According to economist Michael Roberts, DeepSeek R1 is not just another AI model; it’s a seismic event in the tech world. Developed by Chinese firm DeepSeek, R1 rivals the capabilities of models from OpenAI, Anthropic, and Meta while costing a fraction to train. What makes it even more disruptive? It doesn’t rely on cutting-edge Nvidia chips and is completely open-source.
The DeepSeek R1 shockwave
Until now, AI supremacy seemed locked behind the enormous costs of advanced chip development and proprietary AI models. The U.S. government and its leading tech corporations assumed that their grip on AI innovation was secure, thanks to state-of-the-art Nvidia chips and billion-dollar training budgets. DeepSeek has shattered that illusion.
Instead of relying on the latest GPUs, DeepSeek built R1 using older, slower Nvidia chips—precisely the ones that U.S. sanctions still allowed for export to China. The assumption was that without access to the best hardware, China’s AI capabilities would lag behind. DeepSeek proved otherwise, demonstrating that powerful AI models can be trained without cutting-edge resources or astronomical budgets.
This has major implications. If DeepSeek R1 can perform at the same level as its American competitors while using a fraction of the computing power and cost, the entire economic model of AI development comes into question. The AI industry has been operating on the premise that bigger data centers and more expensive chips equal better AI. DeepSeek R1 suggests otherwise.
The open-source factor
DeepSeek R1 isn’t just about efficiency—it’s about accessibility. Unlike OpenAI’s ChatGPT or Google’s Gemini, which operate behind proprietary paywalls, DeepSeek R1 is open-source. This means anyone can run, study, and develop it freely. The analogy is similar to branded versus generic pharmaceuticals: while U.S. firms fiercely guard their AI “black boxes” for profit, DeepSeek has made its model freely available. This could democratize AI development, lowering entry barriers for smaller companies and independent developers worldwide.
For the Magnificent Seven (Apple, Microsoft, Alphabet, Amazon, Nvidia, Tesla, and Meta), this is an existential threat. Their business models hinge on the idea that AI must be monopolized and controlled by a handful of elite firms. If high-quality AI can be created at a fraction of the cost and shared openly, the logic behind their trillion-dollar valuations starts to unravel.
The stock market fallout
Investors took notice. On the same day DeepSeek R1 was announced, Meta unveiled another $65 billion in AI investment. Days earlier, Donald Trump had pledged $500 billion in tech subsidies as part of the so-called Stargate project, aimed at keeping U.S. firms ahead in the AI race.
But what if these investments are misguided? Billionaire investor Ray Dalio likened the current AI bubble to the dot-com boom of the late 1990s. While AI itself is transformative, he warns that investors are mistaking technological potential for guaranteed financial returns. The U.S. market reacted accordingly—five of the biggest AI-driven tech stocks (Nvidia, Alphabet, Amazon, Microsoft, and Meta) collectively lost nearly $750 billion in market value in a single day.
If DeepSeek R1 and future low-cost competitors can deliver AI applications at a tenth of the cost, the current AI arms race may become unsustainable. AI hyperscalers, chip manufacturers, and data center operators—all of whom have been expecting exponential growth—now face an uncertain future.
The U.S.-China AI war: A sanctions backfire?
U.S. efforts to restrict China’s AI capabilities seem to be backfiring. Despite sanctions and export controls designed to block China from accessing the most advanced chips, Chinese companies like Huawei and DeepSeek are making rapid progress. Huawei, for instance, has emerged as Nvidia’s top competitor in China for inference chips, working with firms like DeepSeek to optimize AI models that run on its Ascend chips.
This underscores a fundamental difference in approach. While the U.S. relies on private companies and free-market incentives, China is deploying state-led, strategic investment in AI and semiconductor development. The result? A more coordinated and resilient AI sector that is finding ways to bypass Western technological barriers.
Ray Dalio captured the essence of this shift when he stated: “Capitalism alone—the profit motive alone—cannot win this battle.” The AI war isn’t just about technology; it’s about economic models. Can the decentralized, profit-driven approach of the US compete with China’s state-backed, long-term industrial strategy?
The hidden cost of AI expansion
While DeepSeek R1 highlights how AI can be developed efficiently, the broader AI industry continues its reckless expansion. Companies like OpenAI, Google, and Meta are pouring billions into scaling AI capabilities—but at what cost?
AI’s environmental impact is staggering. A single ChatGPT query consumes nearly ten times the electricity of a Google search. Training AI models requires vast amounts of energy—Goldman Sachs estimates that AI data centers consume as much power as entire cities. Furthermore, AI-driven water consumption is becoming a major issue. Cooling AI data centers requires immense quantities of water, and Silicon Valley giants are increasingly securing control over local water supplies to meet their needs.
The AGI arms race: A race to the cliff?
DeepSeek R1 has shaken the AI landscape, but it hasn’t slowed the larger push towards AGI. Research firm Rosenblatt predicts that, rather than scaling back, tech giants will only accelerate their AI investments. “We expect the bias to be on improved capability, sprinting faster towards artificial general intelligence, more than reduced spending.”
But not everyone sees this as progress. Berkeley professor Stuart Russell warns that AI leaders themselves admit that achieving AGI could pose an existential threat to humanity. “The AGI race is a race towards the edge of a cliff,” he argues. Yet, the momentum behind AI’s development is so strong that even the risk of catastrophic consequences doesn’t appear to be slowing the industry down.
The real AI revolution: A tool for humanity or profit?
Michael Roberts’ analysis suggests that DeepSeek R1 demonstrates AI doesn’t have to be monopolized by a handful of trillion-dollar corporations. Its open-source model shows that AI can be accessible, efficient, and useful for society without requiring massive investment and environmental degradation.
As one commentator put it: “I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.” The question is whether AI will be used to empower individuals and communities—or simply to maximize profits for an elite few.
For now, the race continues. But, as Michael Roberts highlights, DeepSeek R1 has changed the game. The future of AI no longer belongs solely to the highest bidder.
This article is based on the views expressed by Michael Roberts on his blog, The Next Recession (https://thenextrecession.wordpress.com/).