Google agreed to pay $920 million a month until mid-2029 for 110,000 Nvidia GPUs. Cerebras signed a deal to provide more than $10 billion worth of compute to OpenAI through 2028.
"If Anthropic lists first and reports its first profitable quarter, OpenAI prices against a profitable competitor at a higher valuation," Harrison Rolfes, senior research analyst at PitchBook, said. "There is no scenario where that comparison favors OpenAI."
The order of the IPOs will influence pricing across the sector. A poor showing for OpenAI "reprices private AI across the board," Rolfes said. Perplexity CEO Aravind Srinivas said the ripple effects would be broad: "Their success means they'll be able to invest more into frontier model development."
This article is for informational purposes only and does not constitute investment advice.
