Gavin Baker, a well-known technology investor and portfolio manager who has closely tracked artificial intelligence infrastructure spending, has publicly highlighted a difference in cost efficiency between two of the industry's leading foundation model developers: Anthropic and OpenAI. According to the report, Baker pointed to Anthropic's lower cost per token relative to OpenAI, a metric that measures how much it costs a company to process the basic units of text that large language models use to generate responses.
Cost per token has become an increasingly important benchmark in the AI sector as companies scale their model deployments across consumer products, enterprise software, and developer tools. Because token processing costs are tied directly to compute usage, chip efficiency, and model architecture, even modest differences in per-token economics can translate into meaningfully different margins as usage volumes grow into the billions or trillions of tokens processed monthly.
Baker's commentary arrives at a moment when investors and analysts are paying closer attention to the underlying unit economics of AI companies rather than just headline revenue or user growth figures. Both Anthropic and OpenAI have raised substantial capital at high valuations, and each has pursued aggressive infrastructure buildouts involving specialized chips, cloud partnerships, and custom data center capacity. How efficiently each company converts that infrastructure spending into usable model output has become a key differentiator in a market where compute costs remain one of the largest line items for AI labs.
The report does not specify the exact figures Baker cited, nor does it detail which specific models or product tiers were compared. Cost-per-token comparisons across AI providers can vary significantly depending on model size, context window length, whether the comparison involves input or output tokens, and pricing tiers offered to different customer segments. As such, broad claims about one company being cheaper than another should be understood as directional observations rather than precise, apples-to-apples benchmarks unless underlying methodology is disclosed.
Anthropic, the maker of the Claude family of models, and OpenAI, the developer of ChatGPT and the GPT model series, are widely regarded as the two most prominent players in the frontier large language model race, alongside other competitors building rival systems. Both companies have pursued enterprise and developer-focused pricing strategies, and cost efficiency has increasingly been cited by industry participants as a factor that could influence which providers win larger shares of enterprise contracts over time.
The claim attributed to Baker has been made. Readers should treat the comparison as a notable data point raised by a known market commentator rather than an independently confirmed industry-wide finding.
Market Impact
If accurate and sustained, a meaningful cost-per-token advantage could strengthen Anthropic's competitive positioning in enterprise deals, where large-scale token usage directly affects customer costs and vendor margins. Investors tracking the broader AI infrastructure trade, including chipmakers, cloud providers, and adjacent technology firms, often use such efficiency signals as one input when assessing which AI labs may be better positioned to scale profitably.
For now, the observation appears to be a single commentator's assessment rather than a fully verified industry benchmark, so any market reaction should be viewed cautiously until additional data or corroborating analysis from other sources emerges.
The comparison highlighted by Gavin Baker underscores growing investor focus on the operational economics behind frontier AI models, even as the specific figures and methodology remain unconfirmed beyond a single report.
Frequently Asked Questions
Who is Gavin Baker?
Gavin Baker is a technology-focused investor known for commentary on artificial intelligence companies, infrastructure spending, and market trends within the AI sector.
What does 'cost per token' mean in AI models?
Cost per token refers to the expense incurred by an AI company to process the basic text units, or tokens, that large language models use to generate and interpret responses. It is a common way to measure the computational efficiency of running an AI model.
Has the claim about Anthropic's lower costs been independently verified?
As of this report, the claim has been corroborated by only one source, and specific figures, methodology, or cross-source confirmation have not been provided.
Why does cost per token matter for AI companies?
Lower cost per token can improve profit margins as usage scales, potentially giving a company more flexibility in pricing for enterprise customers and strengthening its competitive position against rival AI providers.