New Reality: How the emergence of China's AI DeepSeek is transforming the future of humanity.
Last year, at a private dinner in Silicon Valley, the CEO of one of the leading artificial intelligence development companies was asked what set his company apart from others creating "foundation models" — the systems underlying chatbots like ChatGPT. Does he have a protective moat? Yes, he replied, according to another CEO who was present. No one else has been able to raise as many billions of dollars as he has. That was his moat.
This is why giants like Meta Platforms Inc. have panicked over DeepSeek, a Chinese company that built an AI model on a budget roughly equivalent to the salary of a single AI development leader in the U.S. Its breakthrough now signifies a shift in the balance of power and reckoning for tech giants, who suddenly find themselves no longer guaranteed winners in the AI arena.
The reason is clear. Recently, OpenAI attracted the largest venture capital round in history ($6.6 billion), and Mark Zuckerberg stated that this year Meta will allocate up to $65 billion for AI projects. Tech giants are pouring vast amounts of money into AI, and now it seems like a waste.
The latest model R1 from DeepSeek, released on January 20, was built using AI computing power and chips costing only $6 million, which is significantly less than what companies like OpenAI and Alphabet Inc.'s Google spent. R1 has not only matched ChatGPT in AI model capabilities but also excels among the public, ranking at the top of app stores in the U.S., U.K., and Canada. It is open-source and free for personal use, and affordable for businesses. Silicon Valley oracle Marc Andreessen dubbed its emergence the "Sputnik Effect."
Does this mean that all the billions invested in chips, talent, and energy infrastructure in the U.S. have gone to waste? Not entirely, but these investments alone are no longer sufficient. DeepSeek has demonstrated that future winners in artificial intelligence will need both computing power and innovations in efficiency, which Chinese companies have been forced to develop due to U.S. restrictions on chip manufacturing.
The bleak global market on Monday also highlights serious "blind spots" that both Wall Street and tech companies have regarding AI. Firstly, the rising capabilities of DeepSeek have been known for several months; my Bloomberg Opinion colleague, Catherine Thorbecke, wrote about this company in June. However, it has been clear for over a year that companies like OpenAI, Anthropic, and Google lack moats around their foundational business models. Enterprises and consumers could easily switch from one to another. This situation was already risky, as a Google engineer warned in a notorious memo nearly two years ago. Despite this, OpenAI continues to burn cash without an obvious path to profitability.
Part of the problem is that Silicon Valley moguls have focused too much on creating AI that tops benchmark rankings in a self-serving competition of "my model is bigger than yours," and much less on turning these innovations into well-thought-out products that businesses and consumers could use. DeepSeek's success should serve as a wake-up call for companies like OpenAI to differentiate themselves by focusing on what their clients need. This still leaves them the opportunity to outpace competitors in China, where talent and the thirst for superiority in AI remain abundant.
Silicon Valley has another reason to remain optimistic. Just a few months ago, the tech industry was concerned about an impending limit to AI capabilities, as so-called scaling laws posed a brick wall to AI. But DeepSeek has shown that this is not the case, and that there are workarounds to overcome these limitations.
Most importantly, there is a shift that DeepSeek has caused in the competitive landscape. In the two years since ChatGPT's launch, major tech companies have become the largest financial beneficiaries of the generative AI boom. However, making AI cheaper, rather than just scaling it up or improving it, holds much greater significance for the rest of the world, something tech leaders in their insular circles have failed to appreciate. When smaller organizations can achieve great things with AI, it promises a healthier and more dynamic market.
The great irony is that Sam Altman of OpenAI and other AI leaders now have the chance to engage with workers displaced by their AI systems. They must strive to do more with less, or risk being displaced themselves.
The author expresses a personal opinion that may not align with the editorial stance. The author is responsible for the published content in the "Opinions" section.