It cannot have been a coincidence that China’s DeepSeek R1 was released on the day of Donald Trump’s inauguration. America’s greatest technology leaders were lined behind him, from Elon Musk to new recruit Mark Zuckerberg.
China’s leadership must have known the maelstrom they were about to unleash. Within days, DeepSeek pricked US technological triumphalism, stunning the world and shifting the geopolitical technology power base.
News broke last weekend that DeepSeek’s app had overtaken downloads of all US competitors. Not only was it on par with ChatGPT, but word quickly spread of how much cheaper DeepSeek’s AI reasoning model was, using less computing power and fewer chips. The reaction was seismic.
US tech and energy stocks plunged, losing $1trn in market value – with chip manufacturer Nvidia also routed, losing $600bn off its value in a matter of hours. Only days before, Sam Altman and Trump had unveiled their $500bn investment fund to power more AI infrastructure and development for Open AI, with Trump calling it “the largest AI infrastructure project by far in history” that would keep the “future of technology here in the US”.
Well, not anymore.
Here was cheeky start-up DeepSeek, saying they had worked out a way of doing more with much less and, furthermore, they made it open source and free to use for everyone. It was a stunning victory for China and they revelled in it.
Overnight, global assumptions on who will lead the artificial intelligence race were upended and America’s tech titans had to eat humble pie. DeepSeek’s R1 launch was called “AI’s Sputnik moment” by the Silicon Valley venture capitalist Marc Andreessen, his reference to the Soviet Union satellite launched in 1957. Trump was forced to call DeepSeek “positive” and a “wake-up call” for America’s tech industry.
Meanwhile, as downloads increased in their millions, Altman vengefully accused DeepSeek of copyright infringement – a bit rich given its history of data scraping. Doubts about the possible security issues of downloading DeepSeek began to quickly surface with very little basis. And the controversy continues. Australia is the latest country to urge its citizens to be careful while using DeepSeek, over fears of data leaks to China.
But at its root, DeepSeek is a classic disruption story: a smaller boot-strapped company oversteps the innovations of its far larger, better funded competitors. Not exactly David and Goliath – DeepSeek was self-funded via its owner and CEO Liang Wenfeng and his profitable Chinese quantitive trading hedge fund. But not far off.
One week on – the initial volcanic reaction over and the race for AI supremacy between China and the US firmly back on – does DeepSeek still live up to the hype? Is this the chatbot the world has been waiting for? And, given its lower energy costs to run and how energy hungry these chatbots are, can DeepSeek help save the planet from ever growing emissions?
For many, DeepSeek’s entry into the market is a true game-changer, and is discussed alongside the release of ChatGPT in 2022, because its lowered costs will boost AI adoption across every economy. That means more jobs, higher productivity, and increasing investment across multiple businesses, as they compete to use this technology. It plays straight into all those wonderful positives Rachel Reeves has been talking up this week in her growth speech, and the announcement of the Oxford-Cambridge Arc. Europe’s tech stocks rose, after all, while America’s plunged.
As Alexandra Mousavizadeh, CEO and founder of UK start-up Evident Insights, which benchmarks the adoption of AI globally in the banking sector, pinpoints: “This is a story about open-source models. DeepSeek is challenging the business model of Open AI and Dario Amodei’s Anthropic because they charge for the use of their most sophisticated AI chatbots, while DeepSeek, like Meta’s Llama, is free. And because of its highly innovative reduction in costs, this is going to accelerate the adoption of AI all over the world.”
In other words, DeepSeek has blown open the competition, and called into question the assumption many of us were being forced to swallow: that having a few tech titans controlling the tech ecosystem was a positive. Which one of us didn’t feel increasingly uncomfortable with the power and sway these tech bros were hoarding, including Nvidia and its control of AI chip production?
Brent Hoberman, chairman and co-founder of the Founders Forum Group, a global UK-based community of businesses and technology entrepreneurs, agrees: “The application layers that can be built on top of DeepSeek’s chatbot is where the real value is going to come – software solutions that layer on top of DeepSeek’s model for AI powered consumer and enterprise software and that is going to power innovation.”
Iain Mackie, the CEO of Malted AI, which builds smaller LLM models for enterprises and is also based in the UK, agrees. “Model companies represent just one facet of the AI market's potential. While language models are powerful, the real impact will come from technologies that drive AI adoption to solve high-value problems,” he says.
Hoberman has long been trumpeting the importance of our booming UK tech start-up scene. We are now the third largest AI market in the world, with global talent based here alongside pioneering AI firms. And this is despite not having developed our own large language model.
Our advantage in the UK post-Brexit also lies in our ability to deliver our own regulatory framework outside of the EU, which has meant many huge US technology companies have made the UK their largest European outpost – from Meta to Google and, more recently, AI firms like Scale AI. This in turn increases the pull of talent and more investment in a virtuous circle. Hoberman expects there will be a surge in new investment and applications in response.
It also means France’s large language model, Mistral AI, which caused huge excitement last year, but then fell out of favour, could at any moment produce a new version that could act like DeepSeek, flipping the AI wheel again. Any uplift in Europe will also help the UK. DeepSeek has meant those invested in the AI race feel they have everything to play for again.
But what of our planet? This is a voraciously hungry technology. And the fear is the development of AI is going to move far faster than the innovation around green energy, meaning as we increasingly adopt artificial intelligence and these large language models into every area of our lives from education to defence and health, we are setting ourselves on a direct collision course with climate goals.
If DeepSeek can use far less GPUs, surely this is better for our emissions? Sady not, says Alexandra Mousavizadeh. “The energy consumption rises on how fast we run these models – seconds over minutes – and that is going to get ever faster. And then Agentic AI is coming down the track and this uses a ton more energy…”
And this is the conundrum we still face: how are we going to drive up innovation, productivity, and jobs, without AI also hoovering up our energy sources, when green energy development is trailing so far behind.
It’s back to those pesky bat tunnels again.
