← Back to Home

DeepSeek and Alibaba rescue China’s office landlords

Technology firms are reviving (a few) Chinese commercial-property markets

DeepSeek and Alibaba rescue China’s office landlords

A COURT IN Hangzhou, AI capital of China, ruled in late April that companies cannot fire their staff and replace them with artificial intelligence. This may come as a relief to plenty of people in a metropolis that is home to Alibaba, one of Chinese tech’s mightiest titans, with a total payroll of 128,000 employees. It is also good for Hangzhou’s commercial landlords, who have little use for AI agents and are desperate for human desk jockeys.

China’s dazzling skylines are worryingly lifeless. Since a property bubble burst in 2021, sales, prices and investment in residential and commercial real estate have tumbled. Housing has attracted most of the attention during the crisis; some 90m flats are thought to sit empty across the country. But a decade-long office-building boom has also left cities and ghost business parks littered with unfinished or unoccupied towers.

In some inland cities more than 40% of grade-A offices are empty. The central Chinese city of Wuhan, for instance, may have nearly 200-football-pitches’-worth of idle prime office properties. Commercial projects in which developers invested before the covid-19 pandemic are still coming to the market, further pushing up supply and pulling down rents.

Hangzhou has seen the worst of it. The government’s crackdown on tech that started around the same time as the property-market downturn led to sweeping lay-offs at companies such as Alibaba (whose founder, Jack Ma, was a particular target of the state’s ire). Shenzhen was rocked when its top employers, including Tencent, a digital behemoth, trimmed ranks. A dearth of tech tenants caused new square footage in the southern technology hub to collapse by half in the second quarter of 2022, compared with a year earlier.

In both cities, vacancies at high-end offices subsequently rose from around 20% in 2019 to 30% at their peak, which in Shenzhen was as recently as late 2025. Similarly plush digs in central London were 93% occupied at the end of last year.

China’s property crisis has yet to bottom out. But for prime office space in AI hotspots things are improving. In March demand from AI firms across China was triple what it had been the year before, calculates John Lam at UBS, a bank. This rise was from a low base—the country had many fewer machine-learning startups a year ago. But since their activity is concentrated in a handful of cities like Hangzhou and Shenzhen, it makes a difference. Grade-A office vacancies there are back down to around 20%.

In early 2025, when DeepSeek shocked the world with an AI to rival America’s best, insiders joked that the lab might single-handedly revive Hangzhou’s real-estate market. That looks like an exaggeration: Alibaba, which is reinventing itself as an AI powerhouse, chipped in. And the local government has been rolling out rent subsidies for innovative firms, according to analysts at Savills, a property consultancy.

Tech companies are expected to fill many of the new offices at Westlake 66, an office-and-mall project in the centre of Hangzhou that opened in late April. The rapid expansion of AI firms is also fuelling demand for space from accountants, consultants, lawyers and other corporate hangers-on, notes Derek Pang of Hang Lung Properties, which built Westlake 66. The government has started constraining the supply of commercial real estate and selling less land for development, he adds. This should eventually nudge prices up.

In the short run the court ruling in Hangzhou, which has spared the job of a quality inspector responsible for checking the accuracy of AI output at his company, may help keep the office boomlet going. Whether this turns into a more enduring boom will depend on how many office jobs AI eventually replaces. The countrywide explosion in demand from AI firms in March was due to a need to sit not software engineers but back-office and support staff, Mr Lam points out. This is precisely the sort of job that AI agents are expected to eliminate.