Trending Now
We have updated our Privacy Policy and Terms of Use for Eurasia Group and its affiliates, including GZERO Media, to clarify the types of data we collect, how we collect it, how we use data and with whom we share data. By using our website you consent to our Terms and Conditions and Privacy Policy, including the transfer of your personal data to the United States from your country of residence, and our use of cookies described in our Cookie Policy.
{{ subpage.title }}
Is DeepSeek the next US national security threat?
Before DeepSeek released its R1 model last month, America’s long-term AI dominance felt like a sure thing.
DeepSeek is a Chinese startup, born from a hedge fund, that claims to have used a fraction of the computing power of US competitors while making an artificial intelligence model that rivals the best that Northern California’s labs have to offer. Critics have alleged that the company has been dishonest about claims it only spent $6 million training the model. But for anyone taking DeepSeek at face value, it has been a revelation that sent shockwaves not only through Silicon Valley but also through Wall Street and Washington.
The Biden administration spent the past few years clamping down on powerful US-made chips flowing into China, but evidently, DeepSeek figured out how to build a great model with a dearth of high-tech resources.
“It shines a spotlight on the limits of the US export control system,” said Xiaomeng Lu, geo-technology director of Eurasia Group. “Technology has evolved in a way that regulators failed to anticipate.”
“Necessity is the mother of invention” – that’s how Jack Corrigan of Georgetown’s Center for Security and Emerging Technology put it. “US efforts to hobble China’s AI sector created a need for Chinese developers to innovate a more efficient approach to AI.”
But DeepSeek’s impact goes beyond its own efficiency. It’s an open-source model, meaning its code is available for anyone to use and modify. “Due to the open-source nature of their model, it will be much harder to restrict access to it entirely,” said Valerie Wirtschafter, a fellow at the Brookings Institution. “The other more pragmatic question is whether Congress has the appetite for more whack-a-mole-style tech regulations, given the chaos that has unfolded since the passage of the TikTok ban.”
US government agencies such as NASA and the Navy have banned DeepSeek models on their devices, as did Congress, but there’s been no US effort to try and ban it more widely among the public, as Italy did on Thursday, citing unresolved data privacy concerns. And America’s top cloud providers, including Microsoft Azure and Amazon Web Services, have already added access to R1.
Justin Sherman, founder of Global Cyber Strategies, says that the Trump administration has a toolbox to “screen, restrict, and even expel non-US tech from the US tech supply chain on national security grounds,” particularly through the Commerce Department’s Information and Communications Technology and Services. Still, he cautions against letting “stock market temperaments, reductive China panic in Washington, and media overinflation of industry AI claims” steer nuanced policy decisions.
DeepSeek’s true threat is likely strategic rather than technical. “DeepSeek’s latest model raises the question of what happens if China becomes the leader in providing publicly available, freely downloadable AI models,” said Sam Winter-Levy, a fellow at the Carnegie Endowment for International Peace. “While the US is obsessed with the race to see who can build the single biggest and most powerful model, perhaps even artificial general intelligence, the Chinese might win the race to see who can build really useful and cost-effective models that will be used by people and companies around the world.” At a minimum, China’s overnight success has quickly leveled the playing field for US-China competition over technology.
Perhaps then the answer to DeepSeek requires a rethinking of what American dominance in AI really means. Banning any specific app or model would just be a Band-Aid on a bullet wound.
This illustration photo shows the DeepSeek AI application logo on a black background displayed on a cell phone with a kaleidoscope-effect China flag in the background.
What DeepSeek means for the US-China AI war
A Chinese startup might have achieved what many thought was impossible: matching America’s best artificial intelligence systems at a fraction of the cost.
DeepSeek's latest AI model, DeepSeek-R1, was released earlier this month. The open-source model performs as well as top models from OpenAI and Google while using just a fraction of the computing power and cost to develop; it’s also a fraction of the cost to use.
DeepSeek claims that it only needed $6 million in computing power to develop the model, which the New York Times notes is 10 times less than what Meta spent on its model. The R1 model received the fourth-highest score on Chatbot Arena, which crowd-sources evaluations to rank large language models by capability, only behind two of Google’s Gemini models and ChatGPT-4o and ahead of Anthropic’s Claude 3.5 Sonnet.
If you take DeepSeek at its word, then China has managed to put a major player in AI on the map without access to top chips from US companies like Nvidia and AMD — at least those released in the past two years. Joe Biden’s administration placed strict export controls on these chips, so if the company has had access it may not be forthright about that.
For now, the US markets are indeed taking DeepSeek at its word. Nvidia stock fell nearly 17% on Monday, erasing a record sum from its market capitalization — $589 billion in a single day. The Nasdaq stock exchange ended the day down 3%, as a result.
The revelation about DeepSeek has come as Donald Trump tries to spur AI infrastructure in the United States, heralding the $500 billion Stargate project. But China’s new open-source model might have just changed the landscape when many thought the United States was running away with the race.
In a speech Monday evening, Trump called news of the DeepSeek model a “positive” due to its cheap cost but said American industry needs to compete. “The release of DeepSeek, AI from a Chinese company should be a wakeup call for our industries that we need to be laser-focused on competing to win.”Biden has one week left. His chip war with China isn’t done yet.
On Monday, the US Commerce Department announced new export controls on advanced chips used to train and run artificial intelligence, the latest in a series of increasingly tough restrictions enacted by the Biden administration in the past few years. It’s primarily a mechanism for maintaining American dominance in artificial intelligence while also cutting off adversaries — first and foremost China — from access to the chips they need to level the playing field commercially and militarily.
Under the new export regime, the US will use a three-tier system: Companies from the US and 18 close allies, including the UK, Germany, and Japan, are fully free to buy these chips. Those from countries subject to US arms embargo, such as China and Russia, are completely cut off from acquiring high-powered chips. And countries from every other country will face a set cap restricting the number of AI processors they can buy each year without special permission. This latter group even includes strategic partners such as Mexico, Switzerland, and Israel.
“This policy will help build a trusted technology ecosystem around the world and allow us to protect against the national security risks associated with AI, while ensuring controls do not stifle innovation or US technological leadership,” US Secretary of Commerce Gina Raimondo said in a press release.
But the chip industry isn’t happy — particularly America’s most important chip designer, Nvidia, which said that the rules are “misguided.”
“While cloaked in the guise of an ‘anti-China’ measure, these rules would do nothing to enhance US security,” Ned Finkle, Nvidia’s vice president of government affairs, wrote in a statement. “Rather than mitigate any threat, the new Biden rules would only weaken America’s global competitiveness, undermining the innovation that has kept the US ahead.” The Semiconductor Industry Association criticized the timing — just days before the presidential transition.
Tinglong Dai, a professor at Johns Hopkins Carey Business School, said that US companies are increasingly aligned with national security concerns and “face ever greater challenges to making a profit or even operating in China.” In that way, he said he expects that short-term discomfort will give way to companies falling in line with national security priorities.
Xiaomeng Lu, director of geo-technology at Eurasia Group, expressed concern about implementation. “Under this regime, only a small number of trusted allies have unrestricted access to high-end semiconductors,” she said. “Data centers located in over 100 countries will have to apply for licenses through an onerous process, even if they are owned and operated by US cloud service providers.” She added that the rules could discourage many countries from buying US products.
“This rule is not just about China, but it may make China’s chip smuggling efforts more difficult,” said Jacob Feldgoise, a data research analyst at Georgetown’s Center for Security and Emerging Technology. “The regulation could hurt American businesses, including AI chip companies and cloud service providers, but perhaps more importantly, the rule may hurt the United States’ relationship with a set of allies to which the rule didn’t give preferential treatment.”
Jeremy Mark, a nonresident senior fellow at the Atlantic Council’s GeoEconomics Center, said that China is deeply embedded in the semiconductor supply chain, such as supplying important rare earth metals to chip makers. In this way, the US is risking a “serious trade conflict” in further cutting off China.
But there may be more to the story. “There has been clear, bipartisan support in Washington for export controls, and that has empowered several government agencies to act,” Mark said. “However, it’s not clear how much the latest initiatives have been driven by a White House decision to hand Trump a fait accompli.”
Trump won’t want to appear weak on China, but he also doesn’t want to alienate important industry players. Dai said he expects Trump to maintain a tough stance on China and to “tweak the current export rules — because he’s Trump after all.” The rule doesn’t go into effect for companies until May 15, Feldgoise noted, giving Trump time to revise the rules if needed.
As for China, the country is forging ahead even without access to the top chips. Recently, the Chinese startup DeepSeek released an open-source large language model that has impressed outsiders. This development proves it’s “possible to build a quality model with less advanced chips,” Lu said. Meanwhile, the new Commerce Department restrictions will “undermine US companies’ competitiveness in the face of robust competition from China,” she adds.The NVIDIA logo seen at the American GPU manufacturer NVIDIA Taipei office.
Nvidia forges deals in American Southwest and Southeastern Asia
The California-based chip giant is negotiating with Taiwan Semiconductor Manufacturing Company, or TSMC, the world’s top contract chipmaker, to manufacture its top-of-the-line Blackwell AI processors at TSMC’s Arizona facility. TSMC has invested billions to bring its high-tech manufacturing to the Southwest US, thanks in part to a $6.6 billion cash infusion from the Biden administration as part of the CHIPS and Science Act. Apple and AMD have reportedly already signed on to get their chips made in the Arizona plant when it starts production in the first half of 2025. That said, the chips won’t be entirely made in America: Final packaging is done back in Taiwan, which complicates and prolongs an already lengthy manufacturing process.
Halfway around the world, Nvidia CEO Jensen Huang met with the Thai and Vietnamese prime ministers last week as the company makes inroads in Southeast Asia. Nvidia also announced plans to establish Nvidia’s first research and development center in Vietnam, along with the acquisition of Vietnamese healthcare startup VinBrain for an undisclosed sum. In Thailand, the company signed a cloud deal with a company called SIAM.AI Cloud. Huang also emphasized the importance of “sovereign AI,” meaning that every country should have its own AI infrastructure and models.
In China, however, Nvidia is facing new scrutiny: The State Administration of Market Regulation is reportedly investigating whether the chipmaker violated antitrust laws when it acquired the Israeli-American company Mellanox in 2020. China previously gave conditional approval of the nearly $7 billion deal, but more than four years later, with the US restricting Nvidia from selling its most powerful chips to Chinese companies, the country is seeking new ways to gain leverage. A Nvidia spokesperson said the company is “happy to answer any questions regulators may have about our business.”
The logo of Huawei's global flagship store is displayed in the Huangpu district of Shanghai, China.
The US is thwarting Huawei’s chip ambitions
The US government under President Joe Biden has imposed significant export controls not only on US-made chips but also on semiconductor manufacturing equipment necessary for Huawei to mass produce its own chip designs. US rules have largely cut Huawei off from the most powerful machines made by Dutch lithography company ASML, which essentially makes stencils to imprint miniature designs on chips for mass manufacturing, and TSMC, the world’s largest contract chipmaker. (The US Commerce Department is investigating how Huawei chips recently ended up on TSMC assembly lines.) Instead, Huawei relies on the Chinese chip manufacturer SMIC, which uses less powerful models of ASML machines.
But despite Huawei’s ambitions, Reuters reports that the company has been struggling with these restrictions to make effective chips at scale. For the Ascend 910C, the yield rate — the percentage that comes off manufacturing lines fully functional — is reportedly only 20%, while experts say a 70% yield rate is needed to be commercially viable. China’s top chip designer will need to make a breakthrough with limited resources to make good on its public promises to compete with Nvidia.Recently launched Amazon artificial intelligence processors that aim to tackle Nvidia and the chips made by the other hyperscalers such as Microsoft and Google are shown at an Amazon lab in Austin, Texas, in July 2024.
Amazon’s grand chip plans
Amazon is working on the third generation of its AI chips, called the Trainium2, which industry insiders told Bloomberg was a “make-or-break moment” for the company’s chip ambitions.
Luckily, they already have one important customer’s buy-in: Anthropic, which makes the chatbot Claude. On Nov. 22, Amazon announced it’s investing another $4 billion into Anthropic, doubling its total investment to $8 billion. As part of the deal, the Claude maker — perhaps the main rival to OpenAI — will continue to use Amazon’s Trainium series of chips. Amazon makes and invests in AI software and has the cloud infrastructure needed for AI – so if it can conquer the chip industry and produce chips comparable to the top models from Nvidia, it could become a dominant player in artificial intelligence.Hard Numbers: Doctor vs. machine, Pony rides to an IPO, Hot chips, Foxconn’s crazy demand
4.5 billion: A Chinese self-driving car company, called Pony AI, is attempting to go public on the Nasdaq stock exchange. The company, which is backed by the Japanese automaker Toyota among others, is seeking a $4.5 billion valuation for its initial public offering. The company previously tried to go public in the US through a blank-check company, but plans fell apart when China cracked down on such deals.
72: Nvidia's new Blackwell AI chips are reportedly overheating when installed in server racks designed to hold 72 chips. The company has already faced delays due to design flaws with these chips and is now asking suppliers to modify the designs of the racks numerous times. This issue could further delay sales to the largest tech companies in the world, such as Google and Meta.
A microchip and the Taiwanese flag in an illustration.
TSMC set to get its CHIPS money
The award marked the first finalized disbursement of the CHIPS Act since it was passed in 2022 and will go toward building TSMC's three new chip factories in Arizona — helping offset the $65 billion cost.
A total of $36 billion has been approved by Congress and directed by the Commerce Department to foreign companies such as TSMC and Samsung, as well as US companies including Intel and Texas Instruments. The delays, in addition to the normal snail’s pace of bureaucracy, stem from the fact that the Commerce Department spent much of the past two years negotiating with semiconductor companies, procuring specific commitments before finalizing the amounts they’d receive.
President Joe Biden needs to disburse the payments quickly because the future of the CHIPS Act is in question. When Donald Trump takes office in January, he may fulfill campaign promises to dismantle the Biden initiative or ask the Republican-controlled Congress to repeal it. Alternatively, the president-elect could carry on with the disbursements, which could further a bipartisan goal of beating back China’s AI ambitions.