Moore Threads reported H1 revenue of 1.736 billion RMB (+147% YoY) with narrowed losses but negative operating cash flow.

On August 9, Moore Threads released its first half-year report since going public. Revenue reached RMB 1.736 billion, up 147% year-on-year, surpassing its full-year 2025 total. Net profit attributable to shareholders narrowed losses by 95.73% to RMB 11.56 million.

While doubling revenue and significantly narrowing losses appear to be positive signals on the surface—with many market voices suggesting Moore Threads is approaching an inflection point—analyzing financial statements requires looking beyond growth rates and loss reduction. It is crucial to examine what is driving the revenue and whether the narrowing losses stem from operational improvements or accounting factors.

Moore Threads’ revenue grew from RMB 46 million in 2022 to RMB 124 million in 2023, RMB 438 million in 2024, and RMB 1.506 billion in 2025, reaching RMB 1.736 billion in the first half of 2026. Moving from tens of millions to billions means that recognizing any single large project can artificially inflate growth figures. With a relatively low base of approximately RMB 703 million in H1 2025, the 147% surge was largely amplified by that small denominator. Although H1 2026 revenue exceeded the full-year 2025 total, the absolute incremental gain was only RMB 230 million, indicating that the scale of growth remains modest.

Whether subsequent quarters can sustain this growth rate depends on the actual delivery pace of signed orders. The financial report shows that revenue associated with deferred performance obligations totals only RMB 87.52 million—a significant gap compared to the RMB 1.736 billion revenue scale.

In 2025, sales to the top five customers accounted for 91.36% of Moore Threads’ total revenue. While H1 2026 did not separately disclose revenue concentration, accounts receivable and contract assets from the top five customers still represented a staggering 99%. In H1 2025 alone, sales to Customer R amounted to RMB 397 million, representing 56.63% of total revenue for that period.

High customer concentration is somewhat common across the domestic GPU sector. According to raw data from the 2025 annual reports, Cambricon’s top five customers accounted for 88.66% of sales, while Hygon Information’s stood at 90.28%. Objectively speaking, the entire industry operates with high client dependency, though Moore Threads’ concentration level remains relatively pronounced among listed domestic GPU companies.

As a semiconductor industry analyst noted, when over 50% of a company’s revenue comes from a single client, the sustainability and predictability of that income require careful assessment. This implies that any adjustment in a major client’s purchasing rhythm would significantly impact Moore Threads’ quarterly revenue.

In H1 2026, cloud intelligent computing products contributed RMB 1.693 billion to revenue, accounting for 97.5% of the total, while edge and terminal businesses combined generated just RMB 40.52 million. Edge AI chips, however, represent one of China’s fastest-growing segments. Market institutions estimate the 2025 market size at approximately RMB 12 billion, projecting it to exceed RMB 30 billion by 2028 with a CAGR of over 35%. Currently, this space remains dominated by international vendors, with domestic GPU manufacturers maintaining a relatively low overall penetration rate.

Moore Threads has made virtually no revenue contribution in this segment. Should demand for cloud intelligent computing or policy-driven procurement shift, its other product lines are unlikely to provide meaningful performance supplementation in the short term.

Moore Threads’ net profit attributable to shareholders narrowed losses from RMB 271 million in H1 2025 to RMB 11.56 million, appearing to bring the company close to profitability at first glance. However, understanding the true operational picture requires dissecting non-recurring gains and losses.

In H1 2026, non-recurring gains and losses totaled RMB 139 million, primarily driven by government subsidies of RMB 88.23 million and fair value changes plus disposal gains/losses on financial assets amounting to RMB 59.52 million. After excluding these items, the adjusted net profit still registered a loss of RMB 151 million. This indicates that the dramatic narrowing of reported losses was mainly propelled by government grants and wealth management investment returns, with core operations yet to see fundamental improvement.

Government subsidies are widespread in China’s domestic chip sector, with most A-share listed semiconductor companies receiving varying levels of support. Whether enterprises can achieve independent profitability through core operations once subsidies taper off remains a critical issue requiring long-term tracking.

By the end of June, Moore Threads’ inventory surged from RMB 1.332 billion at the end of 2025 to RMB 3.55 billion, while prepayments reached RMB 1.344 billion. Combined, these two asset categories approached RMB 4.9 billion—nearly 2.8 times the H1 2026 revenue. The company attributed the inventory buildup to proactive upstream capacity reservation and advance stocking. However, with deferred performance obligations corresponding to only RMB 87.52 million in revenue, there is a clear mismatch between order absorption capacity and inventory scale.

Elevated inventory levels are a pervasive phenomenon across the domestic GPU industry. Institutions such as Goldman Sachs have noted that domestic AI chip market demand growth slowed in Q2 2026 compared to Q1. If downstream demand continues to weaken, inventory digestion pressures across the sector will intensify. For Moore Threads, stockpiling is no longer merely a latent risk; the company already booked RMB 39.05 million in inventory impairment provisions in H1 2026, directly eroding current-period profits.

In H1 2026, Moore Threads recorded a net operating cash flow of negative RMB 2.169 billion. Operating cash flow has remained negative since 2022, with the cash outflow gap widening annually. The RMB 6.492 billion in cash and cash equivalents on the balance sheet stems primarily from IPO proceeds rather than operational accumulation. This indicates the company has yet to achieve self-sustaining cash generation from operations and continues to rely on external financing to maintain activities.

Revenue can be temporarily boosted by concentrated orders, and book profits can be polished through government subsidies and investment returns. Yet, operating cash flow remains the definitive metric for assessing a company’s true operational quality.

Zooming out from Moore Threads to the broader landscape reveals clear performance divergence within China’s domestic GPU industry, alongside multiple shared challenges facing all players in the sector.

On August 7, Cambricon released its H1 2026 financial report. Revenue reached RMB 5.996 billion, up 108.13% year-on-year; net profit attributable to shareholders hit RMB 2.311 billion, a 122.61% increase; and adjusted net profit stood at RMB 2.166 billion, up 137.30%. The company has now achieved profitability for seven consecutive quarters, with H1 2026 operating cash flow netting RMB 311 million.

Both Moore Threads and Cambricon belong to the top tier of domestic AI chipmakers, yet their financial fundamentals have diverged significantly. Moore Threads’ current revenue represents roughly one-third of Cambricon’s, and their operating cash flows stand in stark contrast—one deeply negative, the other strongly positive. Huatai Securities’ research report analyzes that this performance split marks the domestic GPU industry’s transition away from a phase of “sector-wide continuous investment and losses” into a cycle where leading firms deliver results and market stratification accelerates.

CAICT’s *AI Chip Software Ecosystem White Paper (2026)* notes that while domestic AI chips are rapidly closing the hardware performance gap with international leaders, the maturity gap in software ecosystems remains substantial. Industry consensus holds that the hardware parameter gap has shrunk from three to five years in 2020 to just one to two years today, whereas the software ecosystem maturity gap persists at over five years. Across dimensions such as developer toolchain completeness, compatibility with mainstream AI frameworks, and open-source community developer activity, domestic GPUs still lag considerably behind NVIDIA.

Industry experts note that while hardware compute power and chip specifications can be rapidly caught up through sustained capital investment, building a software ecosystem demands long-term developer accumulation and continuous version iterations, making short-term catch-up difficult.

Moore Threads publicly disclosed that its platform developers have surpassed 800,000, with its MUSA architecture achieving deep compatibility with CUDA. However, the financial report omits key conversion metrics, such as the scale of clients migrating from NVIDIA, operator completeness in large model training and inference scenarios, and commercial repurchase rates. Consequently, there is a lack of quantitative evidence validating the commercialization effectiveness of its ecosystem.

According to shipment-based statistics from Frost & Sullivan, the domestic market share of homegrown general-purpose GPUs rose from 8.3% in 2022 to 17.4% in 2024. Institutions project that localization shipment proportions could exceed 50% by 2029.

Nevertheless, multiple brokerage research reports warn that a significant portion of recent localization gains stems from centralized IT application innovation (Xinchuang) procurement by government and enterprise sectors. Following the procurement surge in 2024–2025, Xinchuang purchase growth faces downward pressure in 2026. Should procurement rhythms slow, domestic vendors’ revenues will face volatility risks.

Impacted by entity list restrictions, Moore Threads’ overseas revenue share dropped from 0.13% to zero in H1 2026. Cambricon’s overseas revenue accounted for just 0.5% in 2025, while peers like MetaX and Biren Technology also report negligible foreign sales. Industry estimates place the 2025 domestic AI chip market size at approximately RMB 180 billion, representing roughly 22% of the global AI chip market. This implies that nearly 80% of the global addressable market remains out of reach for some domestic GPU makers, tying their growth heavily to domestic policy-driven procurement and creating a structural ceiling.

In 2025, NVIDIA shipped 2.2 million AI acceleration cards in China, capturing a 55% full-year market share. In May 2026, Jensen Huang publicly stated that U.S. export control policies prevent NVIDIA’s high-end training-grade AI chips from entering the Chinese market, effectively ceding the premium segment’s market space to local vendors.

IDC data shows that China’s total AI acceleration card shipments reached approximately 4 million units in 2025, with domestic vendors collectively shipping 1.65 million units, bringing the overall local market share to 41%. Breaking it down: Huawei Ascend shipped 810,000 units (20% share); Cambricon shipped 120,000 units (3%); Hygon Information shipped 80,000 units (2%); and six mainstream domestic GPU manufacturers combined held roughly 35% of the market.

Despite steadily rising domestic market shares, absolute monopoly advantages have not yet formed, leaving overseas vendors as a critical variable. While U.S. restrictions on NVIDIA’s high-end chips have created a short-term development window for domestic GPUs, this substitution cycle is not permanent. Multiple brokerages jointly analyze that NVIDIA’s overseas advanced process capacity bottlenecks are expected to gradually ease in H2 2027. Should export control policies adjust, overseas vendors may resume expanding supply in the Chinese market, leaving domestic GPU makers with an estimated 12 to 18 months to complete commercialization and ecosystem building.

Tianyancha data indicates that founded in 2020, Moore Threads listed on the STAR Market in 2025 and is advancing an H-share listing in 2026. While the company’s expansion pace has been rapid, multiple structural shortcomings remain unresolved during its growth phase. Cambricon’s consecutive profitable quarters demonstrate that domestic GPUs can achieve commercial profitability. However, both leading firms and other players in the sector continue to face shared challenges, including customer concentration, elevated inventory levels, and overseas market restrictions.

China’s domestic GPU industry stands at a crossroads between policy dividend realization and commercial deployment. While domestic market opportunities continue to expand, core hurdles regarding software ecosystems, overseas distribution channels, and sustainable profitability models remain unovercome. The competitive landscape is far from stabilized. Only companies that systematically address these multifaceted weaknesses will secure their footing in the industry’s long-term competition.