Building and Pricing Intelligence: The New Foundations of the AI Economy

Bryhni talked about the AI data center project operated with Bitdeer in Norway. The operating model involves dynamically adjusting the load in response to real-time energy market signals, using large-scale batteries, and reusing heat.

Data centers and the future of AI pricing

Sarah Backhouse: To get more data centers built, providers need more price certainty and essentially a futures market for compute. What needs to be in place for that to work?

Haakon Bryhni: The pricing of AI compute is moving away from simple inputs like power, GPUs, and square footage. Over time, it will be priced closer to how AI is actually consumed at the token level and increasingly through market-based mechanisms such as auctions. That shift is necessary to support large-scale capital investment in new data center capacity.

From static load to grid-responsive compute

Sarah: How does this change how data centers interact with energy markets?

Haakon: In the Nordic energy market, power is traded in short intervals and through futures. A modern data center can interact directly with those markets. With batteries and software, the data center can adjust energy consumption dynamically, reduce load when prices are high, and ramp-up when power is abundant. It can also move jobs to other locations when needed.

Bitdeer’s Norway data center in practice

Sarah: How does this work in a real deployment?

Haakon: The data center being built in Norway is an AI data center operated by Bitdeer and is on track to become one of the largest in Europe. It is tightly integrated with the Nordic energy market. The facility includes large-scale batteries, on-site hydro generation located directly below the data center, and software that enables rapid adjustment of compute loads. This setup allows the facility to support the grid while optimizing operating costs.

Moving compute, not just electrons

Sarah: What makes this possible from a compute perspective?

Haakon: Modern GPUs can change load almost instantly, and workloads can be moved between data centers. At Bitdeer, data centers are being built globally alongside a software platform that can move jobs quickly, not only to save energy, but also to optimize resources across locations. This makes it possible to treat compute as a flexible, mobile resource rather than something locked to a single site.

Heat reuse and circular design

Sarah: Beyond power and compute, how are you thinking about efficiency?

Haakon: Energy reuse is built into the design. Excess heat from the data center is sent to adjacent agricultural operations, with plans that include fish farming. The goal is to create a circular ecosystem around the data center. In the future, energy reuse should not be optional; it should be an expected standard for large-scale data center development.

The takeaway

Next generation AI data centers will not be passive consumers of power. They will interact with energy markets, adjust load dynamically, reuse energy, and move compute across geographies.

Bitdeer’s Norway project offers a concrete example of this shift: AI infrastructure designed to operate as part of the energy system itself, rather than sitting outside it.