Invest1 distinct publisher2 min readUpdated
Equity plans at Chinese chip and AI companies now cover most of the payroll rather than a handful of stars. That changes what a counteroffer costs anyone hiring against them.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
The per-head sums are large because the payrolls are not. Cambricon's earlier plan put 5 million shares in front of 944 people, 85.3% of the company [3], which implies a total headcount of about 1,107 [2]. Moore Threads added 1,080 staff, 85% of its people, in April [6], implying roughly 1,271 in all [3]. Work backwards from the smaller grant and Cambricon's stock was being valued at about 1,151 yuan a share on the day [1]; at that level the 5-million-share plan carries a value near 5.8 billion yuan, about 5.2 million yuan a head across the entire company [4]. For an employer of a thousand people, near-universal coverage is affordable in a way it never would be for an employer of a hundred thousand.
The money then widens sharply above the chip designers. Zhongji Innolight's 99 core recipients averaged gains of more than 26 million yuan each [4], roughly 2.6 billion yuan between them [7]. Zhipu's 426 pre-listing equity holders now average more than HK$100 million apiece [8]. The source's own won conversions imply about 207 won to the yuan and 177 won to the Hong Kong dollar [5], which puts the Zhipu average near 86 million yuan equivalent, some fifteen times Cambricon's core-staff grant [6]. A model researcher holding pre-IPO paper and a chip designer holding an RSU are being held in place by amounts an order of magnitude apart.
Where the shares come from is the second mechanism, and it is not uniform. Zhu Yiming has pledged 768 million shares, half his partnership stake in CXMT, over ten years, with himself explicitly excluded [7]: that pool comes out of a founder's holding rather than the float. The rest of the wave is financed by public shareholders while the market is up, which is what the filings are timed around, according to the South China Morning Post report relayed by en.sedaily.com [1]. Tencent's grant of more than 38.6 million shares equals 0.42% of the company [10]. Alibaba is moving the other way, lifting the share of long-term cash so pay is less exposed to the stock and lands sooner [11]. Both say they are trying to stop staff leaving for startups [12]. One of those two hedges will read badly within a year.
Coverage at 85% and above [3][6] is the part that travels. There is no thin stratum of stars to raid at these companies; the bonded population is the payroll. Anyone recruiting against them, in Hefei or in Hsinchu or in California, is no longer bidding against a salary band but against unvested paper struck at A-share and Hong Kong prices set during a rally.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Chinese AI and semiconductor companies are rolling out sweeping stock-reward programs to retain top talent, with recent corporate filings featuring a wave of stock-incentive plans from free share grants to company-wide schemes; companies are taking advantage of the buoyant stock market. Reported by the South China Morning Post on the 22nd, relayed by en.sedaily.com.
Cambricon, an AI chip designer, recently granted 600,000 shares to 124 core employees, worth an average of 5.57 million yuan (about 1.15 billion won) per person based on the share price on the day of the announcement.
Cambricon last month granted 5 million shares to 944 employees, 85.3% of its total workforce, as part of an incentive plan running through 2028.
Zhongji Innolight, a maker of optical transceivers for AI data centres, allocated 2.48 million shares to 99 core staff in April, delivering average gains of more than 26 million yuan (about 5.37 billion won) per person.
Chip equipment maker AMEC unveiled a restricted stock unit plan in March covering more than 97% of its workforce.
GPU developer Moore Threads added 1,080 employees, 85% of its total staff, to its reward programme in April.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Specific figures, single relayed source
The numbers are unusually concrete for a talent story — share counts, coverage percentages, per-head values, and Tencent's grant sized against shares outstanding — and they are attributed to corporate filings. But the cluster contains one publisher relaying one originating report, with no primary filings, company statements, or independent corroboration, and the interpretive layers (an 'unprecedented' wave, startup-driven attrition) are asserted rather than evidenced.
Payroll-wide plans across eight named firms
Adoption of broad-based equity as a retention tool is documented at eight distinct companies spanning chip design, chip equipment, optics, GPUs, DRAM, AI model labs and big tech, with coverage quantified as a share of headcount at four of them (97%+ at AMEC, 85.3% at Cambricon, 85% at Moore Threads, 9.8% of equity to 426 people at Zhipu). What is not observed is outcome adoption — no evidence that grants were accepted, vested, or changed retention.
Mildly overstated by framing and FX optics
The underlying disclosures are real and specific, but the presentation inflates them. Values are converted into billions of Korean won at roughly 207 won per yuan, making per-head grants read as spectacular; the headline values depend entirely on currently buoyant share prices; and the 'stock-reward war' and 'unprecedented' framing plus the startup-attrition motive go beyond what the article evidences. No dilution, expense, vesting or retention-outcome data is offered to test the retention claim.
Recruiting-signal disclosures, secondhand relay
Observable incentives run in one direction: the companies described gain from publicising generous, payroll-wide plans because such announcements are themselves recruiting and retention signals in a contested labour market, and pre-listing equity stories bolster listing narratives. On the reporting side, a Korean business outlet relaying South China Morning Post coverage has a readership interest in framing Chinese pay escalation as competitive pressure, evident in the won-denominated figures. Nothing in the cluster discloses commercial ties or undisclosed interests, so the score reflects structural incentive alignment only.
Moderate-low: one publisher, one origin
Confidence is constrained by structure rather than plausibility. Ten quantified, internally consistent disclosures point the same way, and the arithmetic checks out, but there is a single publisher relaying a single originating report, no primary filings, no dates precise beyond 'March', 'April' and 'last month', and no independent verification of any figure. The factual grants are credible; the retention effect and the 'unprecedented' scale are not yet testable.
build
SMIC's first $3 billion quarter comes with a wafer price increase attached1 distinct publisher
build
China's accelerator swap makes Cambricon supply, not export policy, your ship-date risk1 distinct publisher
invest
Unitree's $905M Shanghai listing prices humanoids at 35x sales while profit halves1 distinct publisher
build
H200s reach China at 2.5% of the order book, and Hong Kong holds the rest2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
en.sedaily.com
1 article · August 22, 2026