There is always some brouhaha happening on the AI front. But this ‘amazing’ made-in-China model has far-reaching implications—for the US tech industry and the future of US hegemony. We look at the big picture—the nerdy tech analysis can be found in the reading list.
Wtf is DeepSeek?
On January 20, DeepSeek—a mostly unknown AI research lab—released an open source AI model called R1. It became an almost instant sensation—taking Silicon Valley by storm. VC demi-god Marc Andreessen tweeted: “Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen.. And as open source, a profound gift to the world."
First, the backstory: DeepSeek was founded in 2023—and is funded entirely by High-Flyer, a quantitative hedge fund founded by Liang Wenfeng. He is a fully made-in-China prodigy, as well. Born in 1985, Liang grew up in Guangdong—and graduated from the prestigious Zhejiang University. Soon after graduation, Liang founded High-Flyer with two college friends in 2015. It is now one of the best-performing quant hedge funds in China—as in, it uses quantitative analysis to place its investment bets. According to Wall Street Journal:
Liang prefers to be thought of as an engineer rather than a trader, according to people close to him. His High-Flyer was a pioneer in China in applying deep learning to computerized trading. The technique, modeled on the human brain, allows computers to analyze more diverse types of data.
Point to note: DeepSeek is unique in China in that it doesn’t rely on the usual big boy Chinese tech investors—like Baidu or Bytedance. It also was founded to focus on research not profit—a little like OpenAI (see details here).
And why is this R1 such a big deal?
It upends—or at least puts into question—the core premises of the AI industry—which until now has been shaped and driven by Silicon Valley.
Premise one: China is far behind the US in terms of AI capability. Not any more. Back in November, the Stanford University Index claimed “the gap is actually widening” The US is funding and creating more AI firms—by a mile. Hence, its AI models are outperforming the world.
Maybe not: Over a span of a week, two DeepSeek models are now ranked in the top 10 in the world in performance. A Google Gemini model is still #1—but DeepSeek bested Anthropic’s Claude and Grok from Elon Musk’s xAI.
Premise two: The US has maintained its dominance by simply spending more—because developing advanced AI is very, very, expensive:
It has far outpaced China in private AI investment, which hit $67.2 billion in the U.S. last year compared to $7.8 billion in China, according to the report.. It’s no surprise that the home of commercial AI powerhouses such as Google and Meta, along with relative newcomers like OpenAI and Anthropic, has produced many notable AI models that have influenced how the technology is being developed and applied.
Maybe not: The most stunning thing about DeepSeek is its balance sheet:
DeepSeek said training one of its latest models cost $5.6 million, compared with the $100 million to $1 billion range cited last year by Dario Amodei, chief executive of the AI developer Anthropic, as the cost of building a model.
Premise three: The assumption has always been that AI development requires super-advanced chips with great processing power (hence the hefty pricetag):
Training large language models (LLMs) requires a team of highly trained researchers and substantial computing power. In a recent interview with the Chinese media outlet LatePost, Kai-Fu Lee, a veteran entrepreneur and former head of Google China, said that only “front-row players” typically engage in building foundation models such as ChatGPT, as it’s so resource-intensive.
This is why Washington tightly controlled access to advanced chips—banning or restricting exports—a strategy that seemed to be working wonderfully back in 2022:
OpenAI’s release of ChatGPT took the world by surprise—and caught China flat-footed… Meanwhile, the Biden administration’s export controls, unveiled just a month before ChatGPT’s debut, aimed to cut China off from the advanced semiconductors essential for training large-scale AI models. Without cutting-edge chips, Beijing’s goal of AI supremacy by 2030 appeared increasingly out of reach.
Maybe not: Scarcity instead became the mother of innovation. Liang first stockpiled as many Nvidia chips as he could—even before he set up DeepSeek. They are now banned from export to China. Even so, these are hardly state-of-the-art chips today. DeepSeek’s engineers instead made of the most of what they had:
DeepSeek had to come up with more efficient methods to train its models. “They optimized their model architecture using a battery of engineering tricks—custom communication schemes between chips, reducing the size of fields to save memory, and innovative use of the mix-of-models approach,” says Wendy Chang, a software engineer turned policy analyst at the Mercator Institute for China Studies. “Many of these approaches aren’t new ideas, but combining them successfully to produce a cutting-edge model is a remarkable feat.”
Behold the bloodbath! US tech companies were the immediate casualties of the DeepSeek dhamaka—especially Nvidia. The crash on Monday wiped out nearly $600 billion in value—with CEO Jensen Huang's net worth plummeting $18 billion. But all AI superstars are in trouble: “Venture capital firms poured truly unheard-of sums into American AI companies — all based on what might suddenly be a stale thesis.” The VCs themselves are hardly safe:
This could be an extinction-level event for venture capital firms that went all-in on foundational model companies. Particularly if those companies haven't yet productized with wide distribution. The quantums of capital are just so much more than anything VC has ever before disbursed, based on what might be a suddenly-stale thesis. If nanotech and web3 were venture industry grenades, this could be a nuclear bomb.
But why does this matter to anyone other than tech bros?
Here is a quick but hardly exhaustive list of reasons.
Sastaa & tikaau: If AI tech costs less to develop, it will be cheaper for developers who use it to make products—and hence for us, the users:
Early last year, DeepSeek cut its prices for this service to a fraction of what other vendors charged, prompting the industry in China to start a price battle. Anthony Poo, co-founder of a Silicon Valley-based startup using generative AI to predict financial returns, said his company moved to DeepSeek from Anthropic’s Claude model in September. Tests showed DeepSeek performed similarly for around one-fourth of the cost. “OpenAI’s model is the best in performance, but we also don’t want to pay for capacities we don’t need,” said Poo.
DeepSeek has already started a price war inside China. Expect the same kind of pressure on global players like OpenAI. Also this: Unlike its rivals—with the exception of Meta—DeepSeek’s model is open-source, hence free to developers.
Global South rising: Lower prices and the open-source model give hope to countries that don’t have access to America’s deep pockets, say experts: “This could be a truly equalizing breakthrough that is great for researchers and developers with limited resources, especially those from the Global South.” Reminder: Donald Trump just announced a $500 billion Stargate project with the likes of OpenAI and Oracle—to deploy dollars to blow the competition out of the water.
The power of open source: Silicon Valley AI companies closely guard their research—to hoard their power. Chinese companies, ironically, have opened all their research to take on the American Goliath:
For many Chinese AI companies, developing open source models is the only way to play catch-up with their Western counterparts, because it attracts more users and contributors, which in turn help the models grow. “They’ve now demonstrated that cutting-edge models can be built using less, though still a lot of, money and that the current norms of model-building leave plenty of room for optimization,” Chang says. “We are sure to see a lot more attempts in this direction going forward.”
Also this: An open source, cheaper model may not be a coincidence. Chinese companies are consciously targeting countries like us: “[Alibaba’s] Qwen was deliberately imbued with fluency in ‘low-resource’ languages such as Urdu and Bengali, whereas American models are trained using predominantly English data.”
Point to note: DeepSeek isn’t the only Chinese company making waves. Another AI startup 01.ai was trained with only $3 million. TikTok parent company ByteDance released an updated model that claims to outperform OpenAI’s o1 in a key benchmark test.
The technology arms race: Technology is the secret sauce of global hegemony—as history teaches us over and again. The US has placed all its bets on AI to extend its global muscle. That’s why Joe Biden’s last act in office was to pass an unprecedented rule restricting the access to advanced chips—even to allies like India and the UAE. The world would soon run on AI—according to rules set by America. DeepSeek puts an end to that dream:
Beyond a source of national prestige, who leads on AI will likely have ramifications for the global balance of power. If AI agents can automate large parts of the workforce, they may provide a boost to nations’ economies. And future systems, capable of directing weapons or hacking adversaries, could provide a decisive military advantage. As nations caught between the two superpowers are forced to choose between Chinese or American AI systems, artificial intelligence could emerge as a powerful tool for global influence.
DeepSeek and other Chinese companies may never outdo their American peers—OpenAI’s Sam Altman is promising “PhD-level super-agents.” But they now offer an alternative, and a cheaper one at that.
Data point to note: Right now, there are 1,328 AI large language models in the world—of which 36% originate in China—which is the second-largest contributor to AI after the United States.
The bottomline: DeepSeek is the bearer of the worst kind of news for Washington: Pax Americana cannot be saved by a ban. That said, China has long dominated the global market by making stuff that is cheaper—not necessarily better.
Reading list
Forbes is very good at explaining exactly how DeepSeek managed its feat—breaking down the jargon. Indian Express rounds up some of the “wild use cases” making the rounds on social media. MIT Technology Review has the backstory on DeepSeek—while TIME explains how it succeeded despite US export controls. For the geopolitical analysis of these export controls, read Brookings and Carnegie Endowment. Economist has the big picture on the US-China rivalry.
