China's AI chip bellwether has delivered a doubling quarter. On the evening of October 23, Cambricon (688256) published its Q3 2026 report: quarterly revenue of 3.531 billion yuan, up 104.47% YoY, and net profit attributable to shareholders of 1.404 billion yuan, up 147.74%. The quarter alone contributes nearly 40% of nine-month revenue—and momentum is accelerating.

Rising bar charts with AI chip wafers, tech illustration
Doubling revenue and profit put domestic AI chip demand onto the balance sheet

1. 9.5B yuan in nine months, profits doubling across the board

Zoom out to the first three quarters: ① revenue of 9.526 billion yuan, up 106.76%; ② net profit of 3.715 billion yuan, up 131.49%, with non-GAAP profit of 3.523 billion yuan up 148.30%; ③ H1 revenue was 5.996 billion and Q3 alone 3.531 billion—a quarter-by-quarter climb. The company, whose portfolio spans cloud inference chips and cards, smart servers, edge products, IP licensing and intelligent computing cluster systems, attributes growth to market expansion and AI application adoption.

2. Cash flow turns positive as R&D scales

Two details stand out: ① operating cash flow for the nine months reached 228 million yuan versus -29 million a year ago, turning positive; cash collected from sales hit 10.671 billion yuan, up about 111.9%, signaling better collection quality; ② R&D spending totaled 1.225 billion yuan, up 45.35%, with Q3 alone up 73.66% to 523 million—yet with revenue growing faster, R&D intensity fell from 18.29% to 12.86% of revenue. Total assets reached 22.364 billion yuan, up 66.43% from year-start.

Q3 report at a glance

· Q3 revenue 3.531B yuan (+104.47%), net profit 1.404B yuan (+147.74%)

· 9M revenue 9.526B yuan (+106.76%), net profit 3.715B yuan (+131.49%)

· Operating cash flow 228M yuan (positive), collections 10.671B yuan (+111.9%)

· Inventory 11.665B yuan (+136% vs year-start), debt ratio ~32.4%

3. The 11.6B yuan inventory: ambition on one side, impairment risk on the other

The most striking line is inventory: 11.665 billion yuan at quarter-end, up about 136% from 4.944 billion at year-start; prepayments rose from 745 million to 3.044 billion yuan, and notes payable jumped from 351 million to 4.966 billion yuan. Two readings: the optimistic one sees forward stocking and capacity lock-in amid scarce AI chips—inventory is next year's revenue; the cautious one tracks inventory turnover and impairment risk. A debt ratio of about 32.4% remains manageable.

4. From storytelling to statements

For the domestic compute industry, this report matters beyond one company: revenue, profit and cash flow all turning positive together shows demand for domestic AI chips is real and scaling. For government and enterprise buyers, “capacity certainty”—not mere “usability”—is becoming the core consideration for domestic chips. NineZenith champions heterogeneous hybrid scheduling in compute platform construction: training, fine-tuning and inference workloads differ, and combining domestic and imported chips per workload is the pragmatic answer.

(Sources: Cambricon filings, Duchen Finance, Wallstreetcn, Phoenix Tech and other public reports)