China’s domestic AI compute is being tested where it counts — the income statement. According to Securities Star, Sina Finance and others, Cambricon’s Q3 2026 report shows revenue and net profit doubling YoY, with quarterly net income of about RMB 1.4 billion (+148%), a single-quarter record; nine-month cumulative profit reached ~RMB 3.7 billion. The stock briefly rose over 5% after disclosure.
1. From Narrative to Numbers
By the reported figures, the company has completed 70.56% of its annual revenue target, and operating cash flow turned positive, up 8.8x YoY. Cash flow is a harder test than profit — it means revenue is collected in cash rather than piled up as receivables. For domestic compute firms long reliant on external funding, positive operating cash flow marks the start of self-sustaining economics.
2. Two Controversies: Inventory and Shipment Rumors
On the earnings call, chairman Chen Tianshi addressed two focal points: ① high inventory (reportedly ~RMB 11.7B) — in the compute business, ample stocking often maps to order backlogs and delivery cycles; whether it is risk or reservoir depends on revenue conversion in coming quarters. ② The company denied market rumors of “1 million+ chips shipped,” saying it had not communicated such figures. Expect more of these clarifications while results scale faster than sentiment can track. Some overseas outlets noted profit doubled but still fell short of certain analysts’ estimates — the market already prices in aggressive growth.
3. What’s Next: Ecosystem and Cadence
Long-term challenges remain clear: ① closing the software-ecosystem gap with CUDA through toolchains and industry adaptation; ② chip generational cadence to capture training and inference demand; ③ supply-chain capacity amid global memory and advanced-node tightness. Delivering earnings is the first half; ecosystem maturity is the second.
For government and enterprise buyers, domestic compute usability is improving fast — weigh ecosystem maturity and delivery capability alongside performance specs. NineZenith follows the domestic compute ecosystem closely, supporting clients with full-stack planning from models to compute. (Compiled from public reports.)