Cathie Wood highlights AI token price collapse as demand surges in ‘virtuous cycle’

The cost of using AI just got dramatically cheaper, and Cathie Wood wants everyone to know she called it. The ARK Invest founder pointed out that the average price per million tokens for large language models has dropped from $2.07 to $1.02, a decline of more than 50% that happened in a matter of weeks.

Wood framed the data as evidence of what she calls “enormous price elasticity of demand” for AI-driven productivity tools. Translation: cut the price, and usage doesn’t just tick up. It explodes.

The economics of getting smarter, cheaper

The $2.07 figure was the going rate on May 28. Roughly a month later, that number sits at $1.02. For context, a “token” in AI parlance is a chunk of text, roughly three-quarters of a word, that a language model processes. A million tokens is about 750,000 words, or the equivalent of feeding the model roughly ten full-length novels.

That kind of price compression doesn’t happen by accident. Wood attributes it to two forces working in tandem: aggressive price cuts from OpenAI, the market’s dominant closed-source provider, and the arrival of cheaper open-source alternatives from companies like DeepSeek and Kimi.

DeepSeek and Kimi are reportedly offering models at several multiples cheaper than their closed-source competitors.

Wood describes this dynamic as a “virtuous cycle” in AI’s early development stage. Lower prices attract more users, more users generate more data and revenue, and that funds further development, which eventually pushes prices even lower.

A thesis four years in the making

Wood has been beating the AI deflation drum since 2021, well before ChatGPT turned large language models into a household concept. Her argument has consistently been that AI’s transformative potential would follow a pattern familiar from other disruptive technologies: costs fall exponentially, adoption curves steepen, and the companies positioned at the center of that shift capture outsized value.

What the price war means for the AI landscape

OpenAI’s decision to slash prices is strategic, not charitable. By lowering the barrier to entry for developers and enterprises, the company is trying to lock in market share before open-source alternatives become good enough to make closed-source models optional.

DeepSeek’s emergence earlier this year rattled the AI establishment by demonstrating that high-performance models could be built at a fraction of the cost Silicon Valley’s incumbents were spending. Kimi has followed a similar playbook.

For enterprises evaluating their AI strategy, the math keeps getting more favorable. A task that cost $2.07 per million tokens a month ago now costs $1.02. Scale that across an organization processing billions of tokens per month, and the savings become material enough to justify expanding AI integration into workflows that previously didn’t pencil out.

That’s the price elasticity Wood is highlighting. It’s not just that existing users are paying less. It’s that entirely new categories of use become economically viable each time the price drops.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.