It is the electricity, stupid

Most of us already see the enormous impact large language models are having on daily life—both at work and at home. They make it faster and easier to build on our own writing and data, enrich it with new perspectives, and continuously learn.

They also make it significantly easier to understand customers and suppliers, and to approach them from entirely new angles.

The next—and even larger—step will come when AI agents carry identification and authorisation credentials. At that point, they can move beyond recommendations to executing strictly controlled, legally binding actions. This—and much more—becomes possible once identity wallets are widely adopted.

There is no reason to wait. FIDES already lists 34 business wallet alternatives, and the EU is moving toward mandatory wallet adoption across the public sector. Early use is eye-opening: it quickly leads to familiar innovation dynamics, with recurring questions like, “Can this also be used for X or Y?”

However, all of this progress will **dramatically increase the demand for a stable electricity supply. **This creates tension with climate goals. For example, despite major advances in renewables, China still relies on coal for over 50% of its energy.

In fact, AI leadership may depend as much on the** cost of electricity** as on access to advanced chips.

Recent estimates (snippets from Economist article) highlight the scale of the challenge:

• Global data-centre power demand could reach 68 GW by 2027 and 327 GW by 2030 (RAND)

• America’s ageing grid is already struggling, with a large backlog of data centres awaiting connection

• China, by contrast, continues rapid expansion: its grid—already the world’s largest—grew by over 500 GW in a single year, reaching 3,800 GW (more than double the US)

• Over the next five years, China is expected to add six times more capacity than the US

• Growth is driven by massive investments in wind, solar, nuclear—and still, coal

Cost differences are equally striking:

• Chinese data centres can secure electricity at around $0.03/kWh—roughly half typical US costs

• Residential pricing is managed separately, reducing public resistance to energy-intensive infrastructure

Yet China’s advantage is not absolute. A key constraint remains advanced chips. Since 2019, US export restrictions have limited access to leading-edge semiconductors (7nm and below).

Domestic alternatives, such as Huawei’s 7nm AI chips, are less powerful—but can compensate through scale (stacking many chips together). This increases energy consumption, but that trade-off becomes far less significant when electricity is cheap.

Exciting times - with huge opportunities to grab and challenges to be solved.