DeepSeek: What tumbling energy stocks say about AI’s power consumption

In this week’s Power Playbook: Chinese startup DeepSeek has become a force to be reckoned with, seeing stocks crash for major energy players and US tech companies.

DeepSeek – the Chinese startup that’s been on the lips of all industries this week – has had us pondering some very interesting questions; about power demand requirements for AI, about where companies are putting their money and about US big tech’s competitiveness with China.

What’s all the fuss about?

Earlier this week DeepSeek stunned markets and AI experts with a claim to have built its popular chatbot at a much cheaper cost than North American tech giants. Additionally, claims the company, its power demands are significantly less.

The result?

Sinking stocks - across the motherboard of US tech companies.

According to Forbes, tech titan Nvidia took one of the biggest hits, losing nearly $600 billion.

In the energy sector specifically, reported the Financial Times, stock sell-off was seen across major energy and infrastructure companies.

The likes of Siemens Energy, GE Vernova, Mitsubishi Heavy Industries, Schneider Electric, and ABB were all affected, as investors reevaluated AI-driven electricity demand projections.

Siemens Energy specifically, which had benefited from a surge in demand for gas turbines to power AI data centres, saw its stock plummet 21% in a single day despite reporting strong revenue growth.

This all results from claims by DeepSeek concerning the superiority of the tech when it comes to its efficiency. But still, what it highlights about AI power consumption is huge.

Have you read:
Powering data centres for an AI-driven future
Data centre executives pivot toward onsite power, per new report

The energy consumption cruncher

The cruncher – the crucial question – is all about the necessity of data centres and the massive compute power needed for AI.

A blog post by the IEA back in October 2024 explains it well.

Investment in new data centres, they say, surged in recent years, with much of the spending concentrated in the US, where annual investment in data centre construction has doubled in the past two years alone.

To illustrate, in 2023, overall capital investment by Google, Microsoft and Amazon, which are industry leaders in AI adoption and data centre installation, was higher than that of the entire US oil and gas industry – totalling around 0.5% of US GDP.

Average data centres are quite small in power terms, the Agency adds, with demand in the order of 5-10MW. But large hyperscale data centres, which are increasingly common, have power demands of 100MW or more, with an annual electricity consumption equivalent to the electricity demand from around 350 000 to 400 000 electric cars.

Which is why DeepSeek’s claim that its model needs less electricity overall received so much attention.

"If proven true, the efficiencies used within DeepSeek's open-source model can be applied by the hyperscalers to their models, which would result in a more moderated demand," analysts with Evercore ISI said in a note, reported by Reuters.

Although the questions raised by DeepSeek’s emergence are of huge importance, we also need to acknowledge that, as tech gets more efficient, so too will its use increase.

So said Microsoft chairman and CEO Satya Nadella in a Linkedin post: “Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of.”

The market is likely to shift significantly with developments such as those we see with DeepSeek. As these developments continue to unfold, I will endeavour to keep you updated with the Power Playbook.

Cheers,
Yusuf Latief
Content Producer
Smart Energy International

Follow me on LinkedIn