During the AI Impact Summit in New Delhi on February 19, 2026, Indian Prime Minister Narendra Modi posed for a group photo with prominent AI company leaders, including Sam Altman from OpenAI and Dario Amodei from Anthropic. As these companies prepare to make their IPO prospectuses public after confidential filings, investors will encounter a term that may be new to many on Wall Street: tokens. These tokens represent the payment mechanism in the artificial intelligence sector, facilitating transactions between developers and the companies that create AI models. However, converting these tokens into a monetary value that investors understand proves to be complex. D.A. Davidson tech analyst Gil Luria highlighted that this evolving landscape is challenging for everyone involved.
OpenAI recently announced it had confidentially filed its prospectus with the SEC, shortly after Anthropic did the same. With SpaceX anticipated to launch its IPO soon, these offerings could set records, underscoring the need for investors to understand how tokens function. The current shift toward AI mirrors the shift to subscription-based models seen with cloud computing, fundamentally altering the methods of payment and value exchange in technology.
Each time users engage with AI services like ChatGPT or Claude, they utilize tokens to perform tasks such as building spreadsheets or generating images. A token is roughly equated to three-quarters of a word. Companies like OpenAI and Anthropic structure their business models around token-based subscriptions and API access, charging users based on the number of tokens they consume. For example, OpenAI offers its Codex application for free while providing subscription-based access to more sophisticated tools. Additionally, users who exhaust their token limits have the option to purchase more.
OpenAI's pricing indicates that for its premier model, GPT-5.5, users are charged $5 for every million tokens used in inquiries and $30 for each million tokens processed in responses. Anthropic's Claude Opus 4.8 has similar pricing models. For those unfamiliar with this system, it can seem intricate, but large entities like Cerebras and SpaceX have made efforts to clarify the significance of tokens in their IPO filings. SpaceX's documentation defines tokens as essential units of processed text and images used in AI, emphasizing their importance in measuring the effectiveness of AI models.
While tokens are less critical to SpaceX's financial success—where a significant portion of revenue stems from its Starlink internet service—the company recognizes AI as a growing sector, albeit a niche compared to major players like OpenAI and Anthropic. In contrast, Google has seen token usage surge significantly as its models process billions of tokens every minute, with recent reports indicating impressive growth in its cloud revenue fueled by AI services.
Cerebras recently went public and contributes to the token economy through its hardware, enabling faster token generation and reckoning throughout AI processes. They have secured substantial contracts with OpenAI, illustrating the market's demand for robust processing capabilities.
SpaceX noted in its prospectus that its operational efficiency heavily relies on its computing power investments. By leveraging its own infrastructure and data centers, the company aims to maintain a cost edge in token production. Despite being early in its AI journey, SpaceX is already leasing out significant amounts of its data center capacity, signifying a strategic move to monetize its hardware investments.
As companies like Anthropic continue to pay substantial sums for capacity at SpaceX's infrastructure, and Google commits to long-term GPU leases, the dynamics of pricing and competition among AI service providers remain fluid. As the AI landscape evolves, understanding token economics will increasingly be essential for investors eyeing this transformative technological frontier.
