Alibaba, Tencent cast wide net for AI upstarts

Since 2023, investors—including the country’s biggest tech companies—have valued at least six China-based startups developing large language models at more than $1 billion each. Most of these unicorns, dubbed China’s six “Little Artificial-Intelligence Dragons," have received capital from Alibaba and Tencent.

While both companies have dramatically reduced their overall investments in the past two years, their interest in AI has intensified. Since 2023, 40% of Alibaba’s deals in China and 30% of Tencent’s have targeted AI startups—both record highs for the two tech titans, according to Beijing-based deal data provider ITJuzi.

Alibaba and Tencent are “positioning themselves as frontrunners in the generative AI space," said Wei Sun, senior analyst at market research firm Counterpoint Research. “Investment in AI startups is a strategic effort to strengthen their technological edge and expand market influence."

Chinese tech giants, once known for their aggressive dealmaking, have shifted focus to making profits in recent years after being hamstrung by Beijing’s regulatory clampdown on the sector, the Covid-19 pandemic and a faltering Chinese economy. Generative AI has emerged as a rare exception to their newfound conservatism.

Alibaba and Tencent have made AI a strategic priority, seeking to cash in on the technology, develop their own foundation AI models, and sell access to both those and AI computing power via cloud services.

At the same time, China’s AI-model arena is getting increasingly crowded, with more than 190 products now on the market, raising fears among some companies that they are falling behind. By investing in AI unicorns, companies hope to get firsthand insight into research and development, plus access models that could bolster their product offerings and business practices, solidifying their edge in the industry.

“The investments of big tech vendors in AI startups are crucial for their long-term strategies, helping them stay competitive and innovative in a rapidly evolving tech landscape," said Charlie Dai, an analyst at market research company Forrester.

Alibaba and Tencent didn’t respond to requests for comment for this article.

Similar to how Microsoft and Amazon backed OpenAI and Anthropic, respectively, both Alibaba and Tencent have offered significant parts of their investments to portfolio companies in the form of computing power.

Alibaba said in May that it recently invested around $800 million for a 36% share in Beijing-based Moonshot AI. More than $300 million of that was provided in the form of credits to use Alibaba’s cloud-computing infrastructure, people familiar with the matter said. These credits, when used, can be booked as revenue for Alibaba, benefiting its financial results, they said.

Alibaba Chief Executive Eddie Wu said in an analyst call last week that the company expects its revenue from external cloud customers to return to double-digit growth in the next one to two quarters, driven by the expanding adoption of AI cloud services.

Providing computing resources can be more valuable than cash in China given that the U.S. has effectively blocked access to cutting-edge semiconductors in a bid to contain Beijing’s advancement in AI capabilities.

Computing tasks typically form the majority of capital expenditure of LLM startups. An Nvidia processor for AI workloads can have the same price tag as a Cadillac sedan, and training LLMs requires thousands of such chips.

Investors provide computing power support in different ways, including by renting out hardware or selling cloud computing at discounted rates and offering funds earmarked for purchasing chips.

Some startups also tap their backers’ connections to obtain restricted chips.

In December, a Beijing-based LLM unicorn bought several hundred of Nvidia’s H100 chips through an affiliate of one of its investors in Singapore, people familiar with the purchase said. Last September, a Shanghai-based startup developing a healthcare-focused LLM rented computing power provided by Nvidia’s A100 chips from a state-run AI data center in the southwestern city of Chongqing after an investor helped negotiate for a below-market rate, according to people familiar with the deal.

People close to AI investments say capital-hungry startups can access funding from tech giants more easily now than a few years ago, when powerful investors often sought exclusive agreements that prevented startups from accepting money from competitors. Beijing has tightened antitrust rules and cracked down on what regulators called tech companies’ “disorderly capital expansion" in recent years.

“Generative AI holds massive business potential, and no tech giant would want to miss the opportunity," Forrester’s Dai said. “Investing in leading startups, even [ones that have already received investment from] competitors, will help tech giants to grasp the opportunities as much as they can."

Write to Raffaele Huang at raffaele.huang@wsj.com and Tracy Qu at tracy.qu@wsj.com