All news

China's AI plan makes the state the industry's anchor customer

Long symmetrical rows of dark server cabinets receding down the aisle of an empty data centre hall

A trillion-yuan target for 2030 turns China's government into the main buyer of AI, changing who grows fastest and how rivals abroad must compete.

One trillion yuan, nearly $150 billion, is the size China's State Council has set for the country's core artificial intelligence industry by 2030, with related industries ten times larger. The New Generation AI Development Plan, dated 8 July and published on 20 July, sets interim marks of 150 billion yuan by 2020 and 400 billion by 2025, and aims to make China the world's primary center of AI innovation by 2030.

The figures are targets, not a budget, and that is what matters for capital. Chinese industrial plans set outputs and let ministries, provinces and cities compete to deliver them with tax breaks, investment funds and, above all, demand. This one makes the state the anchor customer of Chinese AI. Its list of applications runs through smart government services, smart courts, smart cities, traffic control and intelligent security and police products. Companies that sell to that customer will grow faster and raise money more cheaply than their revenue alone would justify.

Private money moved first. On 11 July SenseTime, a Beijing computer vision company whose clients range from banks and phone makers to city security, closed a $410 million Series B, a record for a single round in AI. The plan now tells every local government that backing such companies is policy: it directs existing government investment funds to qualifying AI projects, offers tax incentives to AI start-ups and promises pilot programs that open public data for their use. In vision systems for security and cities, the investor and the customer now pull in the same direction.

The contrast with the United States lies in form more than in scale. Washington spent about $1.1 billion on unclassified AI research in 2015, and the strategy it published last October sets research priorities, not industry sizes or deadlines. American AI companies earn their revenue from enterprises and consumers who renew only when the product works. Their Chinese peers will increasingly earn theirs from procurement budgets, which grow on political schedules rather than on product results.

That difference sets the terms of competition for founders outside China. A rival with guaranteed domestic demand can price its exports low in vision, speech and city systems, so Western start-ups in those fields need products whose buyer is a private company paying for results, not a government paying for scale. For Fanaura, which invests in the US, the EU and Latin America, the plan separates markets where competition will be state-backed from those where it will not, and the second group is where independent AI companies keep their margins.

Chinese targets have often produced subsidies and overcapacity rather than durable companies, and local funds chasing this goal will inflate valuations of firms with little revenue. That is the best case against this reading. The first checkpoint is close. By 2020 the plan wants 150 billion yuan of core industry, roughly $22 billion, and the mix of who pays for it will show whether the plan built markets or only customers.

Chinese AI now has a buyer that does not need a return on its spending. Anyone competing with it should price that in.

For founders

Building AI products that private companies pay for by results? Send us the project.

Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.