Token

The basic unit of text (roughly 4 characters or 0.75 words) processed by large language models.

In natural language processing and generative AI, a token is the fundamental unit of data processed by a model. It does not correspond exactly to a single word; rather, it represents a common sequence of characters or sub-words. In English, a helpful rule of thumb is that 100 tokens are equivalent to about 75 words. For investors tracking the artificial intelligence economy, tokens are the primary currency of monetization and cost. Foundation model providers charge developers and enterprises based on the volume of input tokens (the prompt) and output tokens (the generation) processed. Consequently, token volume serves as a key proxy for actual software usage and demand. Analyzing token pricing trends and the cost to generate a million tokens helps investors assess the unit economics of software companies, the pricing power of model developers, and the overall demand growth in the cloud infrastructure layer. A rapid decline in token pricing can signal intense competition among model providers, while rising token volume indicates expanding enterprise adoption.

An enterprise software company processes a document containing 75,000 words. Using the standard industry conversion, this equates to approximately 100,000 tokens (75,000 words / 0.75 words per token). If the API provider charges $0.015 per 1,000 tokens, the total cost to process the document is calculated as: (100,000 tokens / 1,000) $0.015 = $1.50.