China’s AI industry has delivered a seismic shock to the world. DeepSeek, a little-known startup, has unveiled a generative AI model that is reportedly as powerful as OpenAI’s ChatGPT—but developed at a fraction of the cost. This breakthrough, described by Silicon Valley investor Marc Andreessen as a “Sputnik moment,” has rattled global tech markets.
The fallout was immediate. US tech stocks suffered a bloodbath, losing nearly $1 trillion in value in a single day. Nvidia, the world’s most valuable chipmaker, saw $589 billion wiped off its market capitalization in what is now the largest one-day loss for any company in US history. The panic was fueled by fears that China’s AI could compete—and even surpass—US advancements while circumventing Washington’s chip restrictions.
DeepSeek’s rise represents a stunning reversal of expectations. Just two years ago, China’s AI scene was seen as lagging behind the US. Beijing’s early attempts at ChatGPT-like models, such as Baidu’s Ernie and Tencent’s Hunyuan, were dismissed as inferior copies. Now, DeepSeek has positioned itself as a real contender, proving that China’s AI talent and research ecosystem are capable of groundbreaking innovation.
Why it matters
- DeepSeek’s AI is more than just another chatbot—it’s a symbol of China’s growing technological self-sufficiency. Its breakthrough challenges the long-standing assumption that the US holds an insurmountable lead in artificial intelligence.
- At the core of US confidence was the belief that China needed American technology—particularly advanced semiconductors from Nvidia—to train competitive AI models. To maintain its dominance, Washington banned the export of high-end AI chips to China in 2022 and further tightened restrictions in 2023. Many assumed these moves would kneecap China’s AI ambitions.
- Yet, DeepSeek has shattered that certainty. The company claims it trained its AI model using only 2,000
Nvidia chips , whereas OpenAI and Google typically require 16,000+ chips for comparable models. Even more astonishingly, OpenAI reportedly spent $1 billion on training ChatGPT, while DeepSeek did it with just about $6 million. - If these claims hold true, it could fundamentally reshape the AI industry. The assumption that cutting-edge AI requires enormous computing power, billion-dollar investments, and thousands of advanced chips may no longer be valid. Instead, DeepSeek has demonstrated that a highly efficient approach to AI training could level the playing field—allowing smaller players, and even adversarial nations, to catch up quickly.
Between the lines: How did DeepSeek pull this off?
1. A different approach to AI training:DeepSeek’s success hinges on an unconventional training method. Traditional AI models process vast amounts of data, requiring massive computational resources. DeepSeek’s approach is different—it prioritizes knowing where to look for answers rather than memorizing everything.
This is akin to a search engine librarian:Traditional AI models behave like a librarian who has read every book in the library and pulls from memory to answer questions. DeepSeek’s model doesn’t read every book in advance. Instead, it quickly finds the right book when asked a question—a more efficient and cost-effective approach. This technique, combined with a “mixture of experts” strategy—which assigns specialized AI models to different types of questions—dramatically reduces the computing power needed to train AI.
2. A strategic stockpile of Nvidia chips
While the US tried to block China from accessing advanced AI chips, DeepSeek found a loophole. Before Washington closed the door in 2023, DeepSeek and other Chinese firms stockpiled tens of thousands of Nvidia’s A100 and H800 chips.
These chips, though slightly less powerful than Nvidia’s cutting-edge H100, were still good enough for training DeepSeek’s AI. Some experts, including Elon Musk, have speculated that DeepSeek may have secretly acquired more high-end chips than disclosed.
3. China’s ‘AI super geeks’ approach
The brains behind DeepSeek is a 39-year-old Chinese hedge funder named Liang Wenfeng. As per a CNN report, in interviews with the state-linked financial outlet Yicai, early business associates described the future DeepSeek founder as somewhat “nerdy” and recalled “a terrible haircut” he once had.
Liang Wenfeng believes in hiring fresh graduates over experienced professionals. His rationale?
Experienced engineers follow conventional approaches. Young engineers are more willing to experiment and think outside the box. This philosophy appears to have worked. Liang’s team of under 140 researchers—mostly graduates from China’s elite universities—designed a breakthrough AI system in record time. Their success has boosted morale in China’s tech industry, with Liang now celebrated as one of China’s “AI heroes.”
A Trojan horse?
- The rise of DeepSeek raises a troubling question for Washington: Could this technology be a Trojan horse? Under Chinese law, all tech companies are required to “cooperate with national intelligence efforts.” This means that any data fed into DeepSeek’s chatbot could, in theory, be accessible to the Chinese state.
- Some analysts fear that DeepSeek’s chatbot could function as an AI-powered intelligence-gathering tool, subtly siphoning off user information under the guise of casual interactions. Others point to its built-in political censorship—when asked about topics such as the Tiananmen Square massacre or China’s treatment of Uyghurs, DeepSeek either refuses to answer or echoes the Chinese government’s official stance.
- With TikTok already under scrutiny in the US for alleged data harvesting, DeepSeek’s explosive rise is triggering similar concerns. The White House’s AI task force is reportedly investigating whether the platform poses a national security risk. “We should be alarmed,” said Ross Burley, co-founder of the Centre for Information Resilience. “Beijing has repeatedly weaponized its tech dominance for surveillance, control, and coercion,” Burley told Guardian.
What they’re saying
- DeepSeek’s sudden rise has sparked polarizing reactions across the tech world:
- President Donald Trump: Called DeepSeek a “wake-up call” for US industries, warning that China’s AI leap could threaten American economic and national security.
- Sam Altman (OpenAI CEO): Praised DeepSeek as “impressive” and welcomed the competition, despite previously framing AI development as a battle between democracy and authoritarianism.
- Elon Musk: Expressed doubts, suggesting that DeepSeek downplayed its use of banned US chips.
- US National Security Officials: Are now investigating potential data security risks, given that DeepSeek stores user data on servers in China—raising concerns about state surveillance.
What’s next?
- DeepSeek’s rise throws a wrench into Washington’s AI strategy. The US has aggressively restricted China’s access to advanced AI chips, believing that these controls would slow Beijing’s progress. But DeepSeek’s breakthrough suggests that China has found ways to innovate despite these restrictions.
- Stronger US export controls: The Trump administrations will likely tighten AI-related export bans.
- Restrictions may expand beyond hardware, targeting software and cloud-based AI services to further limit China’s progress.
- Increased scrutiny of Chinese AI companies: US and European regulators may investigate DeepSeek’s funding sources to determine whether the Chinese government played a role.
- Concerns over data security could lead to bans on DeepSeek’s AI in Western markets, similar to TikTok’s ongoing legal battles.
- US push for AI supremacy: The DeepSeek shock may accelerate government funding for AI innovation.
- Trump’s $500 billion “Stargate” AI initiative could gain bipartisan support as Washington scrambles to ensure US leadership in AI.
Bottom line
DeepSeek’s AI breakthrough signals a shift in the global AI landscape. While it’s unclear if China will ultimately overtake the US in AI, one thing is certain: Silicon Valley’s AI monopoly is no longer guaranteed.
(With inputs from agencies)