China's ChatGPT-killer DeepSeek has OpenAI, Microsoft, Meta and Google worried

Up until now, the US has been the undisputed leader in artificial intelligence, with tech giants like OpenAI, Google, Meta, and Microsoft at the helm of driving innovations in the sector. Starging from making people familiar with GenAI to conversations around AGI or Artificial General Intelligence, it has been US-based tech companies that have been the core of all conversations. Their massive investments, cutting-edge hardware, and top-tier talent have kept them ahead of the pack.

But 2025 is proving to be a turning point. A new player from China, DeepSeek, is making headlines with its release of DeepSeek-V3, a powerful AI model that has already begun to rival the best models from US companies—without the hefty price tag typically associated with such breakthroughs.

More from Tech

How ChatGPT is becoming everyone’s BFF and why that’s dangerous America ready for self-driving cars, but it has a legal problem

DeepSeek’s surprising success has upended long-held assumptions in the AI world. While the US has maintained its AI dominance through billions of dollars in funding and top-of-the-line resources, DeepSeek has shown that ingenuity and smarter use of resources can achieve equally impressive results. The question now is whether China’s AI prowess could shift the global balance of power in tech.

DeepSeek’s low-cost, high-impact model

DeepSeek’s real achievement lies in its ability to develop a cutting-edge AI model while spending a fraction of what its US counterparts have. OpenAI’s development of GPT-4 reportedly cost upwards of $100 million, yet DeepSeek claims it spent just $5.6 million to bring DeepSeek-V3 to life.

How? By using older, less advanced hardware that the US would typically disregard—such as GPUs that are already considered outdated by Western standards. DeepSeek’s ability to get top-tier performance from these resources is a testament to the team’s technical prowess. Despite the lower budget and older hardware, DeepSeek-V3 has surpassed US models like GPT-4 in key benchmarks, such as coding, solving complex mathematical problems, and identifying bugs in software.

Editor’s Picks

1

How a Chinese start-up is changing how AI models are trained and outperforming OpenAI, Meta 2

OpenAI cosying up to Trump, recommends investing heavily in AI if US wants to beat China

This breakthrough has shocked the tech world. For years, it was believed that only massive investments in the latest hardware could produce AI models of this calibre. But DeepSeek has proved that with the right mix of creativity, resourcefulness, and clever design, significant breakthroughs are still possible—even when working with less.

The open-source advantage

What’s also catching the attention of the global AI community is DeepSeek’s commitment to open-source technology. While US companies like OpenAI and Meta have opted for proprietary models, keeping their cutting-edge AI behind paywalls, DeepSeek has made its technology available to the world for free. This approach opens the doors for developers everywhere to build on, tweak, and improve DeepSeek’s models. It’s a bold move that could reshape the future of AI development, promoting collaboration over competition. The open-source nature of DeepSeek-V3 not only democratises access to this technology but also allows for faster iteration and innovation globally.

The decision to go open-source also highlights a stark contrast to the increasingly closed and commercialised nature of AI development in the West. This move could be pivotal in how AI progresses, particularly since DeepSeek’s models are now in the hands of developers worldwide, empowering them to create, adapt, and refine at a speed that’s difficult for larger, more resource-heavy companies to match.

The potential pitfalls

While DeepSeek’s rise has garnered widespread praise, it’s not without its controversies. Concerns are swirling around the potential influence of the Chinese government on the lab’s AI systems. Critics fear that DeepSeek’s AI models could reflect China’s political views and restrictions, especially around sensitive topics such as Taiwan or the Tiananmen Square protests. There is also the larger question of whether AI developed by countries with more authoritarian regimes could uphold democratic values or be used as a tool for control and censorship.

These concerns are compounded by the fact that DeepSeek, despite its open-source approach, may still be subject to government regulations that influence its design and outputs. While the lab’s open-source model has been praised for promoting collaboration, it raises important questions about the kind of values that will be embedded into the AI that is shaping our future.

US’ response: A rushed countermeasure

In light of DeepSeek’s breakthrough, the US has moved quickly to shore up its position in the AI race. The newly announced Stargate Project is a massive $500 billion initiative aimed at creating cutting-edge AI infrastructure in the US. With key players like OpenAI, Oracle, and SoftBank involved, the project’s goal is to ensure the US remains the global leader in AI development. The urgency behind this project signals that the US is not taking the challenge from China lightly.

DeepSeek’s success has forced the US to reconsider its position in the AI race. Until recently, the belief was that only companies with deep pockets and access to the latest hardware could succeed in AI development. But DeepSeek has shown that success in AI is about more than just resources—it’s about innovation, creativity, and making the most of what you have. As the global AI race heats up, it’s clear that the traditional power dynamics are shifting. Whether the US can adapt quickly enough to reclaim its leadership is yet to be seen.