SourceTechNode•Stalecollected in 15h
China Sets 9,800 EFLOPS AI Target
#compute-capacity#data-centers#ai-infrastructurechina-intelligent-computing-infrastructurechinaministry of industry and information technology
China’s infrastructure target signals where AI compute capacity and supply-chain demand may surge next.
30-Second TL;DR
What Changed
The target is 9,800 EFLOPS of intelligent computing capacity by 2030.
Why It Matters
The investment could significantly expand China’s domestic capacity for training and serving large AI models. It may also intensify global competition for GPUs, networking equipment, power, and data-center construction.
What To Do Next
Map your model workloads to expected GPU, networking, and power requirements before expanding into China or other high-growth AI regions.
Who should care:Enterprise & Security Teams
Key Points
- •The target is 9,800 EFLOPS of intelligent computing capacity by 2030.
- •China plans RMB 3.8 trillion in information infrastructure investment from 2026 to 2030.
- •The plan calls for computing clusters with capacity of 10,000 or more.
Deep Insight
Background and context from public sources — not the original article. 9 sources cited.
Enhanced Key Takeaways
- •China's intelligent computing capacity surged 177% year-over-year to 2,185 EFLOPS by June 2026 and rose to approximately 2,450 EFLOPS by late July 2026.
- •The 9,800 EFLOPS target for 2030 represents a 6.2-fold expansion over China's 2025 baseline capacity of 1,590 EFLOPS.
- •Beyond 10,000-card sites—of which China already operated 52 by mid-2026—the blueprint outlines the construction of massive megaclusters containing 100,000 or more accelerator cards.
- •Domestic hardware adoption is accelerating to bypass foreign export curbs, marked by JD Cloud's deployment of a 100,000-GPU cluster utilizing Moore Threads' Universal GPUs.
- •The 2030 targets mandate advanced storage expansion to 1,700 exabytes (up from 540 exabytes in 2025) and enforce a strict Power Usage Effectiveness (PUE) cap below 1.2 for new hyperscale facilities.
Technical Deep Dive
- Compute Precision & Performance: Target capacity of 9,800 EFLOPS is measured in half-precision (FP16) compute, up from a 1,590 EFLOPS baseline in 2025.
- Megacluster Topology: Architectural guidelines define deployments scaling from standard 10,000-card clusters to unified megaclusters housing 100,000+ accelerator cards.
- Domestic Silicon Integration: Broad adoption of sovereign hardware architectures, including Moore Threads Universal GPUs integrated into Tier-1 hyperscale cloud environments.
- Storage Infrastructure: Storage targets set to scale from 540 exabytes (2025) to 1,700 exabytes by 2030 to address data I/O bottlenecks in distributed training workloads.
- Efficiency & Thermal Standard: Mandatory sub-1.2 Power Usage Effectiveness (PUE) for all newly constructed large and hyperscale data centers, tightened from the previous 1.25 benchmark.
- Workload Routing Architecture: Workload segregation routing compute-heavy model training to western renewable energy hubs (Inner Mongolia, Ningxia, Guizhou) via the 'East Data, West Computing' national grid.
Future ImplicationsAI analysis grounded in cited sources
Accelerated substitution of Western AI hardware across tier-1 domestic clouds
The national mandate for 100,000-card megaclusters will fast-track production-level deployment of domestic GPUs like Moore Threads across major Chinese cloud providers.
Aggressive relocation of greenfield AI infrastructure to western provinces
The stringent sub-1.2 PUE requirement will make it practically impossible to build new hyperscale AI training clusters in eastern urban centers without direct access to western renewable grids.
Timeline
2022-02
China formally launches the 'East Data, West Computing' national data center initiative
2025-12
China reaches baseline intelligent computing capacity of 1,590 EFLOPS and 540 exabytes of storage
2026-06
National computing capacity reaches 2,185 EFLOPS across 52 operational 10,000-card facilities
2026-07
National Data Administration reports intelligent computing capacity expanding to roughly 2,450 EFLOPS
2026-09
MIIT unveils 15th Five-Year Plan establishing the 9,800 EFLOPS and RMB 3.8 trillion infrastructure targets
- 2022-02China formally launches the 'East Data, West Computing' national data center initiative
- 2025-12China reaches baseline intelligent computing capacity of 1,590 EFLOPS and 540 exabytes of storage
- 2026-06National computing capacity reaches 2,185 EFLOPS across 52 operational 10,000-card facilities
- 2026-07National Data Administration reports intelligent computing capacity expanding to roughly 2,450 EFLOPS
- 2026-09MIIT unveils 15th Five-Year Plan establishing the 9,800 EFLOPS and RMB 3.8 trillion infrastructure targets
Sources (9)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
1China Targets 9800 Eflops of Intelligent Computing Capacity by 2030technode.com2Ws6a9e750be4b06d4aa055cc0achinadaily.com.cn3China 9800 Eflops Compute Plan 2030 2026valueaddvc.com4China Sets 9800 Eflops Intelligent Computing Target by 2030 Plans 62x 17865189gate.com5China Miit 15th Five Year Plan AI Computing Power Target 9800 Eflops 5c76557b En Keyname 2f245983bit.fan6China Targets Fourfold Boost AI Computing Capacity 2030 Major Tech Pushscmp.com7Jd Cloud Moore Threads Plan 100000 GPU Chinese Silicon Clusteraiweekly.co8Miit Plan Targets 9800 Eflops of Intelligent Compute by 2030unite.ai9Miit Targets 9800 Eflops of AI Compute by 2030 532b Planaiweekly.co
Weekly AI Recap
Read this week's curated digest of top AI events →
Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: TechNode ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.