The Nature of the Token Divide
Currently, analysts warn that the artificial intelligence revolution threatens to create a new inequality gap driven by tokens. Specifically, unlike the digital divide that centered on the access and affordability of internet connections, this new gap is based on the continuous cost of processing intelligence. Furthermore, tokens are the fundamental unit through which artificial intelligence processing is organized, measured, and sold. Consequently, every request and response generated by AI models consumes compute capacity that must ultimately be paid for by the user or the provider. Ultimately, this cost meter never stops running, effectively dividing users based on the quantity and quality of intelligence they can afford to acquire.
Cost Challenges and the Shifting Frontier
Meanwhile, the business model for artificial intelligence differs fundamentally from traditional broadband networks. For instance, although per-token prices for advanced models have recently fallen due to international competition, the total cost for users often continues to rise. In addition, the economic phenomenon known as the Jevons paradox explains that lower unit costs naturally invite increased usage, more complex interactions, and a surge in autonomous activity. Indeed, public institutions like schools and rural hospitals face the severe challenge of absorbing recurring costs of unknown size every time they utilize these technologies. Therefore, the constant advancement of the technological frontier means that affording yesterday’s obsolete model no longer represents truly competitive access.
Policy Solutions and Mitigation
Moreover, experts note that the historical solutions implemented for electrification or broadband internet will not be sufficient to resolve this new paradigm. In fact, addressing the token divide requires standardized measurement beyond simple token counting, such as establishing an index for capability per dollar. Similarly, policymakers must explore diverse mitigation strategies, including institutional support through buying consortiums, public compute resources, and safety certifications for open-weight models. However, the artificial intelligence market is increasingly concentrated in the hands of a few corporate titans that heavily control the underlying compute infrastructure. Ultimately, ensuring this technology does not widen social and economic inequalities will depend on robust policies that protect competition and promote fair, nondiscriminatory access.
Reference
Wheeler, T. (2026, September 14). The AI revolution threatens a ‘token divide’. Brookings. https://www.brookings.edu/articles/the-ai-revolution-threatens-a-token-divide/