Chinese AI Startup DeepSeek Shakes up Silicon Valley, Wall Street

Shares of Nvidia, the darling stock of the generative AI boom, are down 15.3% in midday trading on Monday (Jan. 27) as Wall Street grapples with the revelation that it could be overpaying for this emerging technology. Google, Microsoft and Amazon shares are also down.

The culprit: Chinese AI startup DeepSeek’s artificial intelligence (AI) models that perform on par with OpenAI’s and Google’s top models but at a fraction of the cost and with far fewer of Nvidia’s GPUs. DeepSeek was founded by Liang Wenfeng, who also co-founded a quantitative hedge fund in China called High-Flyer.

DeepSeek has released a family of models: V3 (AI chat) and R1 (reasoning models). The startup said R1 had cost $5.58 million and used 2,048 of Nvidia’s H800 chips to train, according to a paper published by its researchers. Anthropic CEO Dario Amodei has said that AI models currently cost $100 million to train, but it could hit $100 billion. AI executives have also said training would need thousands of AI chips, principally those made by Nvidia.

The sell-off comes as OpenAI, SoftBank, Oracle and MGX announced a project called Stargate last week that plans to spend $100 billion to half a trillion dollars to build AI infrastructure, mainly data centers.

Those and other AI cost estimates are now in question. Shares of Microsoft were down 3.7%, Google fell 3% and Amazon gave up about 1%.

“Artificial intelligence has reached a critical inflection point. The industry stands at a crossroads where escalating costs, environmental concerns, and innovation appear intertwined, threatening to stifle accessibility and adoption,” Gokul Naidu, senior manager at SAP, told PYMNTS. “Enter DeepSeek-R1, the model that’s turning heads in Silicon Valley and beyond for proving that high performance and affordability aren’t mutually exclusive.”

“DeepSeek challenges the narrative that innovation must come at an unsustainable cost,” Naidu said. “For businesses, this means AI could soon be accessible to small and medium enterprises, not just tech giants with deep pockets.”

DeepSeek could “catalyze a shift in Silicon Valley’s approach to AI,” Naidu said. “Historically, the tech epicenter has prioritized growth at all costs, often ignoring questions of efficiency. But DeepSeek forces a recalibration. With its dramatically lower cost structure, businesses and developers alike are reevaluating their priorities, focusing on efficiency-driven growth.”

DeepSeek is now the most downloaded app in the Apple App Store.

OpenAI CEO Sam Altman pushed back in a post on X last month, when DeepSeek V3 first came out, saying, “It is (relatively) easy to copy something that you know works. it is extremely hard to do something new, risky, and difficult when you don’t know if it will work. individual researchers rightly get a lot of glory for that when they do it! it’s the coolest thing in the world.”

DeepSeek also reportedly is subject to Chinese censorship, refusing to answer questions about Taiwan, for example. (Developers can get around this by revising the open-source models.)

Inexpensive Open source a Game-Changer?

Marc Andreessen, from VC firm Andreessen Horowitz, posted on X that “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.”

Naidu said DeepSeek’s open-source strategy “further amplifies its impact. OpenAI’s proprietary approach, while effective, limits transparency and collaboration. By contrast, DeepSeek invites developers worldwide to contribute, experiment, and build on its platform. This open-source ethos democratizes AI capabilities, fostering a more inclusive innovative ecosystem.”

Meta Chief AI Scientist Yann LeCun said in a post on LinkedIn that people concerned that China is overtaking the U.S. in AI are misreading DeepSeek’s impact. “The correct reading is: Open source models are surpassing proprietary ones. DeepSeek has profited from open research and open source. … They came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.”

Meta, whose strategy was to distribute open-source AI models, saw its shares up 1%. With open source, any developer can download and fine-tune, or retrain to customize, their AI models. However, developers typically still have to pay fees to access the model through APIs. While Meta doesn’t charge to access its API, DeepSeek does.

But Naidu said DeepSeek is also deeply discounted here: “DeepSeek charges $0.55 per million tokens, compared to OpenAI’s o1 model, which demands $15 per million tokens. For businesses wary of runaway operational bills, this is a game-changer.”

Many developers build their own AI applications atop the foundation models from OpenAI, Google, Anthropic and others. To access and use these models, such as inputting prompts and getting responses, developers have to pay AI pricing based on token volume. (A thousand tokens translates to roughly 750 words.)

What Chip Export Controls?

Naidu also pointed out that DeepSeek was also able to get around President Joe Biden’s export controls on advanced AI chips, which he recently expanded to carve out different levels of access for more than 120 countries.

“DeepSeek’s achievement is particularly impressive given its constraints,” Naidu said. “As a Chinese company facing export restrictions, it was unable to access the latest Nvidia GPUs, such as the H200, and instead relied on older H800 GPUs. Yet, the R1 model still managed to rival — or, in some tests, outperform — more expensive counterparts.”

“This feat speaks volumes about the ingenuity behind DeepSeek’s approach. It’s not just about throwing money at the problem; it’s about finding smarter, leaner ways to train and deploy AI systems,” Naidu added.

Naidu cited the following ripple effects from DeepSeek:

• Competitive pressure: U.S.-based companies may need to embrace cost efficiency as a competitive advantage rather than an afterthought.
• Wider adoption: Lower costs make AI viable for sectors that previously couldn’t afford it, from education to small-scale retail.
• Environmental responsibility: By reducing energy consumption, models like DeepSeek-R1 encourage sustainability in AI development.

“DeepSeek is more than a model — it’s a wake-up call for the entire AI industry,” Naidu said. “It challenges entrenched assumptions about the cost of innovation and offers a path forward where cutting-edge technology is both affordable and sustainable. As we move deeper into 2025, the conversation around AI is no longer just about power — it’s about power at the right price.

“And in this race, DeepSeek might just have redefined the finish line.”

The post Chinese AI Startup DeepSeek Shakes up Silicon Valley, Wall Street appeared first on PYMNTS.com.