We are in the early days of a seismic shift in the global AI industry. DeepSeek, a previously little-known Chinese artificial intelligence company, has produced a “game changing”“ large language model that promises to reshape the AI landscape almost overnight.

But DeepSeek’s breakthrough also has wider implications for the technological arms race between the US and China, having apparently caught even the best-known US tech firms off guard. Its launch has been predicted to start a "slow unwinding of the AI bet” in the west, amid a new era of “AI efficiency wars”.

In fact, industry experts have been speculating for years about China’s rapid advancements in AI. While the supposedly free-market US has often prioritised proprietary models, China has built a thriving AI ecosystem by leveraging open-source technology, fostering collaboration between government-backed research institutions and major tech firms.

This strategy has enabled China to scale its AI innovation rapidly while the US – despite all the tub-thumping from Silicon Valley – remains limited by restrictive corporate structures. Companies such as Google and Meta, despite promoting open-source initiatives, still rely heavily on closed-source strategies that limit broader access and collaboration.

What makes DeepSeek particularly disruptive is its ability to achieve cutting-edge performance while reducing computing costs – an area where US firms have struggled due to their dependence on training models that demand very expensive processing hardware.

Where once Silicon Valley was the epicentre of global digital innovation, its corporate behemoths now appear vulnerable to more innovative, “scrappy” startup competitors – albeit ones enabled by major state investment in AI infrastructure. By leveraging China’s industrial approach to AI, DeepSeek has crystallised a reality that many in Silicon Valley have long ignored: AI’s centre of power is shifting away from the US and the west.

It highlights the failure of US attempts to preserve its technological hegemony through tight export controls on cutting-edge AI chips to China. According to research fellow Dean Ball: “You can keep [computing resources] away from China, but you can’t export-control the ideas that everyone in the world is hunting for.”

End of the ‘Silicon Valley effect’?

DeepSeek’s emergence has forced US tech leaders to confront an uncomfortable reality: they underestimated China’s AI capabilities. Confident in their perceived lead, companies like Google, Meta, and OpenAI prioritised incremental improvements over anticipating disruptive competition, leaving them vulnerable to a rapidly evolving global AI landscape.

In response, the US tech giants are now scrambling to defend their dominance, pledging over US$400 billion in AI investment. DeepSeek’s rise, fuelled by open-source collaboration, has reignited fierce debates over innovation versus security, while its energy-efficient model has intensified scrutiny on AI’s sustainability.

Yet Silicon Valley continues to cling to what many view as outdated economic theories such as the Jevons paradox to downplay China’s AI surge, insisting that greater efficiency will only fuel demand for computing power and reinforce their dominance. Companies like Meta, OpenAI and Microsoft remain fixated on scaling computational power, betting that expensive hardware will secure their lead. But this assumption blinds them to a shifting reality.

DeepSeek’s rise as the potential “Walmart of AI” is shaking Silicon Valley’s foundation, proving that high-quality AI models can be built at a fraction of the cost. By prioritising efficiency over brute-force computing power, DeepSeek is challenging the US tech industry’s reliance on expensive hardware like Nvidia’s high-end chips.

This shift has already rattled markets, driving down the stock prices of major US firms and forcing a reassessment of AI dominance. Nvidia, whose business depends on supplying high-performance processors, appears particularly vulnerable as DeepSeek’s cost-effective approach threatens to reduce demand for premium chips.