There's been a surge in AI use recently. Here's what's behind it.

• AI token usage soared in recent weeks, while Nvidia GPU pricing has jumped.
• AI is evolving from chatbots to AI agents and other forms of automated computer use.
• That's driving exponential growth in AI inference, suggesting robust demand for advanced models.

AI demand has surged in recent weeks, according to new data that could support tech giants' decisions to dramatically ramp up investment in this area.

OpenRouter, which helps developers access different AI models, has seen activity roughly double in the first weeks of 2026.

This is measured by the number of AI tokens OpenRouter processes. AI models break down words and other inputs into numerical tokens to make them easier to process and understand. One token is about ¾ of a word.

OpenRouter handled 13 trillion AI tokens in the week that ended February 9. That's up from 6.4 trillion during the first week of January.

Seen over a year, this recent surge is even more striking. AI models that OpenRouter taps into include Google's Gemini, Anthropic's Claude, xAI's Grok, OpenAI's GPT range, and a host of open-source offerings from DeepSeek, Moonshot AI, and others.

This data excludes AI tokens processed directly by the major AI companies, so it represents a small fraction of total activity. For instance, Google was processing 1.3 quadrillion tokens a month this past summer. However, the recent rate of change in OpenRouter's data is a valuable signal.

Anand Iyer, a partner at VC firm Lightspeed, said the recent surge has been driven by an explosion in AI agent activity and especially the rapid emergence of OpenClaw, an open-source agentic system.

That's sparked an exponential increase in inference, which is how models, agents, and other AI services are run in the cloud.

"The demand for AI inference right now is coming from agentic platforms, especially OpenClaw," Iyer told Business Insider this week. "We moved from simple chatbot experiences to agentic automation and execution, which has driven exponential growth of OpenClaw installations and OpenRouter usage."

GPU pricing rebounds

There are other signs of rising AI use. Data compiled by Bloomberg suggests demand for Nvidia AI chips has increased lately. The chart below shows the cost to rent Nvidia H100 GPUs. As you can see, prices have rebounded strongly since early December, which implies robust AI demand.

OpenClaw launched in November 2025. That coincides with the H100 rebound here.

Vibe coding jumps

A third data point comes from Barclays analysts, who recently looked at online traffic to vibe coding services, including Lovable, Replit, and Wix's Base44.

The consumer website-based vibe coding services Barclays tracks saw traffic jump 17% month over month in January.

"This was the strongest growth since April 2025," the analysts wrote in a note to investors this week. "The uptick in interest here was driven largely by frontier model improvements in late 2025 that made these platforms more effective."

AI sustainability

The debate still rages over whether AI will be used enough to justify all the heavy investments being made. Big Tech companies recently increased capital expenditure plans to eye-watering levels, raising new concerns about a potential AI bubble.

This new AI usage data may suggest that demand could be strong enough to support such heavy spending.

"AI capex today, as we know it today, is largely driven by training requirements from large frontier labs," Iyer said.

Training is a data- and energy-intensive process in which AI models are created in massive data centers. Inference comes after models are run.

"The sustainability is warranted as long as the revenue produced by inference (which is growing exponentially) will justify the output from the training," Iyer said. "Which seems to be working so far."

Machines don't sleep

OpenClaw is part of a broader evolution of generative AI. The era kicked off with a bang when OpenAI launched ChatGPT in late 2022, and the chatbot became the fastest-growing tech product in history.

More recently, AI models have begun to support autonomous digital agents and similar tools that can use computers independently to accomplish complex tasks.

OpenClaw's open-source technology supports agents that help users automate coding and other workflows by accessing computer files, email, calendars, and messaging services.

In January, Anthropic unveiled Claude Cowork, an AI agent that handles tasks such as document generation and file management.

The startup rolled out Claude Sonnet 4.6, its latest AI model, this week. It's gotten better at what Anthropic calls "computer use," where AI models and agents use computers in similar ways to humans to get tasks done.

"The model sees the computer and interacts with it in much the same way a person would: clicking a (virtual) mouse and typing on a (virtual) keyboard," Anthropic wrote in a blog on Tuesday.

This type of machine-to-machine activity is behind a lot of the surge in AI token use lately. It makes sense: most people can use a computer for seven or eight hours a day. Then, they need a break or a nap. AI models never sleep and can hack away at a task uninterrupted until it's done.

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.