In the rapidly evolving artificial intelligence landscape, a new powerhouse seemingly emerged overnight — DeepSeek, a cutting-edge AI model developed in Hangzhou, China. DeepSeek’s arrival on the Google and Apple app stores on U.S. devices ignited intense debate last week among tech leaders, national security experts, and policymakers, with some describing it as a “Sputnik moment” for AI. […]
DeepSeek’s arrival on the Google and Apple app stores on U.S. devices ignited intense debate last week among tech leaders, national security experts, and policymakers, with some describing it as a “Sputnik moment” for AI. After DeepSeek became one of the most downloaded apps Monday morning, the tech market saw a stark drop in value, marked by shares for Nvidia, the graphics chips manufacturer serving as the backbone for OpenAI’s ChatGPT services, losing more than 12% of their value.
Despite the substantial stock slump earlier this week, Nvidia shares were up 0.6% at 125.41 on Friday. The market could remain volatile into this week as analysts weigh the long-term implications of DeepSeek’s capabilities and the competitive pressures it places on Western AI firms.
President Donald Trump addressed DeepSeek’s emergence in a speech to House Republicans on Monday, stating, “The release of DeepSeek AI from a Chinese company should be a wakeup call for our industries that we should be laser-focused on competing to win.”
Here’s what to know as the week of turbulence from the Chinese AI disruptor comes to a head:
What is DeepSeek and how does it compare to AI platforms like ChatGPT?
DeepSeek is the brainchild of Liang Feng, a hedge fund investor turned AI entrepreneur, whose previous venture, High-Flyer, specialized in quantitative trading.
The transition from finance to AI follows a natural trajectory, as firms seek increasingly sophisticated predictive models to navigate global markets. In March 2023, Liang announced that High-Flyer would pivot toward AI research, culminating in the launch of DeepSeek later that year.
What makes DeepSeek stand out is its efficiency, accessibility, and affordability. While its performance in some ways rivals that of OpenAI’s ChatGPT and Google’s Gemini, DeepSeek’s real advantage lies in its cost-effectiveness. The model reportedly operates at a 90% lower cost than its competitors, making advanced AI capabilities more accessible to a global audience.
Another reason for its overnight success is its cost-free notoriety when compared to other AI platforms that include premium tiers, such as the choice to pay $20 per month to gain access to premium features on ChatGPT.
But when something is labeled as free, there is typically a hidden cost. Experts have expressed caution about forfeiting certain privacy information and data that could be wittingly handed over to the Chinese Communist government, which can request access to the data servers of tech companies based in its country at any point and time.
“DeepSeek is a shared cloud service run in China with data being stored in China — potentially introducing unknown risks to data privacy, compliance mandates, and security controls,” said Bill Conner, CEO of enterprise automation firm Jitterbit and a former security adviser to the U.S. and British governments.
DeepSeek is legit, but experts say it’s shrouded in deception
There also remains an open question about whether the company has been completely forthright about the apparent low cost of its functions. One of the most contentious issues surrounding DeepSeek is its reported training costs and access to advanced AI hardware. The company claims it trained its flagship R1 model for just $6 million, a fraction of the $80 million to $100 million typically cited for training comparable models.
Even industry insiders like Scale AI CEO Alexandr Wang have questioned whether this figure excludes prior research, infrastructure, and algorithm development costs. Wang told CNBC earlier this week he believes China has significantly more Nvidia H100 GPUs, the chips widely used to build leading AI models, than people may think, citing the U.S. export controls that legally prevent the Chinese Communist Party-led nation from obtaining such materials.
Dr. Alfonso Berumen, a practitioner of decision sciences at Pepperdine Graziadio Business School, told the Washington Examiner that while DeepSeek’s efficiency is undisputed, its true costs remain somewhat shrouded in mystery and potential deception.
“There is still some ambiguity on the true cost. However, it is not really debatable that DeepSeek is more efficient than the existing competitors and the source of this is innovation,” Berumen said.
Another leading AI industry expert, Martin Vechev, the director of the Bulgarian Institute for Computer Science, Artificial Intelligence and Technology, or INSAIT, sought to demystify some of the claims surrounding the China-based AI company.
“The [DeepSeek] series of models from China have actually been public for years,” Vechev told the Recursive, an independent tech media platform.
Although the markets reacted as if the technology had been unveiled overnight, DeekSeek actually first made its announcement in December that its V3 model cost just around $6 million to train, indicating that the market panic was largely driven by the public release of its user cost-free product.
Vechev told Recursive the claim of “$5-6M cost of training is misleading” because it comes from claims in its V3 white paper that 2048 of Nvidia’s H800 cards “were used for *one* training, which at market prices is upwards of $5-6M.”
“Developing such a model, however, requires running this training, or some variation of it, many times, and also many other experiments. … That makes the cost to be many times above that, not to mention data collection and other things, a process which can be very expensive,” Vechev said, piling skepticism atop DeepSeek’s claims.
How Washington is reacting
Amid the turbulent week in AI markets, Trump met with Nvidia CEO Jensen Huang at the White House on Friday to discuss AI policy. The meeting, which was set up weeks in advance, provided an opportunity for Trump and Huang to address the evolving role of Nvidia’s advanced computer chips in artificial intelligence development.
DeepSeek’s emergence has also caught the attention of lawmakers in Washington, with some calling for stronger action to curb China’s AI advancements.
Rep. John Moolenaar (R-MI) voiced concerns over national security risks, stating, “The U.S. cannot allow CCP (Chinese Communist Party) models such as DeepSeek to risk our national security [and] leverage our technology to advance their AI ambitions. We must work to swiftly place stronger export controls on technologies critical to DeepSeek’s AI infrastructure.”
Sen. Josh Hawley (R-MO) has taken legislative action by introducing the Decoupling America’s Artificial Intelligence Capabilities from China Act, which seeks to cut off U.S.-China cooperation on AI.
Hawley warned about the broader implications of U.S. involvement with Chinese AI, stating, “Every dollar and gig of data that flows into Chinese AI are dollars and data that will ultimately be used against the United States,” Hawley wrote in a press release. “America cannot afford to empower our greatest adversary at the expense of our own strength. Ensuring American economic superiority means cutting China off from American ingenuity and halting the subsidization of CCP innovation.”
Congressional offices are also being warned not to use DeepSeek due to security concerns. The House’s chief administrative officer issued a notice stating, “At this time, DeepSeek is under review by the CAO and is currently unauthorized for official House use.”
The U.S. Navy has also raised security concerns regarding DeepSeek. In an email sent to personnel this week, the Navy explicitly warned its members not to use DeepSeek’s AI “in any capacity” due to “potential security and ethical concerns associated with the model’s origin and usage.”
What makes DeepSeek technically significant?
At the core of DeepSeek’s innovation is a mixture-of-experts, or MoE, architecture. Unlike traditional models, which activate all their computing power for each query, DeepSeek selectively engages only the necessary components, drastically reducing energy consumption and computational costs, according to an analysis from Built In tech journalist Ellen Glover.
For example, if a user asks DeepSeek to translate a document from English to Mandarin, a traditional AI model (like OpenAI’s GPT-4) would activate its entire neural network, making the process computationally expensive. DeepSeek, however, activates only the relevant language and translation components, making it significantly more efficient.
This approach has major implications for the AI industry, particularly as companies seek ways to reduce reliance on expensive hardware like Nvidia’s high-end GPUs.
China’s global AI disruptor faces competition at home
While DeepSeek’s rapid rise has made waves globally, it has also placed significant pressure on China’s other domestic AI leaders, particularly Alibaba. On Wednesday, Jan. 29, Alibaba’s Qwen 2.5-Max AI model was released, with the company claiming it outperforms DeepSeek-V3 as well as models from OpenAI and Meta, according to Reuters.
The timing of Alibaba’s announcement — on the first day of the Lunar New Year, a national holiday in China — suggests a scramble among Chinese tech giants to counter DeepSeek’s meteoric rise. The race for dominance has also prompted other Chinese firms, including ByteDance and Baidu, to roll out enhanced AI models in the past two weeks.
Bottom line: DeepSeek is no friend to privacy
Jitterbit’s CEO Conner also warned there are still plenty of risks posed by DeepSeek, stating, “In reality, DeepSeek represents a clear risk for any enterprise whose leadership values data privacy, security, and transparency.”
From the technological perspective, DeepSeek’s emergence has shown that bigger isn’t always better — that AI can be developed with less computational power and lower costs.
Still, efforts to create more complex programs in the future, such as artificial generalized intelligence, will require continuing hardware advances, meaning companies like Nvidia and AMD likely will not stay blunted in the long run.