The 85% Gamble: Can Cambricon Turn Every Engineer Into a Millionaire?

On the evening of July 28, 2026, Cambricon filed a draft of its 2026 Restricted Stock Incentive Plan.
945 people. 5 million shares. Grant price: 750 yuan per share. Closing price that day: 1,128 yuan.
Paper spread per person: 1.6 million yuan.

It covers 85.37% of all employees.

This is not a perk. This is a ticket to a company‑wide wager.

You have to buy in.
The average allocation is about 4,233 shares per person. At the grant price, that is roughly 3.17 million yuan out of each participant's own pocket. The shares vest over 54 months. That means an employee must not only put up real money, but also stay with the company for four and a half years, and if performance targets are missed, the paper fortune simply vanishes.

Equity incentives are not unusual among listed companies. What is unusual is the combination of coverage, performance bar, and the competitive field it sits in.

I. Why Now

Cambricon's 2025 revenue was 6.497 billion yuan. The incentive plan demands:

• 2026 revenue ≥ 13.5 billion

• cumulative 2026–2027 revenue ≥ 40.5 billion

• cumulative 2026–2028 revenue ≥ 100 billion

That is a three‑year compound annual growth rate of over 148%.

That is the first number worth staring at.

The second number: 560,000 chips. Alibaba's Pingtouge has deployed its "Zhenwu" series AI chips on Alibaba Cloud, with shipments reaching that level. At the same time, ByteDance is developing its own chips, Huawei's Ascend is iterating, and Baidu's Kunlun is moving forward.

These companies share one thing – they are all potential customers of Cambricon, or already are.

ByteDance has long been one of Cambricon's largest clients. One customer doing two things at once: buying chips from you, and making its own. When self‑developed production capacity ramps up, what happens to the purchase orders?

A headhunter who knows the industry well said in early July: "Right now it's not Cambricon versus Moore Threads fighting over people. It's Cambricon people sitting at the same dinner table with ByteDance people, and Alibaba people handing out business cards."

Among the 945 covered by the plan, R&D staff make up the bulk. Cambricon's R&D headcount has long hovered around 80% of total staff. With this 85.37% coverage, it means essentially every R&D core member is now inside the same bet.

This is not "retaining talent." It is turning talent into a community of interest, while also raising the cost of exit. The 3.17 million yuan outlay plus the 54‑month lock‑up will make anyone thinking of leaving do the math twice.

II. The Money Is Not Free

Buy‑in at 750 yuan per share. Same‑day market price 1,128 yuan. Paper profit 378 yuan per share.
But "paper" means: not yet in hand.

Vesting is conditional, with two tiers:

• Cumulative three‑year revenue reaches 100 billion → full vesting

• Cumulative three‑year revenue reaches 80 billion but below 100 billion → 80% vesting for that tranche

• Below 80 billion → zero vesting for that tranche

Executives face additional net profit growth requirements.

In other words, what employees receive is not a "bonus" but a "profit‑sharing right": the company must first deliver the performance, and only then does that right become effective.

Multiple brokerages give a neutral forecast: cumulative three‑year revenue might land in the 86‑88 billion yuan range. That would likely trigger the 80% tier, but fall short of full vesting.

Eighty percent is still an acceptable result. But here is the catch: the 80‑billion trigger line itself is twelve times Cambricon's 2025 revenue of 6.497 billion. Growing twelve‑fold in three years; that curve demands geometric market‑share expansion, not linear growth.

III. The Real Competitors Are Not on the List

The line that is not written in the incentive plan is this: customers are turning into rivals.

Alibaba's Pingtouge has already completed an internal closed‑loop validation of its Zhenwu series on Alibaba Cloud. Once a chip passes validation in its own scenario, the next step is external sales; price, performance, and adaptation will all become direct pressure on Cambricon.

ByteDance's self‑developed chip team expanded by roughly 40% in the first half of 2026. Headhunter feedback shows that ByteDance's total compensation packages for AI chip engineers are generally 20‑30% higher than Cambricon's for equivalent ranks.

But ByteDance faces the same issue as Alibaba: self‑developed chips take time to validate. During that window, they still need to buy from outside. What Cambricon must do is, before that window closes, shift its customer structure away from reliance on a few large tech giants toward a more diversified base.

On its earnings call for the first half of 2026, Cambricon's management mentioned two directions: government and telecom operators. These sectors have clear policy drivers for domestic chips, and replacement cycles are long; once a chip is adopted, it is not easily swapped out.

But government and telecom decision cycles are also long. The process for a chip to enter a centralized procurement catalogue takes at least one year, often two. And the three‑year performance window leaves no room for a single misstep.

IV. The 1.914 Billion Yuan Cost

There is one more figure in the plan that is easy to overlook: the share‑based compensation expense is estimated at 1.914 billion yuan.

The amortisation schedule:

• 2026: roughly 355 million yuan

• 2027: roughly 806 million yuan

• 2028: roughly 575 million yuan

• 2029: roughly 178 million yuan

In 2027, over 800 million yuan in non‑cash expense will flow straight into the income statement. Cambricon only turned profitable in 2025. In 2026, its income statement will carry an extra 355 million yuan of "paper cost"; the money never leaves the company, but it makes the net profit figure look uglier.

The market looks at two things: whether revenue growth is delivered, and whether cash flow is healthy. Share‑based payments do not affect cash flow, but they do affect the market's perception of profitability. When a company faces both "must grow at high speed" and "book earnings under pressure," investor patience gets tested repeatedly.

V. The Rival's Tempo

Around the same time, Moore Threads, another domestic GPU/AI chip company, also rolled out an incentive plan: roughly 760,000 shares granted to 1,042 core employees, with 38 opting out.

Two independent AI chip companies moving almost simultaneously says one thing: the window is short, and whoever locks in people first gains an extra breath.

Moore Threads leans more toward GPUs, while Cambricon focuses on dedicated AI chips. Their technical paths differ, but the talent pool is the same. A chip architect can move between the two companies at the cost of perhaps a single dinner conversation.

The incentive plans themselves are signals. Cambricon's covers the whole workforce; Moore Threads' covers only core staff. The strategic difference is: Cambricon bets that "everyone must fight together," while Moore Threads bets that "key people cannot leave." Neither is right nor wrong; they are choices suited to their current scales.

VI. What to Watch Over the Next Three Years

The effect of this incentive plan will not show in August 2026.

It will be tested at three points in time.

First checkpoint: early 2027, when the 2026 annual report is released.
Did the 13.5 billion revenue target materialise? If not, the first vesting tranche gets discounted, and the market will question the feasibility of the following two years.

Second checkpoint: early 2028, when the cumulative 40.5 billion target for 2026‑2027 is assessed.
The cumulative revenue over these two years determines whether the 80% vesting trigger can be hit. If the first two years only add up to 30 billion, then the third year needs to deliver 50 billion just to touch the 80‑billion cumulative threshold; that pressure will feed into the stock price a full year earlier.

Third checkpoint: early 2029, when the three‑year cumulative number is finalised.
Whether it lands at 100 billion, 80 billion, or lower will decide whether the 945 people walk away with a handsome return, a passing grade, or four pieces of worthless paper.

But there is a fourth checkpoint, one not written in the plan document: the maturity level of Alibaba's and ByteDance's self‑developed chips by 2028. If by then their in‑house chips have massively replaced external purchases, Cambricon's hundred‑billion goal will lose a critical pillar.

Cambricon wrote one number into the plan: 5 million shares, 0.80% of total equity.

0.80% looks small. But it ties together 85% of the staff, 1.914 billion yuan in cost, and a promise to grow twelve‑fold in three years.

By the end of 2028, Cambricon's stock price will speak for this plan. Not the vision articulated by management at earnings calls, but the number that flashes on the exchange screen. It will compress every judgment, whether talent stayed, whether performance was delivered, whether customers became rivals, into a single answer.

Until then, 945 people, when they get their monthly paychecks, will see a line next to the salary figure: Tranche X restricted stock, vesting progress X/54.

The day that the progress bar hits 54, the answer comes out.