On Monday the tenth of June, engineers at Shopee’s Singapore headquarters were told they were done. The cuts ran globally, about 8 percent of the developer workforce, and they landed on engineering and quality-assurance roles. Sea, Shopee’s parent, did not reach for the usual euphemisms about right-sizing or realignment. It named the reason plainly: a pivot to AI. That candour is the interesting part. Companies rarely say the quiet thing out loud, and a company employing tens of thousands of people across the region just did.

Sea is not a marginal case. It runs Shopee, the Garena games business, and the Monee fintech arm, and at the end of 2025 it carried just over 77,000 staff. Chief Executive Forrest Li has said openly that a one-trillion-dollar market capitalisation, roughly ten times where the company sits now, is reachable if Sea pushes AI through all three businesses, with Google supplying the model layer for search, advertising, and an agentic-commerce prototype. Cutting developers in the same breath as chasing a trillion dollars looks like a contradiction. It is not.

The old roster was sized for a world where each new feature cost a certain amount of human labour. That cost fell, so the roster stopped balancing.

Here is the thing the panic about machines coming for jobs keeps missing. The number that moved at Sea was the cost of producing the next unit of product work, well before it showed up as a headcount. When a model can draft, refactor, and test a service well enough that one engineer now ships what three used to, the price of the next feature drops. Headcount math is downstream of that price. The old roster was sized for a world where each new feature cost a certain amount of human labour. That cost fell, so the roster stopped balancing.

Call it bookkeeping that already happened rather than a forecast about what might.

You can see the same arithmetic working in China, which gives the move its real scale. Alibaba shed about a third of its headcount across 2025, falling from a peak near 250,000 to roughly 124,000, while pouring resources into its Qwen models and cloud. Baidu trimmed close to 7 percent. The pattern is selective: cut the broad engineering and operations base, hire hard into AI and research. Tencent, notably, grew about 5 percent over the same year. The work did not vanish. It moved to where a smaller number of people, leaning on better tools, could carry it.

I would resist the cleanest version of this story, though, because the evidence does not support it cleanly. Reporting on the Chinese cuts found the AI-displacement effect to be fragmented, not systemic, with economic slowdown and post-expansion correction doing much of the work that gets credited to automation. AI is the accelerant and the stated reason, not always the sole cause. The honest read is that the cost of the next feature fell far enough to change the headcount math, and a soft macro environment made the cut easy to take.

The reliability gap is exactly why the cuts hit quality assurance and not the senior architects.

The detail that should hold your attention is which roles went. The cuts hit quality assurance hardest, and that is not an accident. A model is strongest at generating plausible code and weakest at knowing whether that code is actually correct under load. The reliability gap is exactly why the cuts hit quality assurance and not the senior architects. Treating QA as the most automatable function is a bet that the gap between plausible and correct has closed. For a checkout flow moving real money across six markets, that is a live and uncomfortable bet.1

For an engineer in the region, the lesson is narrower and harder than “AI took my job.” The work that priced itself by volume, more tickets closed, more tests written, more boilerplate produced, is the work whose price just fell. What did not fall is the price of judgement about what to build and whether it holds. That is the column the region’s biggest tech employer is still buying. On Monday in Singapore, it stopped buying the other one.

Footnotes

  1. Sea said more cuts may follow. Worth watching whether the next round reaches the people deciding what to build, or stays in the rooms where work is merely produced. ↩