HomeIndustriesChinese startups like DeepSeek are difficult global AI giants

Chinese startups like DeepSeek are difficult global AI giants

Stay up so far with free updates

Last month, a Chinese startup called DeepSeek surprised the international tech community with its latest open-source artificial intelligence model. DeepSeek-V3 delivers performance comparable to better-funded US competitors resembling OpenAI. This week it impressed again with R1, its foray AI reasoning.

DeepSeek will not be an isolated case. Since the center of last 12 months, Chinese technology firms resembling Alibaba, Tencent, ByteDance, Moonshot and 01.ai have steadily narrowed the gap with US rivals, matching their capabilities and surpassing them in cost efficiency.

China's efficiency successes are not any coincidence. They are a direct response to the escalating export restrictions by the USA and its allies. By restricting China's access to advanced AI chips, the US has inadvertently spurred its innovation.

To reduce reliance on high-end chips from overseas, Chinese AI firms have been experimenting with novel approaches to algorithms, architecture and training strategies. Many have taken an “expert mix” approach, specializing in smaller AI models trained on specific data. These can deliver powerful results while reducing computing resources.

DeepSeek-V3 embodies the success of this imaginative approach. According to the technical report, the model was trained in a knowledge center using Nvidia H800 GPUs – a less advanced chip than Nvidia's latest versions. Despite this, DeepSeek accomplished the training in only two months and value $5.5 million – a fraction of the sums reportedly spent by US firms resembling OpenAI.

DeepSeek has also dramatically reduced inference costs, earning it the nickname “Pinduoduo of AI,” a nod to the favored Chinese discount e-commerce giant’s cost-cutting business model. This breakthrough has profound implications. It challenges the widely held assumption that cutting-edge AI requires enormous computing power and lots of billions of dollars. DeepSeek shows how software ingenuity can compensate for hardware limitations.

It also highlights the bounds of US export controls designed to slow China's AI advances. While these measures may cause short-term disruption, their impact will diminish over time as China innovates to adapt.

The inconvenient truth for U.S. policymakers is that strict export controls have forced Chinese tech firms to grow to be more self-reliant, resulting in breakthroughs that may not otherwise have been possible.

However, not everyone agrees with this. Gregory Allen, director of the Wadhwani AI Center on the Center for Strategic and International Studies think tank, claims China has made progress no matter US restrictions.

To make sure, the Chinese government has long sought to make the country an AI superpower. But the record in resource allocation is poor. Government-led initiatives often result in inefficiency, corruption and waste – a reality highlighted by high-profile scandals and slow progress within the country's semiconductor industry.

What America's AI restrictions inadvertently did was provide a robust business incentive for the Chinese private sector to step into the gap. And while the US often dominates in groundbreaking research, Chinese firms excel in implementation, affordability and product distribution. They have demonstrated a formidable ability to scale in e-commerce, electric vehicles, solar panels and batteries.

The rise of DeepSeek is a stark reminder that limitations can drive innovation. For Donald Trump's latest administration, this raises an uncomfortable query: Will increasingly stringent sanctions against China speed up the very advances they’re designed to suppress?

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Must Read