TL;DR — Industrial LLMs scale, county-level digital infrastructure sinks, and three SME digitalization pilots expand: China’s digitalization enters a three-track compound Q4 phase

Q4 of 2026 has opened with three synchronized digitalization accelerations. IDC’s latest data shows that the share of Chinese industrial enterprises piloting or applying LLMs and agents jumped from 9.6% in 2024 to 47.5% in 2025 — a near-fourfold year-on-year increase. MIIT’s “Model-Data Resonance” initiative now covers 20 priority industries including steel, machine tools, automotive, and aerospace. At the same time, “gigabit to every county, 5G to every township” has been completed nationwide — 100% of administrative villages now have broadband and 95%+ have 5G — and the dedicated “Millisecond Computing” campaign now covers 50 prefecture-level cities. Q4 marks the transition from coverage to penetration to county-level sinking across all three tracks.

1. Q4 Indicators Across the Three Tracks

1.1 Industrial LLMs: From pilots to scaled penetration

According to IDC’s 2026 survey, the share of Chinese industrial enterprises that have adopted LLMs and AI agents has risen from 9.6% to 47.5% in just one year. The share of enterprises deploying these technologies across multiple functions (R&D, manufacturing, supply chain) has jumped from 1.7% to 35%. The National Industrial Information Security Development Research Center (NIISDRC) reports in its “Manufacturing Digital-Intelligent Transformation Capability Level (2026)” that the penetration of AI in priority industrial scenarios reached 34.2% in H1 2026, up 5.5 percentage points year-on-year. The nationwide integration-of-informatization-and-industrialization (IIT) level has reached 67.1, entering a coordinated digitalization, networking, and intelligentization phase.

The Industry Think Tank (Yiou) characterizes industrial LLMs as a three-layer architecture: foundation (general industrial LLM), pillar (industry-specific LLM), and pinnacle (scenario-specific LLM). Compact 7B–13B parameter models now run on production-line edge terminals. Multimodal industrial-LLM solutions have grown from 15% of scenarios at end-2025 to 21% in H1 2026.

MIIT and the National Data Bureau have jointly launched the “Model-Data Resonance” initiative targeting 20 priority industries — including steel, machine tools, automotive, and aerospace — with a goal by end-2026 of forming a virtuous “data–model–scenario” cycle. IIM forecasts a 31.2% CAGR for China’s industrial LLM market from 2026 to 2030, with 2026–2027 as a standard-setting period, 2028–2029 as a scaled-diffusion period, and 2030 as the start of self-optimization.

1.2 County-level digital infrastructure: from coverage to millisecond computing

MIIT’s Q1 2026 data show 4.958 million 5G base stations nationwide, 32.01 million 10G PON ports with gigabit capability, 86 cities running 10-gigabit optical-network pilots, and 5G-Advanced coverage in 330 cities. The “gigabit to every county, 5G to every township” milestone is now complete — 100% of administrative villages have broadband and 95%+ have 5G.

The Q4 milestone is the rollout of the dedicated “Millisecond Computing” campaign. MIIT has defined three implementation tasks: millisecond interconnection between computing nodes, millisecond access to computing power, and millisecond reach to applications. The first wave covers 50 prefecture-level cities (counties/districts) across 31 provincial regions, with mid-term and end-of-period assessments planned for end-2026 and end-2027 respectively. This marks the transition of county-level digital infrastructure from “connectivity coverage” to “computing power sinking.”

The Cyberspace Administration of China, the Ministry of Agriculture and Rural Affairs, and MIIT have jointly issued the “Digital Village High-Quality Development Action Plan (2026–2030)”, which sets 2030 targets for village-industry digitalization, village-construction informatization, and village-governance intelligentization.

1.3 SME digitalization: the third pilot wave

MIIT and provincial SME authorities have continued to push SME digitalization pilots, with the first three waves covering industrial LLM applications, 5G factories, lightweight SaaS, and chain-wide digitalization. 5G factories have achieved an average 20.5% product-quality improvement and 18.4% operating-cost reduction. Value-added of computer, communications, and other electronic equipment manufacturing — a category dominated by digitalization suppliers — grew 13.6% year-on-year, the highest among major industrial categories. Q4 marks the assessment phase for the third pilot wave, with expansion expected across more industrial categories and county-level industrial clusters.

2. FAQ — High-Frequency Questions Answered

Q1: Why is Q4 the critical “compounding” period for all three tracks? A: Industrial LLMs move from pilot to scale, county-level infrastructure moves from coverage to millisecond computing, and SME pilots move from wave two to wave three. These three forces synchronize in Q4, and combined with the traditional Q4 manufacturing delivery peak, they form a “scale-up + penetration + sinking” compound phase.

Q2: Is 47.5% industrial LLM penetration overstated? A: No. 47.5% reflects the share of enterprises that have piloted or locally applied industrial LLMs in any function. The NIISDRC figure (34.2%) is the “priority-scenario penetration” rate — a stricter, scenario-specific metric. Both show multi-fold growth versus 2024, and they measure different but complementary concepts.

Q3: What can still sink to the county level after “gigabit to every county”? A: Connectivity is no longer the bottleneck — computing power is. The “Millisecond Computing” campaign sinks compute resources from provincial backbones to prefecture and county levels, enabling industrial internet, AI inference, and real-time control applications that genuinely run in real time at the edge.

Q4: What is the core challenge for SME digitalization pilots? A: The core challenge is not technology but ROI. Honeywell’s 2026 survey of 1,600 manufacturing enterprises shows that 80% have explored AI but only 17% have run it through to the originally planned deployment. The Q4 expansion’s success will be measured by turning “tried it” into “running it daily.”

Q5: How does “Model-Data Resonance” differ from earlier “AI+Manufacturing” policies? A: Earlier policies emphasized capability building; “Model-Data Resonance” targets the closed loop among data, model, and scenario. It binds industrial data circulation, model training, and scenario validation into one action plan, and adds a measurable 2026 milestone.

3. Key Terminology

  • Industrial LLM: A large language/multimodal model (≥7B parameters) pretrained or fine-tuned on industrial data; structured in three layers — foundation, pillar (industry-specific), and pinnacle (scenario-specific).
  • Model-Data Resonance (模数共振): A 2026 MIIT + National Data Bureau initiative covering 20 priority industries, designed to close the data–model–scenario loop.
  • Millisecond Computing (毫秒用算): MIIT’s 2026 special campaign with three implementation tasks — millisecond inter-compute interconnection, millisecond compute access, and millisecond application reach — initially covering 50 cities.
  • IIT (Integration of Informatization and Industrialization): A composite national metric; reached 67.1 in 2026.
  • 5G-A (5G-Advanced): The evolved 5G network; covered 330 cities in Q1 2026.
  • 10G PON: 10-Gigabit passive optical network port; the Q1 2026 national port count reached 32.01 million.
  • 5G Factory: A factory whose production line is end-to-end integrated with 5G, with 20.5% quality improvement and 18.4% cost reduction observed on average.

4. References

Industry reports

  • IDC, “China Industrial Enterprise LLM and Agent Adoption Survey (2026)”
  • National Industrial Information Security Development Research Center, “Manufacturing Digital-Intelligent Transformation Capability Level (2026)”
  • Yiou Think Tank, “2026 China Industrial LLM Industry Research Report”
  • IIM Information, “Global and China Industrial Inspection LLM Application Market Analysis Report (2026)”
  • Honeywell, “2026 Manufacturing AI Adoption Survey (1,600 enterprises)”

Official policies

  • MIIT and seven other ministries, “AI + Manufacturing Special Action Implementation Opinions”
  • MIIT + National Data Bureau, “Model-Data Resonance Action Plan (2026)”
  • MIIT, “City-Wide Millisecond Computing Special Campaign (2026)”
  • Cyberspace Administration of China + Ministry of Agriculture and Rural Affairs + MIIT, “Digital Village High-Quality Development Action Plan (2026–2030)”

Official data

  • MIIT Q1 2026 Industry and Information Technology Development Press Conference (April 21, 2026)
  • Xinhua, “‘Gigabit to Every County, 5G to Every Township’ Fully Achieved (April 2026)”
  • Xinhua, “City-Wide Millisecond Computing Special Campaign Launched (March 2026)”