China Open Source Daily — 2026-09-11

⚠️ Editorial note: The open source ecosystem in China operates under a distinct institutional framework — characterized by state-led initiatives, intranet-like boundaries, and top-down governance. Readers should be aware that this context differs from the community-driven open source model common in other regions. The term “open source” as used in Chinese media may refer to practices that diverge from the conventional definition.

China Open Source Daily — 2026-09-11

📖 Party-Theoretical Codification — Chen Kaihua in Qiushi (2026/17): Open-Source Community Formally Named as a Scientific-Research Subject, Alongside Data Platforms and AI Agents

1. Qiushi / Qiushi WeChat (August 31, 2026, 7:40 PM; published in Qiushi 2026/17) — Chen Kaihua (陈凯华), invited professor at UCAS School of Public Policy and Management, “人工智能如何引领科研范式变革” (“How AI Leads the Transformation of the Research Paradigm”)

Qiushi (求是) is the theoretical flagship of the Central Committee of the Communist Party of China — the journal where, since 1994, the Central Committee has published its most formal ideological formulations. The article is published in Qiushi 2026/17 (August 31, 2026) and simultaneously carried by the Qiushi WeChat account, giving it the same dual-placement formal weight as other party-theoretical formulations documented in this series.

The article’s four-part diagnosis documents four explicit Chinese-side institutional weaknesses against the U.S. Genesis Plan baseline (announced November 2025 as a Manhattan-scale national research mobilization; July 2026 formalized by the White House OSTP’s “Science: The New Golden Age” report):

Weakness 1 (algorithm, compute, data supply) — China’s compute-resource total is world-leading, but compute supply for research scenarios is insufficient, unevenly distributed, lacks a national intelligent compute-scheduling platform, and lacks unified data standards and open-sharing mechanisms.
Weakness 2 (interdisciplinary and talent mechanism) — the current single-discipline, department-boundary classification-management system cannot match the fast-fusion knowledge-innovation demand of the intelligent era; research evaluation and resource allocation are blocked by discipline barriers.
Weakness 3 (subject coordination and element fusion) — “open-source communities, data platforms, research intelligent-agents, and other new knowledge subjects or carriers — their coordinated operation is not smooth; the intelligentization level of various innovation subjects is significantly insufficient; sub-systems lack open-sharing and coordinated coupling; there is no supporting whole-chain innovation-paradigm-reshaping combined force.” This is the passage that formally names open-source community (开源社区) as a 科研主体 (scientific-research subject) — the first documented placement of an open-source community as a formal research subject on a Chinese party-theoretical surface.
Weakness 4 (institutional mechanism and coordinated infrastructure) — current organizational architecture and institutional arrangements are “clearly lagging behind” the AI-driven research-paradigm development process; “the organization and safeguarding mechanism of open-source and open infrastructure is still in the exploration stage” (开源开放的基础设施组织和保障机制仍处于探索阶段); traditional research-funding, evaluation, and organization mechanisms cannot satisfy open and open-source demand.

Against these four diagnosed weaknesses, the article proposes four policy directions. The direction that carries the sharpest institutional-economics implication is the third — “Actively Cultivate an Open and Cooperative Innovation Ecosystem” (积极培育开放协同创新生态):

Build a national-level open-source community for AI-driven research (加快建设国家级人工智能驱动科研的开源社区).
Explore a new mechanism combining code contribution with compute incentive (探索代码贡献与算力激励有机结合的新型机制).
Lead the formulation of key global standards on scientific-data quality and algorithm reproducibility.
Actively participate in international scientific-data sharing agreements.
Increase China’s international voice and influence in research-paradigm transformation.

The article’s own conclusion (as the third of four policy directions) is the institutionally sharpest formulation in the piece — it places open-source community building, alongside data-platform building and AI-research-agent building, as the third formal institutional pillar of a Chinese AI-driven-research-paradigm reform agenda, with an explicit institutional-form proposal (code-contribution-compute-incentive mechanism) that has never appeared in this series’ prior twenty briefings.

Institutional significance: The Chen Kaihua Qiushi article is the first documented placement of an open-source community as a formal 科研主体 (scientific-research subject) on a Chinese party-theoretical surface — the same research-paradigm question that this series’ September 9 briefing’s CCTV codification first broadcast to a general audience, now documented at the party-theoretical-discourse layer with an explicit institutional-form proposal.

From an institutional economics perspective, the Chen Kaihua Qiushi article matters on four axes:

First, the open-source-community-as-科研主体 placement is institutionally the sharpest institutional-form observation of this cycle — it is a formal placement of an open-source community as a scientific-research subject on a Chinese party-theoretical surface, and the same placement is proposed to be operationalized at a national scale (a national-level open-source community for AI-driven research). The prior twenty briefings had documented the Chinese institutional architecture’s open-source framing at multiple institutional surfaces — the state-council-policy-discourse layer (MIIT Liu Yulin, August 28 briefing), the state-media-national-economy layer (CCTV, September 9 briefing), the party-theoretical-journal layer (Xie Lan, August 31 briefing, 中国社会科学报; Ren Xudong, September 5 briefing), and the Western-foundation-governance-seat layer (PyTorch Foundation, September 10 briefing). The Chen Kaihua Qiushi article is the party-theoretical-flagship placement of the same question — but unlike the prior party-theoretical placement (Xie Lan in 中国社会科学报, a social-science journal), the Chen Kaihua placement is on the Central Committee’s theoretical flagship (Qiushi), and unlike the CCTV placement (September 9 briefing), it is a research-recommendation article rather than a broadcast. From an institutional economics standpoint, this state-media-broadcast / party-theoretical-flagship pairing is a structural finding: the same research-paradigm question is being placed at two institutional surfaces — the state-media broadcast surface (CCTV, September 9 briefing) and the party-theoretical-flagship surface (Qiushi, today) — with the party-theoretical-flagship surface carrying a specific institutional-form proposal (national-level AI-research open-source community) that the state-media surface did not make. The institutional-economics reading is that the research-paradigm question has now crossed from state-media broadcast to party-theoretical-flagship formulation — a structurally new institutional-form object that the prior nineteen briefings had not observed.

Second, the code-contribution-compute-incentive mechanism (代码贡献与算力激励有机结合的新型机制) is institutionally a formal proposal for an open-source contribution economy at the state-research level — the first documented proposal for this mechanism on a Chinese party-theoretical surface. The article explicitly proposes exploring “a new mechanism combining code contribution with compute incentive” (探索代码贡献与算力激励有机结合的新型机制) as the operating principle of the national-level AI-research open-source community. This is a structural institutional-form proposal — it names a specific governance-mechanism for the proposed national-level open-source community. From an institutional economics standpoint, this contribution-compute-incentive pairing is a structural finding: the Chinese institutional architecture is proposing a specific mechanism to reward open-source contribution with compute resources — an institutional-form proposal that has never appeared in this series’ prior nineteen briefings, and that has institutionally direct parallels in Western-foundation governance-seat distributions (PyTorch Foundation, September 10 briefing) but at a different incentive layer (compute incentive, not governance-seat incentive). The institutional-economics reading is that the contribution-compute-incentive mechanism is a structurally new institutional-form object on the open-source contribution-economy layer — a formal Chinese-party-theoretical proposal for how open-source contribution will be rewarded inside the proposed national-level AI-research open-source community.

Third, the Genesis-Plan-baseline / asymmetric-catching-up pairing is institutionally the same institutional-form object that this series’ August 13 briefing’s Chatham House “open-weight vs. closed-weight” institutional-model comparison documented at the external-comparison layer — the Chinese-side formulation of the same U.S.-China research-mobilization question, now documented on the party-theoretical-flagship surface. The article explicitly names the U.S. Genesis Plan (announced November 2025, formalized July 2026 by the White House OSTP’s “Science: The New Golden Age” report) as the specific benchmark for the Chinese-side research-paradigm reform agenda, and explicitly frames the Chinese-side advantage as one of “asymmetric catching up” (非对称赶超). From an institutional economics standpoint, this external-comparison / domestic-formulation pairing is a structural finding: the same U.S.-China research-mobilization question is being formulated at two institutional surfaces — the external-comparison surface (Chatham House, August 13 briefing) and the domestic-formulation surface (Chen Kaihua in Qiushi, today) — with the domestic-formulation surface explicitly naming the external comparison (Genesis Plan) and explicitly framing the Chinese response (asymmetric catching up). The institutional-economics reading is that the research-mobilization question has crossed from an external-comparison framing to a domestic-side party-theoretical-flagship formulation — the Chinese institutional architecture is now formulating its own research-mobilization answer on the party-theoretical-flagship surface, rather than responding to external framing on a Chinese-state-media surface.

Fourth, the “still in the exploration stage” (仍处于探索阶段) qualifier on the open-source open-infrastructure organization and safeguarding mechanism is institutionally the sharpest self-diagnostic statement in the article — the Chinese institutional architecture is publicly declaring that its open-source open-infrastructure institutional mechanism is not yet operational, on a party-theoretical-flagship surface. The article explicitly documents that “the organization and safeguarding mechanism of open-source and open infrastructure is still in the exploration stage” (开源开放的基础设施组织和保障机制仍处于探索阶段). This is the sharpest self-diagnostic statement in the piece, and its placement on the party-theoretical-flagship surface gives it the highest institutional weight. From an institutional economics standpoint, this celebratory-formulation / self-diagnostic-qualifier pairing is a structural finding: the same party-theoretical-flagship surface that proposes a national-level AI-research open-source community simultaneously documents that the institutional mechanism to build and operate such a community is not yet in place — the same pattern this series’ September 9 briefing’s CCTV codification documented at the state-media-national-economy surface (four explicit disclaimers alongside the celebratory framing), now documented at the party-theoretical-flagship surface as a research-recommendation article. The institutional-economics reading is that the party-theoretical-flagship surface is not ambiguous about the boundary between the proposal and the current state of institutional capacity — the boundary is explicitly stated, in the same article, in the same section.

Sources:

Qiushi / Qiushi WeChat (August 31, 2026) — Chen Kaihua, “How AI Leads the Transformation of the Research Paradigm,” Qiushi 2026/17
Context: August 13 briefing — Chatham House’s open-weight vs. closed-weight institutional-model comparison; August 28 briefing — MIIT Liu Yulin’s three-pillar open-source strategy; August 31 briefing — Xie兰’s 中国社会科学报 open-source-as-statecraft framing; September 5 briefing — Ren Xudong’s platform-neutrality-as-priced-public-good essay; September 7 briefing — Huawei’s AAIF entry as a first Chinese enterprise; September 9 briefing — CCTV Finance’s open-source-as-national-economy codification; September 10 briefing — PyTorch Foundation’s device-agnostic PyTorch keynote (Cambricon Wei Li)
Institutional-baseline reference: White House Office of Science and Technology Policy, “Science: The New Golden Age” (July 2026), naming the Genesis Plan as the flagship action of U.S. AI-enabled science research

2. Zhishashe (智本社) (September 9, 2026, 9:22 AM; revised) — Qinghe (清和), editor-in-chief of Zhishashe, “中国大厂：两边通吃” (“Chinese Big Cos: Eats Both Sides”)

Zhishashe is an independent Chinese institutional-economics analysis surface, and Qinghe’s September 9 article is a ~7,000-word comparative analysis of Chinese- versus U.S.-listed companies’ commercial-credit behavior (commercial credit = supplier-and-customer-payment-terms financing), explicitly framed through the new-institutional-economics lens.

The article’s specific institutional-economics observations, extracted from the Chinese-language source:

The commercial-credit-divergence baseline. As of 2025, 5,382 A-share non-financial listed companies report 15.30 trillion yuan of payables against 9.97 trillion yuan of receivables — a net surplus of 5.33 trillion yuan, or 8.4% of revenue. The U.S.-side comparison sample (1,523 U.S. non-financial listed companies) reports receivables 1.32 trillion dollars against payables 1.18 trillion dollars — a net deficit of approximately 0.14 trillion dollars, or −1.2% of revenue. The Chinese side is systematically net occupying supply-chain financing; the U.S. side is systematically net supplying supply-chain financing.
The payment-turnover-days gap. From 2017 to 2025, the payment turnover days of a fixed A-share non-financial sample (2,660 firms) rose from 83.9 days to 106.7 days (a 22.8-day increase). The U.S. non-financial fixed sample (533 firms) rose from 66.1 days to 70.9 days (a 4.8-day increase). The China-U.S. gap widened from 17.9 days to 35.8 days over the same period.
The top-20 Chinese manufacturer sample. For the top-20 fixed Chinese manufacturing sample, 2017-2025 net payable surplus rose from 199.7 billion yuan to 654.1 billion yuan, and the net-surplus-to-revenue ratio rose from 7.0% to 13.1% — equivalent to 266.3% of attributable-net-profit, up from 124.9% in 2017. The 13-firm U.S. comparison sample shows a net-surplus-to-revenue ratio of 4.5% in 2025 — the Chinese figure is 2.9x the U.S. figure.
The “shadow bank” reading. The article’s sharpest institutional-economics reading is that Chinese large manufacturers are functioning as shadow banks (影子银行) with quasi-financial characteristics — they are drawing interest income from their net commercial-credit occupation, not from their operating activity. The 20-firm Chinese manufacturing sample’s interest income rose from 9.88 billion yuan (2017) to 36.82 billion yuan (2025), and its interest expense rose from 25.97 billion yuan to 38.01 billion yuan; the interest-income-to-interest-expense coverage ratio rose from 38.0% to 96.9% over the same period. The private/mixed-ownership group’s coverage ratio exceeded 100% in 2021 and reached 115.4% in 2025 — the interest income is fully covering the interest expense.
The counterfactual debt-substitution calculation. The article documents that if the 2025 Chinese 654.1-billion-yuan net commercial-credit surplus were entirely replaced by new debt at 3-5% interest rates (holding interest income constant), the coverage ratio would drop from 96.9% to 52.1-63.9%. Therefore, the net commercial-credit surplus explains most of the coverage-ratio increase.
The U.S.-side contrast. The five-firm U.S. manufacturing comparison sample (Tesla, NVIDIA, Intel, Honeywell, HP) shows interest income rising from 732 million dollars (FY2017) to 4.931 billion dollars (FY2025) against interest expense rising from 1.800 billion dollars to 3.450 billion dollars, with a coverage ratio rising from 40.6% to 142.9%. But the same five firms’ 2025 net commercial-credit surplus is only 9.136 billion dollars — if entirely substituted by debt at 5%, the coverage ratio would drop only 12 percentage points, explaining only about 10% of the coverage-ratio change. The U.S.-side interest-income growth is primarily driven by the interest-rate environment, not by commercial-credit occupation.
The risk-mechanism reading. The article’s risk-diagnosis section explicitly frames the two-sided-credit occupation (national credit + commercial credit) as a source of macro-risk amplification — with specific reference to Evergrande’s 三道红线 (three red lines) property-finance case, and the risk that new-energy vehicle manufacturers may become “the next Evergrande” (新能源车企可能成为下一个"恒大").

Institutional significance: Qinghe’s Zhishashe article is the first documented application of the Chinese-side institutional-economics shadow-bank frame to the domestic large-manufacturer commercial-credit layer — the same shadow-bank diagnostic that has been applied externally to the U.S.-side large-manufacturer layer (the interest-income growth is explained by the interest-rate environment, not commercial-credit occupation), and now documented internally to the Chinese-side large-manufacturer layer (the interest-income growth is explained by commercial-credit occupation, not the interest-rate environment).

From an institutional economics perspective, the Qinghe Zhishashe article matters on four axes:

First, the “shadow bank” reading of Chinese large manufacturers is institutionally the sharpest institutional-form observation of this cycle — the Chinese large manufacturer is being documented as an institutional-form object that is performing quasi-financial functions (drawing interest income from commercial-credit occupation rather than from operating activity) on the domestic-compliance layer. The article’s explicit statement that “Chinese large manufacturers are shadow banks with quasi-financial characteristics” (中国大型制造企业是"影子银行"，具备准金融特征) is a structurally new institutional-form object on the domestic-compliance layer. From an institutional economics standpoint, this shadow-bank-reading / commercial-credit-occupation pairing is a structural finding: the Chinese large manufacturer is not just a manufacturing entity but a quasi-financial entity on the commercial-credit layer — its interest-income growth is being explained by its net commercial-credit occupation, not by its operating activity. The institutional-economics reading is that the domestic-manufacturing layer is producing an institutional-form object (the shadow-bank large manufacturer) that this series’ prior twenty briefings had not documented — a structurally new institutional-form object on the domestic-compliance layer.

Second, the interest-income-to-interest-expense coverage-ratio gap (38% to 97% on the Chinese side vs. 41% to 143% on the U.S. side, driven by different causal mechanisms) is institutionally the sharpest structural-finding of this cycle — the same coverage-ratio growth pattern on two institutional sides, driven by two institutionally different causal mechanisms. The article explicitly documents that the Chinese-side coverage-ratio growth (38% to 97%) is explained by commercial-credit occupation, while the U.S.-side coverage-ratio growth (41% to 143%) is explained by the interest-rate environment. From an institutional economics standpoint, this same-coverage-ratio / different-causal-mechanism pairing is a structural finding: the two institutional sides are producing the same quantitative outcome (rising coverage ratio) through institutionally different causal mechanisms (commercial-credit occupation vs. interest-rate environment), and this causal-mechanism difference is an institutionally structural feature of the domestic-compliance layer. The institutional-economics reading is that the domestic-compliance layer is not producing a coverage-ratio outcome but is producing a coverage-ratio outcome with a specific causal-mechanism signature — a structurally new institutional-form object.

Third, the Evergrande-reference / next-Evergrande-risk pairing is institutionally the sharpest risk-diagnosis framing in the article — the article explicitly names the Evergrande 三道红线 case as the institutional precedent for the two-sided-credit-occupation macro-risk, and explicitly identifies new-energy vehicle manufacturers as the potential next-case. The article’s risk-mechanism section explicitly references Evergrande’s 2020 annual report and the 2024 CSRC administrative-penalty decision (2019, 2020 financial-report false statement), and explicitly identifies new-energy vehicle manufacturers as the potential next-case. From an institutional economics standpoint, this Evergrande-reference / next-Evergrande-risk pairing is a structural finding: the two-sided-credit-occupation macro-risk is not a hypothetical risk but a risk with a documented institutional precedent and a named potential-target sector. The institutional-economics reading is that the two-sided-credit-occupation macro-risk is a structurally-documented institutional-form object on the domestic-compliance layer — with an explicit institutional precedent (Evergrande) and an explicit named-target sector (new-energy vehicle manufacturers).

Fourth, the two-sided-credit-occupation diagnostic pairs directly with this series’ August 26 Heilan “going-global” institutional critique — the same institutional-economics frame applied to two institutionally different surfaces (domestic commercial credit vs. export-level competitiveness), and producing two institutionally parallel findings. The August 26 briefing documented Heilan’s institutional critique of the “going-global” (出海) framing — with the specific Heilan data that overseas gross margin (11.21%) is much lower than domestic gross margin (28.22%), and that overseas revenue is growing (+4.56%) while overseas operating cash flow is collapsing (−39.21%). The Qinghe Zhishashe article applies the same institutional-economics frame (credit-and-cash-flow diagnostic) to a different institutional surface (domestic commercial credit) and produces a parallel finding (domestic large manufacturers are drawing quasi-financial income from commercial-credit occupation, not from operating activity). From an institutional economics standpoint, this Heilan-going-global-critique / Qinghe-shadow-bank-diagnostic pairing is a structural finding: the same institutional-economics frame is producing institutionally parallel findings on two different institutional surfaces — the export-level competitiveness surface (August 26 briefing, Heilan) and the domestic commercial-credit surface (today, Qinghe). The institutional-economics reading is that the institutional-economics frame is a cross-surface analytical tool — the same frame, applied to two institutional surfaces, produces two institutionally parallel findings.

Sources:

Zhishashe (智本社) (September 9, 2026, 9:22 AM; revised) — Qinghe (清和), editor-in-chief, “中国大厂：两边通吃”
Institutional-baseline references: Evergrande 2020 annual report; CSRC administrative-penalty decision against Evergrande Real Estate Group Co., Ltd. and related responsible persons (2024); People’s Bank of China, Fourth Quarter 2024 Chinese Monetary Policy Implementation Report (February 2025); State Council, “Regulations on Guaranteeing Payments to Small and Medium Enterprises” (revised 2025-06-01)
Context: August 26 briefing — Heilan’s “going-global” institutional critique (overseas gross margin 11.21% vs. domestic 28.22%; overseas operating cash flow −39.21%)

🏗️ Competitive Scorecard — Xinchuang-Lutoushe Half-Year Scorecard for 36 Vendors Across 8 Segments: Hardware Turning Profitable, Integration Vendors Collapsing, Huawei “Full-Stack” Eating the Market

3. Xinchuang-Lutoushe (信创露透社) (September 1, 2026, 5:09 AM) — Tou Ge (透哥), “信创8大赛道36家公司上半年业绩对比：谁净赚29亿，谁血亏6亿，谁利润暴跌280倍” (“Xinchuang 8-Segment 36-Company Half-Year Scorecard: Who Netted 2.9B, Who Lost 600M, Whose Profit Fell 280x”)

Xinchuang-Lutoushe is a Chinese-language WeChat-monitor surface dedicated to the Xinchuang (信创, “information technology application innovation,” the domestic-substitution-for-critical-IT-infrastructure) market. The article is a comparative scorecard of 36 Xinchuang vendors across 8 segments (CPU, GPU, Operating System, Complete Machines, Database, Middleware, Network Security, Integration Vendor) based on their first-half 2026 financial disclosures.

The article’s specific institutional-economics observations, extracted from the Chinese-language source:

Huawei’s full-stack dominance. Huawei’s first-half 2026 revenue of 467.819 billion yuan is a company-record high (+9.5% YoY), but net profit fell 36% to 23.809 billion yuan, with R&D spending of 121.38 billion yuan (+25.2% YoY, now 25.2% of revenue). The article’s characterization — “wherever Huawei’s full stack goes, there is ‘grassless BS’” (华为"全家桶"所到之处，几乎是"寸草BS" — a pun on “寸草不生” / “no grass grows”) — documents that Huawei’s full-stack (CPU/GPU/OS/server/desktop/laptop/network-security/database/middleware/cloud) has displaced incumbent domestic vendors across the Xinchuang market.
CPU segment: Hygon dominant, Feiteng and Loongson collapsing. Hygon’s H1 revenue of 9.1 billion yuan is 12.16x Feiteng’s revenue and 33.5x Loongson’s; its net profit of 1.8 billion yuan exceeds the combined revenue of Feiteng and Loongson. Feiteng’s H1 loss of 338 million yuan widened 48% YoY. Zhaoxin (Zhongxin) filed for a STAR-Market IPO in June 2024 but has not yet listed.
GPU segment: The first domestic AI-accelerator profitability inflection. The article’s sharpest finding is that “the picture has changed this year” (今年画风变了) — Cambricon (寒武纪) netted 2.3 billion yuan in H1, and Muxi (沐曦) turned profitable with 600 million yuan net profit, while Moore Threads (摩尔线程) sharply narrowed its losses and Enflame (燧原) launched its STAR-Market IPO subscription. Biren (壁仞) grew revenue ~20x but still lost over 300 million yuan; Enflame grew revenue ~3x but still lost over 600 million yuan. Jingjia Micro (景嘉微) — core team from the National University of Defense Technology, with 70%+ market share in the military graphics-display-control module segment — is called out as an under-watched military-industrial player.
Operating System segment: Kirin leading, Tongxin turning profitable, HarmonyOS entering. Kirin Software continues to lead with H1 revenue of 700 million yuan and net profit of 150 million yuan. Tongxin (UOS) finally turned profitable, and Kirin Xin’an is close to break-even. The article explicitly notes that “HarmonyOS has already obtained the national-test ‘pass’ (国测通行证), and with increasing adaptation effort, its future erosion of Kirin’s market is inevitable” (华为鸿蒙已经拿到了国测"通行证"，随着适配力度的加大，未来对麒麟的市场蚕食是必然的). Fangde (中科方德, one of the top three desktop OS vendors, with a chairman position in Jiashu Investment’s September consolidation documented in the September 9 briefing) is not listed as a separate row because it is not a listed company.
Complete-Machine segment: Inspur dominant, SoftCom in structural decline. Inspur (浪潮) netted 2.9 billion yuan in H1, more than doubling YoY. SoftCom Power (软通动力) is the only losing complete-machine vendor, with its loss widening YoY, following its January 2024 acquisition of Tsinghua Tongfang’s computer business. The article attributes SoftCom’s loss to the ongoing “entanglement with the original Tongfang team” (与原同方团队的纠葛).
Database segment: Dameng (达梦) leading the listed vendors, but not alone. Dameng leads the listed-vendor comparison with H1 revenue of 710 million yuan and net profit of 220 million yuan, but the article explicitly warns that “this does not mean Dameng can be complacent,” because Huawei GaussDB, Alibaba OceanBase, TencentDB, and ZTE GoldenDB are all gradually entering Xinchuang projects without separately disclosing their database-segment financials. Haisheng (神舟通用) revenue fell 70% with a shift from profit to loss; Dianke Jinchen (电科金仓) barely profitable at 7.51 million yuan.
Middleware segment: Three vendors, all losing. Baolan, Puyuan, and Zhongchuang are all losing. Zhongchuang’s loss widened 150%. Dongfang Tong was delisted in January 2025 for financial fraud and has no H1 2026 data.
Network Security segment: Sangfor dominant, Tianrongxin collapsing. Sangfor’s cloud-computing + AI segment is now 55.34% of its revenue at 2.212 billion yuan, with network security (39.69% of revenue at 1.587 billion yuan) no longer the primary growth driver. Qi’anxin (奇安信) revenue fell 15% and its H1 loss narrowed from 770 million yuan. Tianrongxin is the only vendor with a “double decline” — a 270 million yuan loss equal to half its H1 revenue.
Integration Vendor segment: Only Donghua Software profitable, Taiji collapsing. Only Donghua Software (华为钻石合作伙伴 / Huawei Diamond Partner) netted 200+ million yuan; the other four (Neusoft, Shenzhou Information, China Software, Taiji) all lost. Taiji’s profit fell 281x — its Q1 2025 loss widened from a projected 330-495 million yuan (disclosed January 31, 2026) to 763 million yuan (disclosed April 30, 2026), with its market cap falling 72% from its April 2023 peak to the current 9 billion yuan.

Institutional significance: The Xinchuang-Lutoushe scorecard is the first documented half-year-industrial-level institutional-form observation on the Xinchuang layer — a structurally new institutional-form observation that this series had not yet produced, with three institutionally parallel findings: the hardware layer is turning profitable, the integration-vendor layer is collapsing, and the OS layer is on the verge of a Huawei-driven market restructuring.

From an institutional economics perspective, the Xinchuang-Lutoushe scorecard matters on four axes:

First, the hardware-layer profitability inflection (Cambricon 2.3B net profit; Muxi 600M net profit) is institutionally the sharpest institutional-form observation of this cycle — it is a first-documented profitability inflection at the domestic-AI-accelerator layer, and it pairs directly with this series’ September 10 briefing’s DeepSeek × Ascend 950DT 160,000-chip inference-only decoupling milestone. The September 10 briefing documented the DeepSeek × Ascend 950DT order as a domestic-silicon-decoupling event at the inference layer only (training remains on NVIDIA). The Xinchuang-Lutoushe scorecard’s Cambricon 2.3B net profit and Muxi 600M net-profit figures are the empirical instantiation of the same inference-layer decoupling process on the domestic-AI-accelerator-vendor profitability layer. From an institutional economics standpoint, this DeepSeek-Ascend-inference-decoupling / Cambricon-Muxi-profitability-inflection pairing is a structural finding: the domestic-AI-accelerator decoupling process is producing cash flow at the inference layer, and this cash-flow production is being documented on two institutional surfaces simultaneously — the DeepSeek-procurement surface (September 10 briefing) and the domestic-AI-accelerator-vendor-profitability surface (today). The institutional-economics reading is that the domestic-AI-accelerator decoupling process is producing a structurally new institutional-form object — cash flow at the inference layer, not just procurement commitments at the training layer.

Second, the integration-vendor collapse (Taiji’s 281x profit decline, 72% market-cap decline) is institutionally the sharpest institutional-form observation of the entire Xinchuang scorecard — the integration-vendor layer is producing the sharpest loss concentration in the Xinchuang market, and this loss concentration is a structurally new institutional-form object on the domestic-substitution layer. The article explicitly documents that Taiji’s profit fell 281x and its market cap fell 72% from its April 2023 peak. The Q1 2025 loss widening (from a projected 330-495 million yuan to an actual 763 million yuan) is institutionally significant — it is an institutional-form observation about the gap between a listed-company’s projected loss and its actual loss, and the investor-rights-maintenance wave (投资者维权索赔热潮) that followed. From an institutional economics standpoint, this projected-loss / actual-loss-gap pairing is a structural finding: the domestic-substitution layer is producing institutional-form objects (loss-concentration events at the integration-vendor layer) that are not being documented on the state-media-national-economy surface (CCTV, September 9 briefing) but on the WeChat-monitor Xinchuang surface. The institutional-economics reading is that the domestic-substitution layer is producing institutional-form objects that this series’ prior twenty briefings had not documented — loss concentration at the integration-vendor layer.

Third, the Huawei full-stack market-eating observation pairs with this series’ September 9 briefing’s Jiashu-统信-方德 chairman-consolidation observation — two institutionally parallel observations about consolidation dynamics in the same Xinchuang market, but at different layers (Huawei at the horizontal full-stack layer, Jiashu at the vertical desktop-OS layer). The September 9 briefing documented Jiashu Investment’s chairman-consolidation across 统信 (UOS) and 方德 (Fangde) at the vertical desktop-OS layer. The Xinchuang-Lutoushe scorecard documents Huawei’s full-stack market eating at the horizontal full-stack layer. From an institutional economics standpoint, this Huawei-horizontal-consolidation / Jiashu-vertical-consolidation pairing is a structural finding: the Xinchuang market is being consolidated at two institutional layers simultaneously — the horizontal full-stack layer (Huawei) and the vertical desktop-OS layer (Jiashu) — with the two consolidation processes operating in parallel in the same market. The institutional-economics reading is that the Xinchuang market is being consolidated as a structurally new institutional-form object on the domestic-substitution layer — a structurally new consolidation pattern across two institutional layers.

Fourth, the HarmonyOS-entering-the-OS-layer observation pairs directly with the Jiashu-统信-方德 chairman-consolidation observation — two institutionally parallel observations about the OS layer’s structural instability, but at different scales (Jiashu at the private-equity-consolidation scale, HarmonyOS at the state-enterprise-ecosystem-entry scale). The September 9 briefing documented Jiashu Investment’s chairman-consolidation across 统信 (UOS) and 方德 (Fangde). The Xinchuang-Lutoushe scorecard documents that “HarmonyOS has already obtained the national-test ‘pass’ (国测通行证), and with increasing adaptation effort, its future erosion of Kirin’s market is inevitable.” From an institutional economics standpoint, this Jiashu-private-equity-consolidation / HarmonyOS-state-enterprise-entry pairing is a structural finding: the OS layer’s structural instability is being documented at two institutional scales simultaneously — the private-equity-consolidation scale (Jiashu) and the state-enterprise-ecosystem-entry scale (HarmonyOS) — with the two instability processes operating in parallel in the same layer. The institutional-economics reading is that the OS layer is a structurally unstable institutional-form object on the domestic-substitution layer — with consolidation dynamics operating at both the private-equity scale and the state-enterprise-ecosystem-entry scale.

Sources:

Xinchuang-Lutoushe (信创露透社) (September 1, 2026, 5:09 AM) — Tou Ge (透哥), “信创8大赛道36家公司上半年业绩对比”
Institutional-baseline references: Huawei H1 2026 financial report (revenue 467.819 billion yuan, +9.5% YoY; net profit 23.809 billion yuan, −36% YoY; R&D 121.38 billion yuan, +25.2% YoY); Taiji Co., Ltd. Q1 2025 earnings announcement (projected loss 330-495 million yuan, actual loss 763 million yuan); Taichu Co., Ltd. market-cap peak (April 2023) and current market cap (approximately 9 billion yuan, −72%)
Context: September 9 briefing — Jiashu Investment’s chairman-consolidation across 统信 (UOS) and 方德 (Fangde); September 10 briefing — DeepSeek × Ascend 950DT 160,000-chip inference-only decoupling milestone; September 10 briefing — PyTorch Foundation’s Cambricon device-agnostic PyTorch keynote

🔍 Commentary

Three institutional objects on three different surfaces — one ideological (Qiushi), one diagnostic (Zhishashe), one competitive (Xinchuang-Lutoushe) — all converging on the same unsolved institutional object: the institutional identity of Chinese open source and its surrounding economy.

This cycle’s briefing documents, on three different institutional surfaces, three related institutional objects on the same unsolved institutional object that the prior twenty briefings have been documenting: the institutional identity of Chinese open source and its surrounding economy across institutional borders.

The Chen Kaihua Qiushi article (August 31, published in Qiushi 2026/17) documents the institutional identity of Chinese open source at the party-theoretical-flagship layer — for the first time on a Chinese party-theoretical surface, placing 开源社区 (open-source community) alongside data platforms and AI research agents as a formal 科研主体 (research subject), and proposing “a national-level open-source community for AI-driven research” with an explicit mechanism combining “code contribution and compute incentive.”
The Qinghe Zhishashe article (September 9) documents the institutional identity of Chinese open source at the independent-institutional-economics-analysis layer — the first documented application of the shadow-bank frame to the domestic large-manufacturer commercial-credit layer, with the 5.33-trillion-yuan net payable-receivable surplus on the A-share side compared to the U.S.-side 0.14-trillion-dollar net deficit, and the interest-income-to-interest-expense coverage ratio rising from 38% to 97% over 2017-2025.
The Xinchuang-Lutoushe scorecard (September 1) documents the institutional identity of Chinese open source at the WeChat-monitor-competitive-scorecard layer — the first documented half-year-industrial-level institutional-form observation on the Xinchuang layer, with the hardware-layer profitability inflection (Cambricon 2.3B net profit; Muxi 600M net profit), the integration-vendor collapse (Taiji 281x profit decline, 72% market-cap decline), and the OS-layer structural instability (HarmonyOS entering with the national-test pass).

The three events together document that the institutional identity of Chinese open source and its surrounding economy across institutional borders — the same unsolved institutional object that the prior twenty briefings have been documenting — is now being formalized on three new surfaces: as a party-theoretical-flagship formulation (Chen Kaihua in Qiushi), as an independent institutional-economics diagnostic (Qinghe in Zhishashe), and as a WeChat-monitor-competitive scorecard (Xinchuang-Lutoushe).

One structural pattern across all three surfaces: the same institutional frame (open source and its surrounding economy) is being operationalized at three institutionally different layers simultaneously.

The three events together document a structurally new institutional pattern that the prior twenty briefings had not yet observed: the same institutional frame is being operationalized at three institutionally different layers simultaneously — the party-theoretical-flagship layer (Chen Kaihua in Qiushi), the independent-institutional-economics-analysis layer (Qinghe in Zhishashe), and the WeChat-monitor-competitive-scorecard layer (Xinchuang-Lutoushe).

From an institutional economics standpoint, this three-surface-simultaneous-operationalization / institutional-identity-of-chinese-open-source pairing is a structural finding: the institutional identity of Chinese open source and its surrounding economy is not being operationalized layer-by-layer sequentially but across three institutionally different layers simultaneously — and the simultaneous operationalization is the institutional-form object that the prior twenty briefings have been tracking from multiple angles. The institutional-economics reading is that the Chinese open-source institutional architecture has crossed a three-surface-simultaneous-operationalization threshold — the same institutional frame is being operationalized at the party-theoretical-flagship layer, the independent-institutional-economics-analysis layer, and the WeChat-monitor-competitive-scorecard layer simultaneously, and this three-surface-simultaneous-operationalization pattern is a structurally new institutional-form object.

One structural risk across all three surfaces: the Chen Kaihua “still in the exploration stage” self-diagnostic pairs with the Xinchuang-Lutoushe Taiji 281x-collapse — the same institutional architecture is proposing a national-level AI-research open-source community while documenting that its integration-vendor layer is producing the sharpest loss concentration in the Xinchuang market.

The Chen Kaihua article’s explicit statement that “the organization and safeguarding mechanism of open-source and open infrastructure is still in the exploration stage” (开源开放的基础设施组织和保障机制仍处于探索阶段) is the sharpest self-diagnostic statement in the party-theoretical-flagship surface — the same party-theoretical-flagship surface that proposes a national-level AI-research open-source community simultaneously documents that the institutional mechanism to build and operate such a community is not yet in place.

The Xinchuang-Lutoushe scorecard’s Taiji 281x-collapse (projected loss 330-495 million yuan → actual loss 763 million yuan → 72% market-cap decline from April 2023 peak) is the sharpest institutional-form observation on the WeChat-monitor-competitive-scorecard surface — the same WeChat-monitor surface that documents the hardware-layer profitability inflection (Cambricon 2.3B net profit) simultaneously documents the integration-vendor-layer collapse (Taiji 281x profit decline).

Reading all three stories together, the structural risk is that the three institutional objects (Chen Kaihua Qiushi, Qinghe Zhishashe, Xinchuang-Lutoushe) are not obviously coherent with each other:

The Chen Kaihua Qiushi article proposes a national-level AI-research open-source community with a code-contribution-compute-incentive mechanism, but explicitly documents that the institutional mechanism to build and operate such a community is “still in the exploration stage.”
The Qinghe Zhishashe article documents that Chinese large manufacturers are shadow banks with quasi-financial characteristics, but the two-sided-credit-occupation macro-risk (Evergrande-reference; next-Evergrande new-energy vehicle risk) is not being operationalized by the same party-theoretical-flagship surface that proposes the national-level AI-research open-source community.
The Xinchuang-Lutoushe scorecard documents that the Xinchuang market is producing cash flow at the hardware layer and loss concentration at the integration-vendor layer, but this same Xinchuang market is the layer that the Chen Kaihua Qiushi article’s proposed national-level AI-research open-source community would need to operationalize on.

The institutional-economics question — which this cycle’s briefing documents but does not answer — is whether the Chinese institutional architecture can hold the three institutional objects simultaneously without one converting the other. If the Chen Kaihua Qiushi article’s national-level AI-research open-source community proposal converts the “still in the exploration stage” self-diagnostic into a formalized institutional-mechanism, the proposal becomes an institutional-form object that the party-theoretical-flagship surface is now obligated to operationalize; if the Qinghe Zhishashe shadow-bank diagnostic converts the two-sided-credit-occupation macro-risk into a formal-state response, the shadow-bank diagnostic becomes a state-crisis-response object rather than an independent institutional-economics diagnostic; if the Xinchuang-Lutoushe Taiji 281x-collapse converts the integration-vendor layer’s loss concentration into a state-directive-industrial-capacity question (MIIT Liu Yulin, August 28 briefing), the Xinchuang market’s structural instability becomes a national-industrial-policy response object.

One perspective, not a verdict.

All three stories — the Chen Kaihua Qiushi party-theoretical-flagship formulation, the Qinghe Zhishashe independent-institutional-economics diagnostic, and the Xinchuang-Lutoushe WeChat-monitor-competitive scorecard — are best read as observations of institutional movement in progress, not as verdicts on institutional direction. The Chen Kaihua Qiushi article does not guarantee that the national-level AI-research open-source community proposal will be operationalized; the Qinghe Zhishashe article does not guarantee that the two-sided-credit-occupation macro-risk will trigger a state response; the Xinchuang-Lutoushe scorecard does not guarantee that the hardware-layer profitability inflection will extend to the training-layer or that the integration-vendor-layer collapse will trigger an integration-vendor-layer state response. What this cycle’s briefing documents is that the institutional identity of Chinese open source and its surrounding economy has crossed a three-surface-simultaneous-operationalization threshold — from a two-surface-simultaneous-formalization pattern (September 10 briefing’s PyTorch-Foundation / DeepSeek-Ascend / Shanghai municipal-fund simultaneous formalization) to a three-surface-simultaneous-operationalization pattern — and that the three new institutional-form surfaces (Chen Kaihua Qiushi party-theoretical-flagship, Qinghe Zhishashe independent-institutional-economics, Xinchuang-Lutoushe WeChat-monitor-competitive) sit on the same unsolved institutional object: the institutional identity of Chinese open source and its surrounding economy across institutional borders.

Editorial note on perspective: This briefing presents one institutional-economics reading of Chinese open-source developments, not a verdict. The “institution” in these stories — the Chen Kaihua Qiushi party-theoretical-flagship formulation, the Qinghe Zhishashe shadow-bank diagnostic, the Xinchuang-Lutoushe competitive scorecard — is treated as an object of observation, not a target of critique. The Great Divergence 2.0 framework (FLOSS vs. State-Chartered Codebase vs. Intranet Shared Source vs. Cyber-Estate) is a lens, not a universal answer. One perspective, not a verdict.