September 2026

BlueCallom Questions Whether the World Is Building Enough AI Capacity

Enterprise AI utilization could increase AI consumption hundreds of times compared with today’s Desktop AI like chatbot and copilot usage.

Zurich, September 2026 – While technology companies are investing hundreds of billions into AI data centers, chips and energy infrastructure, concerns about an emerging AI capacity bubble are growing. Swiss Enterprise AI company BlueCallom comes to a different conclusion. Based on experience from current Enterprise AI projects, the company believes the more important question may be whether enough capacity is being built.

BlueCallom’s observation comes from developing Enterprise AI applications that operate across complete business workflows. Unlike chatbots and copilots, these applications do not wait for an employee to enter a prompt. AI continuously researches, monitors situations, analyzes changes, prepares actions, responds to events and supports employees throughout a workflow.

This dramatically changes AI utilization.

From Occasional AI Use to Continuous AI Work

Today, an employee may interact with AI ten times per day and receive an answer within 10–20 seconds. This represents only 100–200 seconds of AI activity during a day of 86,400 seconds.

In BlueCallom’s Enterprise AI implementations, that utilization changes fundamentally. AI applications perform tasks in the background, autonomous agents monitor events, and AI supports employees throughout individual workflow steps.

“We realized through our customer projects that the biggest difference between today’s AI and Enterprise AI is not necessarily the intelligence of the models, but their utilization,” says Axel Schultze, CEO of BlueCallom. “As AI utilization increases, compute consumption increases dramatically. But so does the productivity generated by that capacity.”

BlueCallom calls this relationship AI Utilization.

At 50% utilization, an Enterprise AI environment could theoretically perform approximately 43,000 seconds of AI work per day. Compared with 150 seconds of typical AI interaction today, this represents almost 300 times more AI activity.

What If Two Billion Employees Become Fully AI-Enabled?

This observation led BlueCallom to a much larger question:

If two billion knowledge workers eventually become fully AI-enabled, how much AI capacity would the world actually need?

Current infrastructure numbers already appear enormous. Industry estimates suggest AI workloads could require more than 150 gigawatts of global data-center power by 2030.

Yet 2030 will almost certainly not represent a world with two billion employees operating highly utilized Enterprise AI environments.

BlueCallom is therefore developing a global Enterprise AI capacity model that combines observed Enterprise AI utilization, projected inference requirements and expected improvements in computing efficiency.

The objective is to quantify the infrastructure required if Enterprise AI eventually reaches utilization levels of 10%, 25% and 50% across two billion workers.

AI Access Is Not AI Productivity

The analysis also exposes what BlueCallom considers one of the misconceptions surrounding today’s AI adoption numbers.

AI Access ≠ AI Utilization ≠ AI Productivity.

Giving thousands of employees access to a copilot does not automatically create significant productivity gains.

The industrial revolution followed the same economic principle. Purchasing machines did not create industrial productivity. Utilizing them did.

BlueCallom expects AI utilization to become an important enterprise metric as companies move from experimenting with AI toward deploying mission-critical Enterprise AI applications.

“Today we are impressed by the enormous amount of AI infrastructure being built,” Schultze concludes. “But we are comparing that capacity with a world that barely uses AI. Once billions of people work with Enterprise AI throughout the day, today’s infrastructure investments may look surprisingly small.”

The question may therefore no longer be whether the AI industry is building too much capacity.

The question is how much capacity two billion fully AI-enabled employees will eventually consume.

More information is available at https://bluecallom.com/ai-capacity-hype-or-predictable-demand/