Supercomputers/Systems/China Merchants Bank AI Compute Platform
China · operational · mixed
China Merchants Bank AI Compute Platform
Operated by China Merchants Bank .
Phase 1 seed dataset, compiled by hand. These rows were built from public operator, laboratory and vendor sources. A mechanical second-reader pass has since fetched every cited source: 1092 of 1153 systems have a readable citation that names them, and 40 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.
Measured performance
- Rmax
- -
- Rpeak
- -
- Rmax ÷ Rpeak
- -
- Cores
- -
- Power
- -
- Per watt
- - GF/W
Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.
What this machine is made of
One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.
| Role | Part | Supplier at build | Quantity | Confidence | Source |
|---|---|---|---|---|---|
| Integrator | - Nearly 10,000 heterogeneous accelerator cards pooled under one in-house Kubernetes platform (Kueue, KEDA, HAMi, Fluid); built and run by the bank itself | China Merchants Bank | - | Reported | cncf.iobex.co |
| Accelerator | - NVIDIA GPUs (count not stated) within the heterogeneous accelerator pool | NVIDIA | - | Reported | cncf.iobex.co |
| Accelerator | - Multiple Huawei Ascend accelerator models (count not stated) within the heterogeneous accelerator pool | Huawei | - | Reported | cncf.io |
| Accelerator | - Moore Threads accelerators (count not stated) within the heterogeneous accelerator pool | Moore Threads | - | Reported | cncf.io |
The read
China Merchants Bank's private AI compute platform: a unified Kubernetes control plane that pools nearly 10,000 heterogeneous accelerator cards, about 99 percent of the bank's accelerator compute, and shares them across model training, fine-tuning and online inference. The CNCF announcement of 7 September 2026 (the bank won the CNCF End User Case Study Contest) gives the nearly 10,000 card figure, average utilization rising from 35 percent to over 60 percent and inference cost per million tokens falling by more than 60 percent; an independent write-up on bex.co restates the roughly 10,000 card count and 99 percent share. A separate CNCF case study (July 2026) lists the accelerator vendors as NVIDIA, Huawei Ascend, Kunlunxin, Moore Threads, MetaX and PPUs, but no source gives per-vendor card counts, so quantities are left blank. This is a pooled fleet rather than a single fabric-connected training cluster, and neither the site nor the interconnect is documented, so the site is left unlocated. No TOP500 submission exists.
Timeline
No dated events recorded.
Change history
- 2026-10-01 Added China Merchants Bank AI Compute Platform (China Merchants Bank)
Source check
We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.
Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.