In the rapidly evolving world of AI, the true cost of progress is being reshaped. While the compute power driving AI has been a known expense, the valuation of the tokens produced by this compute has remained elusive—until now.
Earlier this year, the ability to trade GPU compute like a commodity was pioneered when Ornn introduced the Ornn Compute Price Index (OCPI) on the Bloomberg Terminal. This move marked the advent of compute futures on the Intercontinental Exchange, transforming how we perceive the value of computational resources. However, the AI ecosystem also relies heavily on inference tokens, the real currency in AI transactions, which have lacked a standardized price.
To understand why a fixed price is crucial, consider the history of commodities like oil. In 1866, oil producers standardized a barrel to 42 gallons, creating a uniform unit that facilitated transparent pricing. Similarly, in AI, the inference token is becoming the standard unit. Yet, without a clear price, the market remains opaque.
Traditionally, AI model rate cards have been public, but actual transaction prices often deviate significantly. This discrepancy echoes the historical oil industry's posted versus transacted price gap, which in AI arises from structural differences such as caching, provider routing, and model selection.
Today marks a pivotal change. Ornn, supported by 021T Capital, is unveiling the Ornn Token Price Indices (OTPI), the first framework to benchmark frontier-lab tokens based on real transactions instead of posted rates. OTPI provides daily indices for leading AI labs, Anthropic and OpenAI, offering a transparent metric for enterprises budgeting AI expenditures, investors gauging demand, and labs assessing monetization strategies.
Each lab's OTPI index calculates the dollar value per million tokens by weighing models according to transacted token volumes. Unlike historical price indices that averaged prices, OTPI reflects the real-time market dynamics by considering actual paid inference, encompassing billions of transactions daily.
While OCPI captures the cost of GPU time, OTPI reveals the cost of the tokens produced, jointly mapping the AI cost curve. The key question is not whether AI is used, but whether its usage is sustainable financially. OpenAI suggests the cost per intelligence unit is plummeting 40-fold annually. OTPI measures how the token price holds up as AI progresses, shedding light on the real deflation in AI economics.
This index provides unique insights into a provider's model traffic distribution and the adoption rate of new releases, data typically guarded by labs. As global data-center investments surge towards nearly $7 trillion by 2030, driven by the belief in growing AI demand, a reliable token price offers the first tangible evidence of this trend's reality.
In a world increasingly defined by AI, understanding the true cost of intelligence becomes vital not just for businesses and investors, but for society at large, influencing policy, ethical considerations, and the future landscape of technology.