Hiroshi Katsuya of HP Japan has provided practical criteria for deciding between cloud and local execution environments through his experience in supporting customer AI implementation and data science operations.
Cloud vs. Local: How to Choose the Right AI Execution Environment, According to HP Japan
This article is a translation. Read the Japanese original
According to Mr. Katsuya, the decision for implementation depends heavily on the nature of the data being handled. He stated that while cloud-based AI is utilized for researching publicly available information, data that cannot be shared externally within business operations is operated entirely in a local environment.
When selecting an AI execution environment, it is necessary to consider business continuity and management structures, rather than just comparing costs. While cloud services carry the risk of usage restrictions due to network environments or changes in service specifications, local environments face challenges such as operational costs related to equipment installation location, power, and maintenance.
Furthermore, as automation via AI agents advances, he pointed out that the bottleneck for processing is not just the GPU, but the entire hardware configuration, including main memory, CPU, and SSD. Designing for the specific use case, scale of use, and operational workflows—including "who approves what and how"—is the key to establishing AI within business operations.
Sources
- Where to run in-house AI? Asking an HP Japan practitioner about the split between cloud and local （ITmedia AI+, 2026-09-16）