On August 13, Nokia management emailed staff that the Hangzhou R&D center closes by year end. The site works on radio technology and carries about 1,600 people. Offices in Beijing, Chengdu, Qingdao and Shanghai were also described as under discussion.
Five days later the scope changed. On August 18 the South China Morning Post reported, citing people familiar with the matter, that Nokia plans to close almost all mainland China sites in stages by year end and cut most of its 7,200 staff there, keeping some after-sales service.
One site became nearly every site in five days.
The ledger runs back seven years
Nokia’s spokesperson said the company has been aligning its China operations with its global mode of operation, and that the China business has steadily declined over several years. The accounts back that up.
Greater China revenue, which includes Hong Kong and Taiwan, fell from almost €2.2B in 2018 to €913M in 2025. That is 58% off in seven years. Headcount over the same stretch went from an average of 13,700 in 2020 to 7,200 in 2025, close to half gone.
Nokia’s global headcount fell from about 92,000 to 78,000 across those same years, down 15%. China fell 47%. The three-fold gap is the tell. This is not a worldwide contraction landing proportionally in China. China was pulled out and handled separately.
Ericsson’s curve is nearly identical: China revenue slid from SEK 18.7B in 2020 to roughly SEK 8.2B last year. When both European vendors retreat to the same position in the same market at the same time, the cause is not execution. Nokia executives said at an event in Oulu last year that the company had been notified of its exclusion from China on national security grounds. Its share there now runs under 3%.
The money did not leave R&D. It left China.
This is the part that reads backward if you skim it.
Nokia raised full-year restructuring guidance from €250M to €800M, with €350M attached to the China overhaul, targeting roughly €200M of savings from folding Nokia Shanghai Bell into the global organization. String those numbers together and the obvious conclusion is a company cutting research.
The accounts say otherwise. Nokia spent almost €4.9B on R&D in 2025, up from €4.5B the year before. First-half 2026 R&D came to about €2.3B, a 6% increase on the year-earlier half.
So a 1,600-person research center closes while the research budget climbs. Nothing is being withdrawn from R&D. It is being withdrawn from China. For the engineer sitting in that building, the two are not remotely the same event: one says the discipline needs fewer people, the other says the discipline hires them somewhere else now.
1,600 radio engineers land in Hangzhou
The site is in Hangzhou and the work is radio. Put those two facts together and you get the actual labor question in this story.
Hangzhou is now one of the densest clusters of embodied-AI and robotics companies in China. Unitree is headquartered there. It filed its listing notice on August 18 with roughly 18,000 bipedal humanoids cumulatively built and an issue multiple of 219.23x. The skills those companies hire for and the skills a radio R&D center trains overlap too cleanly to be coincidence: RF, embedded, real-time control, signal processing, antenna and EMC work. All of it feeds directly into motion control and multi-sensor fusion.
Which means these 1,600 are not a standard displacement cohort. They are senior on average, trained in large-company engineering process, and they sit squarely on the skill band their own city is shortest on. All of them hit the market in the same quarter.
The near-term effect is on the wage anchor. Release 1,600 comparable candidates at once and every local hiring manager’s negotiating position improves, most of all at robotics firms that were spending six months poaching from legacy telecom vendors one engineer at a time.
The medium-term signal is where they go. If most of this cohort ends up at local robotics and embodied-AI companies, the closure registers statistically as a cross-industry skills transfer rather than job destruction. If they scatter into automotive and consumer electronics, or leave Hangzhou entirely, that says local robotics firms cannot yet absorb senior engineering talent at this scale. For an industry already routing more than 70% of its shipments into industrial use, the hiring side would be lagging the shipping side.
This one does not belong in the AI-layoffs column
Worth stating plainly: Nokia did not cite AI. Not once. That separates it from most of what we have logged recently.
Pentera cut 20% in four months and named AI in both statements. Rapid7 raised its profit outlook and cut 12% in the same filing. Cisco cut 4,000 and three months later named the metric: earnings per employee. Same category, all three: a company explaining structural headcount reduction through AI.
Nokia is the other category. The 1,600 R&D jobs disappear because of market access, not productivity. Filing this under AI layoffs would overstate AI’s footprint and understate what the event means for the people in it. Jobs cut by AI usually still exist somewhere else in the same industry. Jobs cut by a geopolitical exit go away across the whole sector in that country at once.
The practical read for anyone tracking their own exposure is one line: before deciding what a layoff means for you, work out which kind it is. Same 1,600 people, but “this role needs fewer of us” and “this role’s employers left the country” run down completely different paths.
