The race for Artificial Intelligence hardware in China has entered a more mature phase: it’s no longer measured solely by prototypes or performance promises, but by something much more uncomfortable for any founder: how much revenue is generated and whether there’s a profit at the end of the period. And in 2025, two of the most closely watched names in the fabless ecosystem (designers outsourcing manufacturing) have shown a contrast that helps understand the sector’s current state: Cambricon has turned profitable, while Moore Threads has grown strongly but still operates at a loss.
The data published in late February and early March 2026 place both companies on the same track —namely, the rapid expansion of revenue driven by domestic computing demand— but with different financial outcomes.
Cambricon: revenue jump that ends in profit
Cambricon Technologies (688256) closed 2025 with a revenue increase of 453.21%, reaching 6.5 billion yuan (CNY), according to its preliminary earnings report. Simultaneously, the company reported a net attributable profit of 2.06 billion yuan, turning profitable after last year’s losses. This figure marks a significant turnaround for a company that, according to market monitoring, had gone years without a full positive fiscal year.
Beyond the headline, the market interprets this as a signal: in an environment where R&D investment remains high, Cambricon has not only increased sales but has also crossed the profitability threshold. The big question now is sustainability: how much of this leap is driven by structural demand (data centers, inference, institutional purchases) and how much depends on intensive procurement cycles.
Moore Threads: rapid growth, but still operating at a loss (though narrower)
Moore Threads Technology (688795) also had an expansion year, with revenue of 1.51 billion yuan in 2025, a 243.37% increase year-over-year. However, the company remains unprofitable: it reported a net loss of 1.02 billion yuan, although this loss was reduced compared to 2024, when it was significantly larger.
In its market communication, the message echoes that of many scaling chip companies: revenue and margin improvements driven by higher product acceptance, but cost structure still dominated by investments in development, ecosystem building, and sales capacity. Even Chinese financial media emphasize that the company continues in a phase of heavy R&D spending and that, compared to international giants, it still lacks the technological muscle and platform maturity to compete directly.
A plausible explanation: product, scale, and “system economics”
At first glance, one might think that Cambricon’s turnaround and Moore Threads’ ongoing losses can be explained solely by “selling more.” But in semiconductors — especially in AI acceleration — the nuance lies in what is sold and how it is sold:
- Product mix and average selling price (ASP): certain contracts and segments (e.g., enterprise or institutional deployments) may boost revenue and margins faster than a fragmented market.
- Cost of building the ecosystem: compilers, libraries, framework compatibility, and customer validation are as important as the chip itself. This investment can delay profitability.
- Scale and operational leverage: as volume grows, some fixed costs are diluted, but the breakeven point arrives at different times for each company.
In other words: 2025 confirms that domestic demand is fueling growth, but not all companies are converting that momentum into profit at the same pace.
The context accelerating it all: domestic demand and technological substitution
This contrast doesn’t happen in a vacuum. Geopolitical pressure and trade restrictions have created an environment where manufacturers and operators seek local alternatives. In this scenario, Chinese AI chip designers find more doors open for pilots, purchases, and deployments, which is reflected in revenue growth.
In fact, this trend isn’t exclusive to these two firms. Other Chinese GPU/AI actors have also reported growth, though with different dynamics in losses and scaling. The takeaway from 2025 is clear: China is pushing AI hardware strongly, but the industry is splitting into two groups: those already monetizing with profits and those still paying the toll of platform development.
Summary table: 2025 revenue and profitability (CNY)
| Company | 2025 Revenue | YoY Change | Net Result 2025 | Quick take |
|---|---|---|---|---|
| Cambricon | 6.5 billion | +453.21% | +2.06 billion | Returns to profits with strong sales expansion |
| Moore Threads | 1.51 billion | +243.37% | -1.02 billion | Rapid growth but still losses (less than 2024) |
Frequently Asked Questions
What does it mean for a company to be “fabless” in AI chips?
It means designing processors and accelerators but typically outsourcing manufacturing to foundries. Their advantage lies in architecture, software, and product, with challenges in R&D costs and access to capacity.
Why is Cambricon profitable while Moore Threads isn’t, despite both growing?
Because profitability depends on product mix, ASP, ecosystem development costs, and the timing when volume makes fixed costs sustainable. In AI chips, the system matters as much as the silicon.
What do these figures imply for the AI hardware market in China in 2026?
That domestic demand is sustaining aggressive growth, but the market will begin to demand results: it’s no longer enough to sell more; margins and business continuity will need to be demonstrated.
Could this pattern repeat among other Chinese GPU/AI designers?
Yes. It’s common for several companies to grow rapidly in revenue while still operating at a loss due to investments in R&D, software support, and platform building. The market will tend to reward those who cross the profitability threshold first.