Moore Threads' half-year guidance moves China's domestic GPU race from one question to another. The issue is no longer only whether the chips can be built. It is whether they can be sold, delivered and kept in production.
On the evening of July 16, the STAR Market GPU company said it expects first-half 2026 revenue of RMB 1.65 billion to RMB 1.75 billion. Even the low end of that range is above its roughly RMB 1.505 billion revenue for all of 2025. For a chipmaker founded in 2020 and listed only at the end of 2025, six months of revenue exceeding the previous full year says more about market acceptance than another launch event would.
The company gave a direct explanation in its announcement: strong market demand for full-function GPUs, faster commercialization of its KUAE intelligent computing cluster and stable product supply. The most important phrase is stable supply. For domestic AI chips, the hardest questions have rarely stopped at PowerPoint performance. Buyers care about delivery, repeat purchases and whether clusters can run reliably.
First-half revenue has already overtaken last year
The guidance packs in several numbers:
- Expected first-half 2026 revenue of RMB 1.65 billion to RMB 1.75 billion;
- 235.12% to 249.37% of the same period last year;
- year-on-year growth of 135.12% to 149.37%;
- an increase of RMB 948.2381 million to RMB 1.0482381 billion from a year earlier;
- first-half 2025 revenue of RMB 701.7619 million.
Follow-up reports from The Paper and Securities Daily added a useful benchmark: Moore Threads generated about RMB 1.505 billion in revenue for all of 2025. That means the company is very likely to have surpassed last year's full-year revenue in the first six months of 2026.
The industrial signal is not just that another Chinese GPU company is growing. It is the shape of the revenue curve. Training cards, inference cards and cluster solutions rarely reach this kind of half-year growth through one-off buying alone. The numbers usually point to continued delivery for large customers, or to a shift from trial orders into scaled deployment.
S5000 has become a delivery clue, not just a product name
Moore Threads also named the MTT S5000 in the announcement. The company said the flagship full-function GPU accelerator for training and inference has reached scaled mass production and is being delivered as part of large clusters built around the product, with support for training trillion-parameter models.
This earnings forecast has not been audited by certified public accountants.
That caveat belongs in the main story too. The announcement did not provide single-card performance, cluster size, customer names or a full income statement. What it does confirm is a different milestone: the S5000 has moved from launch-stage product to a revenue-contributing delivery line.
That matters for domestic AI chips. Chinese model companies and cloud providers often need more than peak performance on one card. They need the software stack, interconnects, scheduling and supply cycle to hold together. The KUAE cluster appearing in the explanation suggests Moore Threads wants to sell a full package of cards, clusters and software rather than scattered hardware.
The growth looks strong, but the risks are in the announcement too
The company also set clear boundaries. It said the final and accurate financial data should be based on the 2026 interim report.
That limit should not be ignored. In the first quarter of 2026, Moore Threads recorded RMB 738 million in revenue and RMB 29.3592 million in net profit attributable to shareholders, but recurring net profit after exclusions was still a loss of RMB 54 million. For the first-half forecast, the revenue step-up is now fairly clear. Profit quality, cash flow and customer concentration still have to wait for the interim report.
Market expectations are already high. The Paper reported that Moore Threads' market value was about RMB 305 billion at the July 16 close. Against first-half revenue of RMB 1.65 billion to RMB 1.75 billion, investors are clearly pricing more than current-period earnings. They are also betting on the coming years of procurement demand for domestically made AI compute in China.
Three things to watch in the interim report
After this update, the next indicators are concrete.
First, watch the customer structure behind the S5000 and KUAE cluster revenue. If the revenue comes from only a few projects, volatility will be high. If customers across multiple industries are starting to reorder, the commercial value of domestic GPUs becomes stronger.
Second, watch gross margin and recurring profit after exclusions. Chip revenue can grow quickly, but long-term value is weaker if growth is built on expensive delivery. Moore Threads reported full-year 2025 gross profit of RMB 987 million and an overall gross margin of 65.57%. Whether that older measure can hold in the first half of 2026 is a hard test.
Third, watch whether the WAIC demos for a model training factory, token production factory and agent production factory come with real customer cases. Launch slogans are easy. Orders and deployment scenarios will decide whether Moore Threads can move from a domestic GPU leader narrative to a durable AI infrastructure company.
The value of this guidance is that it pulls the conversation back from import substitution to the business itself: revenue, delivery, mass production, clusters and audited financial statements. In the end, chip stories have to land in those cold financial tables.
Sources: Moore Threads 2026 interim earnings forecast announcement, Zhidongxi, Securities Daily, CocoLoop, The Paper. The Shanghai Stock Exchange announcement was used to verify the revenue range, year-on-year growth, prior-year base, audit status and MTT S5000 mass-production description; Securities Daily and The Paper were used to verify WAIC demo clues, 2025 full-year revenue, first-quarter 2026 revenue and market-value figures.