By Oct. 16, Huawei said its AI Data Platform had been put to work in more than 20 key service areas, citing cases in AI-assisted diagnosis and treatment, AI coding and intelligent customer service.
At a data storage summit in Florence, Huawei said a hospital's adoption rate for AI-assisted diagnosis reports rose from 50% to 90%. Throughput for AI coding at an unnamed internet company improved by a factor of 2.2, measured in tokens per second. The company also reported a 16% increase in customer satisfaction through intelligent customer service at an unnamed bank.
Huawei did not identify the organizations or give measurement periods. It supplied no starting throughput figure and did not explain the basis or method for the bank's satisfaction result. The examples track different outcomes, so they do not provide a common performance measure across deployments.
DatacenterDynamics separately reported a roughly 40-fold improvement in diagnostic report generation, nearly doubled token throughput at an AI coding company and a 30% gain in a bank's first-contact resolution rate. Those measures differ from Huawei's adoption, throughput and satisfaction figures, so they cannot be compared directly.
Huawei describes the platform as a knowledge base, a key-value (KV) cache store and a memory bank paired with its Unified Cache Manager (UCM). The company says the system supports AI inference by organizing knowledge retrieval and cached data. It designed the cache management to reduce the need to repeat calculations as a model handles requests.
Huawei tied the coding example to a storage constraint: long, multistep AI coding tasks can generate large volumes of KV-cache data, creating bottlenecks. The company says its platform uses cache scheduling and associative search to address that issue. Its account presents data handling as part of the infrastructure behind enterprise AI workloads.
Huawei announced the platform at MWC Barcelona on March 2. The Florence presentation followed months later and focused on deployment examples and customer results.



