Huawei plans more AI pharma partnerships, mainly with Chinese drugmakers

Huawei intends to sign more AI partnerships with pharmaceutical companies, according to William Zhang, president of its healthcare business unit, who told Reuters the projects currently run mainly with domestic Chinese drugmakers.

The company sells compound screening tools alongside its Ascend and Kunpeng chips, the same domestic stack it has been building out since US sanctions forced it to find workarounds.

“As we further deepen our research into AI in the medical field, we’ll have more collaboration and results emerging with pharmaceutical companies from drug manufacturing to clinical to final implementation,” Zhang said.

He also described existing collaborations in clinical practice in hospitals without elaborating on which ones.

There is no target number attached to any of this, and no prospective partner was named. It is a statement of direction rather than an announcement, which is worth establishing before the figures that will inevitably get attached to it start circulating.

The one project Huawei has pointed to is a deal announced in May involving Guangzhou Pharmaceutical Holdings, the state-owned group.

Huawei described it as the industry’s first production validation of independently developed AI drug research models adapted to Ascend and Kunpeng.

That phrase deserves unpacking, because it is likely to be misread. “Independently developed” here translates a Chinese term meaning domestically developed rather than built by Huawei, and the models in question belong to StoneWise, a Beijing AI drug design company that was the third party to the agreement.

Huawei’s contribution was the silicon and the work of porting somebody else’s software onto it. That is a meaningful thing to have done, given the whole point is running a drug discovery pipeline without American chips, but it is infrastructure rather than science.

The company does have a model of its own, and it is not new. The Pangu drug molecule model was released in 2021, developed with the Chinese Academy of Sciences and trained on 1.7 billion existing compounds to predict how molecules bind to targets.

Huawei also has an older tie-up with Yunnan Baiyao dating to 2022, under which the drugmaker supplies botanical compound libraries, and Huawei supplies the cloud and the AI. Neither that nor the Guangyao work has produced a named drug candidate that has entered trials.

The scale gap with the competition is the part that puts the announcement in perspective. Nvidia has struck AI partnerships with Eli Lilly and Novo Nordisk, and the Lilly arrangement alone is a co-innovation lab with up to $1bn committed jointly over five years.

Set against that, Huawei’s disclosed pharma work is one three-party ecosystem agreement at proof-of-concept stage plus a four-year-old cooperation deal. The asymmetry is not hidden, and Zhang did not attempt to hide it.

What Huawei has instead is a policy tailwind. Biopharmaceuticals were named an emerging pillar industry in this year’s government work report, AI in pharmaceuticals is written into the fifteenth five-year plan, and a state-backed body inaugurated in June lists Huawei on the supply side alongside Kingdee and XtalPi.

The domestic focus Zhang described also has an obvious constraint behind it. US export guidance issued last year told the world that using Huawei’s Ascend accelerators anywhere is likely to breach American controls, which limits the addressable market for a pitch built on that hardware, much as it has for Huawei’s data centre bids abroad.

The wider field is less encouraging than the announcements suggest. Roughly $60bn has gone into AI drug discovery globally, and no AI-discovered drug has yet been approved anywhere, though 179 candidates were in pipelines by June against four in 2017, and nine have reached Phase III.

Reuters noted that industry forecasts suggest machine learning could halve early-stage development timelines and costs within three to five years.

That is a projection about the field rather than a claim about Huawei, and the distinction matters given how much of this sector is currently being sold on the strength of the former, as China’s push to run serious AI workloads on domestic silicon keeps demonstrating.