ARK Invest has highlighted a surge in AI inference volumes even as token prices collapse, a divergence the firm says could drive broader AI adoption and open the door to more diverse applications.
Inference demand climbs
Inference is the stage where a trained AI model applies what it's learned to new data. It's the part of AI that runs in real time, powering everything from chatbots to image generators. ARK Invest's observation points to growing demand for that compute power, even as the market for tokens — often used to pay for AI services or participate in decentralized networks — takes a hit.
The firm didn't provide specific numbers, but the direction is clear: inference volumes are up, token prices are down. That gap, ARK Invest suggests, could be a catalyst.
Token prices in freefall
Token prices are collapsing, and that's not necessarily bad news for AI adoption. If tokens are cheaper, the cost of using AI services that rely on them could drop. That could make AI more accessible to smaller developers and startups, which might explain why ARK Invest sees this as a potential driver of broader adoption.
Lower prices don't automatically mean more usage, but combined with rising inference volumes, the trend could encourage experimentation. ARK Invest highlighted that this could foster innovation and a wider range of applications.
A catalyst for adoption
The combination of high inference demand and falling token prices could create a sweet spot. Developers might find it cheaper to build and deploy AI tools, while users could see lower costs for AI-powered services. ARK Invest's analysis suggests this could be the moment when AI moves beyond the early adopters.
For now, the market is watching whether the divergence holds. If inference volumes keep climbing while token prices stay low, the economics of AI could shift. ARK Invest's note adds to a growing conversation about how the cost of AI infrastructure interacts with the value of the tokens that power it.
The firm's view is that this moment could be a turning point, but the outcome depends on whether the demand for inference continues to outpace the token market's struggles.
