ChinAI #314: Can AI save China's independent cloud providers?
Greetings from a world where…
scotcheroos are too good
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: Can AI save China's independent cloud providers?
Context: This week’s feature translation (link to original Leiphone article) is centered on a question that, at first glance, doesn’t seem to have obviously high stakes: Can AI save China’s independent cloud providers?
Why should we care if the big giants — Alibaba Cloud, Huawei Cloud, Tencent Cloud, and Baidu Cloud — continue to dominate China’s cloud market?1 Well, for one, cloud providers that are independent (not subsidiaries of big tech giant) such as QingCloud and UCloud check against concentration in the cloud computing market. In my book, I cite a lot of sources that show that more competitive market structures create healthier environments for technological diffusion across firms.
Here’s one example, taken from the workforce-development programs of the U.S.’s Big Three cloud providers (AWS, Microsoft Azure, Google Cloud), that shows how concentration can distort the incentives for widespread diffusion:
“As with infrastructure expansions, these workforce-development programs can benefit both the cloud provider and the participants in the program, though they may have negative effects on competition in the cloud market. Because companies tailor these workforce-development programs to their specific suites of cloud and AI services, these initiatives can serve to lock participants into a particular cloud provider’s ecosystem. These programs are similar to the courses the Big Three offer related to their platforms, such as the AWS Certified Developer course, which builds competencies not in cloud computing as a whole but rather AWS’ specific offerings.”
Key Takeaways: With the rise of generative AI, independent cloud vendors such as UCloud [优刻得], QingCloud [青云科技], and Kingsoft Cloud have found a new lifeline.
Let’s start with the data. The estimated proportion of AI-related revenues from UCloud’s Q1 2025 financial report: more than 30%. The comparable metric for QingCloud for 2024: about 40% of QingCloud’s annual revenue. Same indicator for Kingsoft Cloud’s 2024 Q4: 34%. CCID Consulting’s cloud service market research report listed these three as “challengers” to China’s dominant cloud computing players.
Here’s the hope for medium and small-sized cloud providers: “Unlike the general computing era where big companies win all and the customers become giants, the decentralized trend of customers in the intelligent computing business leaves room for independent cloud vendors to play. After all, when AI is implemented, it is difficult for big companies to reach all sub-industries. The emergence of open-source DeepSeek has also broken the model moat in competition with the tech giants.”
However, there are three major challenges. Most interesting to me, the last one is all about policy changes, including China’s recent “window guidance” on computing power.
First, smaller cloud vendors struggle to compete in price wars in computing power. According to industry insiders, this year’s prices for computing power rental — a key line of business for these cloud providers — have dropped by about 30-40 percent compared with May and June last year. One source at an independent cloud provider disclosed that 80-90% of its customers do one-time purchases, instead of a subscription-based system.
Here’s how it often works when these small and medium-sized providers go against large cloud giants like Alibaba Cloud: “When an independent cloud manufacturer provides intelligent computing platform services to an insurance company, the list price given is lower than the price of a large manufacturer, but the large firm privately tells the company that it will sell at a 50% discount, that is, if it wants to win this order, the price of the independent cloud provider must be reduced to less than 50%.”
Challenge #2 is the long-tail customer base (large base of smaller customers who make low-volume purchases) bring substantial uncertainty. The article states, “long-tail customers account for 80% of these inference demands, which often bring scattered orders, in units of one month or three months, at most half a year or one year. In the computing power leasing business, the market can no longer have a long-term stable sense of security.”
Notably, large model startups have reduced their consumption of computing power. They used to be big clients (“A large model company can talk about dozens of projects a year, and the contract signing usually starts with at least a three-year period”), but many could not close the loop on their business models. The piece cites a fascinating data point: as of October 2024, almost half of the large models that passed the algorithm registry filing have moved away from training foundation models toward developing specialized AI applications (with lower compute power needs).
Independent cloud providers confront a third major challenge connected to government policy fluctuations. Previously, the central and local governments encouraged an all-out blitz to build data centers, which resulted in a lot of reckless and duplicative construction. In a reversal, in March and April this year, the Chinese government seems to have issued a “window guidance” for computing power infrastructure consumption that restricts state-owned enterprises from investing in small-scale intelligent computing projects, with energy consumption under 2500 KW. *See ChinAI links for more background on this informal industry policy instrument.
As the article concludes, “This is not a good thing for independent cloud vendors: many intelligent computing projects are assets, liabilities or advance funds held by state-owned entities.”
Some other consequences from tightening at the policy level: “A person in the computing power industry who has close ties with state-owned enterprises revealed that the subsidy for investing 1 billion (RMB) to build intelligent computing centers in 2023 can reach 40%. Since 2024, due to the continuous adjustment of approval standards, approval has become more difficult.”
FULL TRANSLATION: Can intelligent computing business save “independent cloud vendors”?
ChinAI Links (Four to Forward)
Should-read: Green monetary policy in China - window guidance and the promotion of sustainable lending and investment
An intriguing detail from this week’s feature translation was a reference to a window guidance document on datacenter construction. This Climate Policy article looks at China’s use of window guidance as a monetary policy tool for green finance.
Should-read: Funding the AI Cloud — Amazon, Alphabet, and Microsoft’s Cloud Computing Investments, Part 1
Here’s the CSET data snapshot, published by Christian Schoeberl and Jack Corrigan, that I cited above. Some great statistics on the Big Three’s global investments in cloud computing infrastructure.
Should-read: After Rubio seeks to revoke their visas, Chinese students say U.S. resembles the country they left
For NBC News, Kimmy Yam reports on the State Departments new policy that will aggressively revoke visas of Chinese students “with connections to the Chinese Communist Party or studying in critical fields”. I’ll just say this: twenty some years ago, my parents came to Iowa City to study biochemistry and computer science. If this policy were in place back then, our family’s opportunities look a lot different.
Should-read: Fading Beacon
For The Chronicle of Higher Education, Karin Fischer and Sasha Aslanian’s 2021 report on the consequences of the U.S. closing its door on international students is especially relevant today.
Thank you for reading and engaging.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Also! Listen to narrations of the ChinAI Newsletter in podcast format here.
These four providers captured 80 percent of China’s cloud computing market in 2021. https://www.china-briefing.com/doing-business-guide/china/sector-insights/china-s-cloud-computing-market-developments-and-opportunities-for-foreign-players