How Does the AI Cost Index SDLLMTK Affect the IPO Outlook of OpenAI and Anthropic?

Put simply, SDLLMTK is the world’s first market benchmark index that tracks the actual, consumption-weighted price per million tokens for LLMs (Large Language Models). It is published by Silicon Data and listed on Bloomberg.

How it works: It combines actual usage and transaction data from the major AI Providers (OpenAI, Anthropic, open-source models and others) to calculate the effective blended token price the market really pays (Blended Price / 1M Tokens).

Trend: A surge in the first half of the year: The index nearly doubled in just five months. A pullback after peaking in May: The index has fallen about 20% from its late-May high, and recent readings are consolidating around $1.62 to $1.67 USD.

Falling cost per token: SDLLMTK shows that LLM inference costs are trending downward. The main reasons include:
Hardware and compute optimisation: Better cost-efficiency from GPU compute such as H100/B200.
The rise of small models (SLMs) and open-source models: Open Weights models force closed-source giants to cut prices to win customers.
Model architecture advances: Quantization and KV Cache optimisation have lowered inference costs.
Shifting market consumption: Enterprises are starting to move non-core work to cheaper Models, creating “hybrid deployment (Model Routing)”.

In the past, people judged the AI giants only by User growth. Now, when Wall Street looks at them, it will look at SDLLMTK’s token price and unit gross margin. Do you think OpenAI or Anthropic can win the IPO valuation battle first? Feel free to discuss in the comments below! 👇