(March 26): DeepSeek did more than just show the AI industry you don’t have to spend billions to build artificial intelligence. It fired up a long-dormant Chinese tech industry — and now Western names from OpenAI Inc to Nvidia Corp may pay the price.

Since DeepSeek upstaged OpenAI in January with a powerful model that purportedly cost just several million dollars to build, China’s tech leaders have flooded the market with a rapid succession of low-cost AI services, undercutting premium offerings from the likes of OpenAI and Alphabet Inc’s Google. Chinese companies have in the past two weeks rolled out no fewer than 10 major product updates or releases — and that’s just the big names.  

Baidu Inc unfurled the Ernie X1 in direct competition with DeepSeek’s R1. Alibaba Group Holding Ltd followed suit with its own AI agents and reasoning model upgrade. Just in the past week, Tencent Holdings Ltd trotted out its AI blueprint and answer to the R1; Ant Group Co shared findings on how Chinese chips can slash costs by a fifth; DeepSeek itself upgraded the V3 model. Even Meituan — best known as the world’s biggest meal-delivery service — announced it was splashing out billions of yuan on AI.

The rapidly accelerating array of enhancements and tune-ups is more than Chinese companies jumping on the DeepSeek bandwagon. Collectively, the AI models — nearly all of them open-sourced — represent the developers’ effort to set world standards and benchmarks, and grab a bigger slice of the global market. While the jury is out on whether these AI releases match or surpass the most cutting-edge systems from Western AI developers, these newer options are putting more pressure on the business models of leading US companies.

OpenAI, for one, is now trying to strike a careful balance. The ChatGPT maker has said it’s mulling giving away some of its technology following DeepSeek’s success with the open-source approach. At the same time, OpenAI is still weighing charging far more for its most sophisticated products. If DeepSeek’s low-cost template is replicated, that may also shrink the profits for Nvidia which specializes in the expensive AI chips, making valuation “adjustments” inevitable, said Amr Awadallah, founder and chief executive officer of Vectara Inc.

Chinese companies have in past years squeezed out global rivals in industries as varied as electric vehicles and solar panels, by out-manufacturing and undercutting their competitors. The pattern may be replicating itself in AI.

It’s a “big problem,” said the Palo Alto-based Awadallah, whose startup helps enterprises build and deploy AI agents and AI Assistants. “We’re about to see a trend towards significant margin compression for companies throughout the ecosystem. Not just AI model builders, but the large AI enablers that are also driving the growth of the industry.” 

The open-source, high-performance, resource-efficient models that followed DeepSeek’s January announcement are being replicated and used globally, including in the US and India — even as businesses and government officials in these countries move to restrict access to DeepSeek itself on employee devices. 

Chinese developers are reshaping the market and calling into question the massive infrastructure investments pledged by big US tech companies such as OpenAI and Microsoft Corp. Global names including Apple Inc and BMW AG are already teaming with Alibaba to integrate its AI services into their products for the country.

“If you assume the Chinese LLM players’ intent is to disrupt the market and steal share, it’s worked,” said James Wilton, managing partner and founder of Monevate, a consultancy for tech companies. “They might be willing to eat up the costs right now, but it won’t stay free forever.”

At the same time, the Chinese cloud providers that host AI development are slashing prices — a competition that threatens to spill beyond its borders. 

“It’s just sort of a natural evolution of that price war within the Chinese ecosystem spreading out into other markets,” said Kevin Xu, a tech investor and founder of US-based Interconnected Capital.

On Tuesday, Alibaba chairman Joe Tsai warned of a potential bubble forming in data centre construction, arguing that the pace of that buildout may outstrip initial demand for AI services. “I’m still astounded by the type of numbers that’s being thrown around in the US about investing into AI,” Tsai said at a Hong Kong conference. “People are talking, literally talking, about US$500 billion [RM2.22 trillion], several 100 billion [US] dollars. I don’t think that’s entirely necessary.”

Some expect the parade of open-source models to spill over into adjacent fields from computer vision to robotics and image generation in coming months. 

With China’s strength in hardware, the cheaper and more accessible the AI models, the more the demand for devices powered by AI, said Balaji Srinivasan, tech investor and former general partner with Andreessen Horowitz. 

“China is trying to do to AI what they always do: study, copy, optimise, and then bankrupt everyone with low prices and enormous scale,” he said in a post on X.

Uploaded by Felyx Teoh