One of the most striking AI recruitment figures in China this week is not attached to model research. It is attached to deployment: an advertised ByteDance forward-deployed engineer role whose upper monthly range, multiplied by its reported 15-month pay structure, reaches ¥975,000 annually.
The South China Morning Post reported on October 25 that ByteDance, Tencent Holdings and Alibaba Group are expanding recruitment and training for forward-deployed engineers, the staff who sit with a customer and turn a model into a working system inside that customer's business. The ByteDance listings, on the Boss Zhipin recruitment platform, put the monthly pay at 35,000 to 65,000 yuan over a 15-month pay structure. That top-of-range multiplication is the number now travelling around the internet: 975,000 yuan a year.
Do the arithmetic before quoting it
That number deserves the treatment this site usually gives a benchmark chart. It is the ceiling of an advertised range, reached only by a candidate who lands the maximum monthly salary and the full 15-month structure. Based on the monthly range and 15-month pay structure reported by the South China Morning Post, ALFW calculates an advertised annual range of ¥525,000–¥975,000. Neither endpoint establishes what successful applicants actually receive. At the exchange-rate approximation used in the reporting, the ceiling is roughly US$146,000; the yuan figure is the primary one.
The other employers' figures are monthly only. Ant Digital Technologies, Ant Group's technology arm, lists business-facing roles at 40,000 to 60,000 yuan a month. Tencent-related positions run from 30,000 to 60,000 yuan a month, and Alibaba Cloud's deployment engineering listings reach 50,000 yuan a month. The 15-month structure is reported for ByteDance's roles; the SCMP reporting gives no annual pay structure for the other employers, and their monthly ranges should not be annualised using ByteDance's.
The calculated ¥975,000 ceiling is about 22.5 times China's 2025 nationwide per-capita disposable income of ¥43,377, according to ALFW's calculation using official statistics. This is a scale comparison, not a like-for-like salary comparison.
George Wang, a senior human resources professional at an AI company in Beijing, made the same point to the SCMP directly: the high salaries advertised for some of these jobs account for only a small part of the market. He still described strong demand for developers able to build AI agents as companies reorganise around the technology. That pairing, a hot ceiling and a thinner middle, is what a young job category looks like before its pay settles. The advertised maximum should not be mistaken for a typical FDE salary; the available reporting does not establish a market median.
The SCMP is owned by Alibaba, one of the companies discussed in its reporting. That relationship is relevant context for readers, though it does not by itself discredit the recruitment figures. The cited job listings originate on Boss Zhipin, and the posting-growth statistic comes from Maimai; ALFW has not independently audited either dataset.
The surge counts adverts, not people
According to Maimai data cited by the SCMP, FDE job postings rose 1,523% in the first seven months of 2026 compared with the same period in 2025. The measure counts job advertisements, not hires. The available reporting does not provide the underlying posting counts, unique-employer totals or number of positions filled.
A posting is an intention, not a hire. Companies may advertise the same role in several cities, repost unfilled jobs, or advertise partly to signal strategy to investors and rivals — possible explanations for part of the surge, none of them measured by the reporting. Read with those limits, the 1,523 percent is best understood as a measure of how fast the job title itself is spreading, which is still information: two years ago, most of these companies did not have the title at scale.
The work itself explains the spread. An engineer identified as Su, who previously led a forward-deployed team at a major Chinese internet company, described the hiring target to the SCMP as someone who can understand clients across different industries, choose appropriate AI tools and carry a solution through to implementation. That is a consulting skill set attached to an engineering salary, and it cannot be produced by a training run, only by sending people into messy organisations and letting them accumulate cases.
Governments and clouds are formalising the role
The SCMP also reports that Tencent Cloud introduced an FDE certification for engineers working with its Agent Development Platform, describing it as the industry's first for the role. Separately, a late-August MIIT notice encouraged customer-site deployment teams and relevant university training. These are institutional signals of interest in the role, not evidence of how many engineers have been hired.
The pattern is not China-only, though the comparison is context rather than corroboration. Earlier this month Anthropic announced a US$100 million initiative to train 10,000 forward-deployed engineers by the end of 2027, a program this site covered on October 3. Two different systems, one state-guided and one corporate, have arrived at the same diagnosis in the same month: AI adoption depends not only on model capability but on people who can wire models into real operations.
That is the durable part of this story, and it survives every discount applied above. Discount the ceiling, discount the posting count, note the newspaper's owner, and what remains is a ministry notice, a certification program, a nine-figure American training fund, and three Chinese platforms competing for the same hybrid role. The listings and institutional initiatives suggest growing organizational attention to deploying AI systems in customer operations. Readers deciding where AI careers or AI budgets go next should watch filled roles and delivered deployments rather than advertised ceilings. But they should not dismiss the signal because the largest number in it is a maximum: maximums are where markets announce what they value most.
Sources
- [1] South China Morning Post (October 25, 2026) — the original report on forward-deployed engineer pay at ByteDance, Tencent and Alibaba, based on Boss Zhipin listings and Maimai data; read by ALFW via the full relay at AI Affairs, with every figure cross-checked against KuCoin's relay of the same reportRead source
- [2] KuCoin News flash relay of the SCMP report (October 25, 2026) — carries the US$149,739 conversion for 1 million yuan behind the approximate dollar figure, and notes that Alibaba owns the South China Morning PostRead source
- [3] National Bureau of Statistics, “Households' Income and Consumption Expenditure in 2025” — nationwide per capita disposable income of 43,377 yuan in 2025, read directly by ALFWRead source
- [4] Anthropic's US$100 million forward-deployed engineer initiative, as covered by alfw.ai on October 3, 2026 — context only, not a source for the China figuresRead source