Chinese AI lab DeepSeek sent a shockwave through the tech sector this week after releasing its R1 large language model (LLM) that was faster, more efficient, and cheaper to train and run than existing models, even though it used underpowered A800 chips from Nvidia (NASDAQ:NVDA).

The impact was immediate. Nvidia and Broadcom (NASDAQ:AVGO) plunged 17% and a host of other AI tech stocks also tumbled by double-digit percentages. There was a $1 trillion selloff in U.S. stocks. If DeepSeek R1 can build a better model using inferior chips, then demand for massive clusters of expensive AI GPUs will undoubtedly fall.

Yet right on the heels of the DeepSeek announcement another aftershock hit AI. Alibaba (NYSE:BABA) announced that it’s own LLM, Qwen2.5-Max, can also outperform ChatGPT and Meta Platform‘s (NASDAQ:META) Llama 3.1 models. Moreover, it surpassed DeepSeek on several benchmarks.

This is equally significant to the DeepSeek news. Where the R1 model runs on dumbed-down versions of Nvidia’s popular H100 chips that are used by data centers and hyperscalers, Qwen2.5 also uses graphics processing units (GPUs) that are comparatively underpowered.

This suggests that sleek and cheap AI models are within reach of just about anyone.

The AI market shakeup continues

Alibaba, of course, is primarily seen as China’s e-commerce answer to Amazon (NASDAQ:AMZN). But the online retailer has its fingers in a lot of pies, including cloud services and logistics. Apparently it is now an AI stock too.

While there aren’t a lot of details yet about Qwen2.5-Max, Alibaba posted about the AI model’s achievements on WeChat and in a blog post. 

“Qwen 2.5-Max outperforms…almost across the board GPT-4o, DeepSeek-V3 and Llama-3.1-405B,” it said on the social media site, according to Reuters, while pointing out on its blog, “The scaling of data and model size not only showcases advancements in model intelligence but also reflects our unwavering commitment to pioneering research.”

While there’s been no independent verification of Alibaba’s claims and its model is closed-source compared to DeepSeek’s open-source code, which allows users to modify the algorithm, fine-tune it, and build upon its foundation, it represents another step forward in cheap, accessible AI.

Doing more with less

Alibaba’s development is still noteworthy. Where DeepSeek uses the A800 chip Nvidia developed to get around technology export restrictions to China, Alibaba reportedly uses the AI chip it developed, the Hanguang 800 GPU, which also has less raw computational power than Nvidia’s H100.

What it shows is how rapidly changing the global AI landscape is. Instead of brute compute force, we will soon see sleek, efficient, and cheap models proliferate, making AI more widely available. OpenAI, for example, reportedly uses 32,000 top-of-the-line GPUs for its latest ChatGPT model.

While that means there is risk to the pricing power Nvidia currently commands for its high-end H100 GPUs and the soon-to-be-released Blackwell superchip, it bodes well for making AI more broadly available to more companies. It is also a positive development for Alibaba.

AI powering BABA’s resurgence

The Chinese conglomerate demonstrated its ability to swiftly respond to breaking developments, all the while adapting to limitations on accessing the latest technology. It shows that more can be done with less by using resources more efficiently and innovating existing architectures.

As Alibaba said in its blog, “We are dedicated to enhancing the thinking and reasoning capabilities of large language models through the innovative application of scaled reinforcement learning.”

Coupled with the return to growth of its legacy businesses, BABA stock is positioned to rise. Third-quarter revenue increased 5% to $77 billion as it improved upon its monetization of Taobao and Tmall, while cloud services raced ahead 7% to $4.2 billion on the strength of its AI offerings. It’s notable that the Hanguang 800 chip was developed by Alibaba to specifically work within its own cloud and e-commerce ecosystem.

Alibaba’s AI-related product revenue grew at triple-digits year-over-year for the fifth consecutive quarter.

Key takeaway

Shares of BABA just broke through the $100 per share threshold for the first time since November and are up 38% over the past year. Trading at just 10 times next year’s earnings, less than twice its sales, and 16 times the free cash flow it produces, Alibaba looks like it’s ready for a magic carpet ride higher.

Thank you for reading! Have some feedback for us?
Contact the 24/7 Wall St. editorial team.