As advances in artificial intelligence (AI) gather pace, the superpowers are all too aware of the ramifications of non-aligned partners gaining ground in the technology – particularly in military uses. The US and China have, as a result, engaged in a tit-for-tat to restrict each other’s development potential.
The US has added advanced computing chips to its export control list. China has banned some firms from buying products from the US’s largest memory chipmaker Micron Technology and announced export restrictions on metals that are used in advanced chip manufacturing.
Now, the Biden administration reportedly intends to make US cloud services providers seek government permission before they can provide Chinese companies with cloud services that use AI processors, particularly where the platforms are being used to train AI models. The move would be a major escalation in the technological trade war between China and the US, but also a step towards the politicisation of cloud provision.
“Cloud computing was political from the beginning,” explains Trey Herr, director of cyber statecraft initiative under the Atlantic Council’s Digital Forensic Research Lab, “and it will be political at the end.”
“It requires a tremendous amount of physical infrastructure and all of that infrastructure has to sit somewhere. So there are the local politics of the jurisdiction it lives in,” he says, adding that any discussion of cloud security quickly becomes politicised.
How cloud restrictions will affect big business
Caught in the middle are the cloud providers doing business with Chinese clients. Of those, Microsoft Azure and Amazon Web Services (AWS) are the biggest and likely most affected.
“It certainly has an impact on earnings. I think everyone is doing business in the Chinese market,” says Herr.
The new AI-related restrictions mean providers will need to change the services available to Chinese consumers. But mitigation against losing that segment of the market has already been under way. “It’s a growing market, not an existing one,” says Gordon McKenna, CTO of public cloud at Ensono. In that sense, cloud providers aren’t losing as much as they might gain, as the need for machine learning data ramps up and companies know they must adopt AI or risk getting left behind.
It’s worth maintaining perspective, comments Dr Fabio Goncalves de Oliveira, a lecturer in entrepreneurship and innovation at Henley Business School and a certified instructor of AI and cloud computing. He notes that Chinese regulations for multinational companies to provide cloud and telecoms services in China dictate there must be at least 50% local ownership of capital.
For example, Microsoft Azure is represented in China by Shanghai Blue Cloud Technology and AWS is represented by Beijing Sinnet Technology Co (Sinnet). It still needs to be determined whether any new rules would impact these ventures. Moreover, the Western stake in Chinese public and private cloud is relatively small. Research by IDC shows that Alibaba has 34% of the market share (the biggest), followed by Huawei, Tencent and China Telecom, all Chinese-owned. By comparison, AWS carries only 9% of the market share.
“For cloud AI developers based in China, this could be a problem if the current Chinese cloud providers do not keep up with the pace of development of AI models,” says Goncalves de Oliveira.
But this outcome seems unlikely. As Goncalves de Oliveira points out: “The Chinese regulations for using data, including data about citizens, benefit AI development, while EU and Western regulations constrain the advances of new AI developers and the use of data to train AI models.”
How cloud providers are responding to customers’ needs
In some ways, the US plans are unsurprising as companies and investors have been weaning themselves off Chinese data, devices and equipment, with many embargoes from Hikvision cameras to Huawei and TikTok.
Cloud computing was political from the beginning, and it will be political at the end
There is a general concern among cloud providers’ clients about security, and in particular that they might be sharing a data centre with Chinese companies. This is not a place that either cloud providers or their clients want to be in, says McKenna.
“I think Microsoft and AWS have understood better these requirements from clients and have architected solutions better to meet these compliances,” he adds.
As an example, AWS has moved beyond the traditional public and private cloud platforms. Clients in highly regulated industries can keep their data in GovCloud platforms, which are more secure and designed to meet the security and compliance requirements of government agencies and organisations.
The rules in question may also apply only to the public cloud, says Goncalves de Oliviera, and multinationals operating in China typically adopt a hybrid approach: “If they have a private cloud and an authorised network carrier, these companies might still gain access to restricted services.”
Will there be further combative restrictions?
That said, given the complexities of regulating such a rapidly advancing market, we are only seeing the beginning of governance in this area.
“Sources in Chinese and American big tech have stated that it is impossible for the American government to avoid the spread of cloud AI technologies in both directions,” says Goncalves de Oliviera. “Chinese AI developers can access pre-trained Western models elsewhere, through their presence in American, British and other Western universities and participation in Western corporations.” For example, the open-source, pre-trained machine learning and deep-learning models offered by the California-based company H2O.ai are free.
Cloud will continue to act as the extension of trade wars and geopolitical wrangling. Cloud providers will be able to pre-empt many future restrictions, and those with larger stakes should have little to worry about in the short term.
