GMIF 2026: 140 Trillion Daily Token Calls Are Reshaping SSD Capabilities and Roles

The 5th GMIF 2026 Global Memory Industry Innovation Summit has sent a clear signal: the explosion of AI inference demand is changing the core logic of the storage industry. China's daily token call volume has surpassed 140 trillion, a figure that directly reflects the scale of large model inference workloads and serves as the fundamental driver of storage market changes.

Morgan Stanley analysts pointed out that, excluding HBM, storage already accounts for nearly half of global cloud data center capital expenditures in 2026, with projections to further rise to 53% in 2027. This means storage is no longer just a supporting player in AI computing power, but is becoming a critical bottleneck determining inference efficiency.

The surge in token call volume has placed entirely new demands on storage architecture. In large-scale inference scenarios, model weight loading, KV Cache read/write operations, and multi-turn dialogue context storage all impose requirements on SSD bandwidth, latency, and capacity that far exceed those of traditional data center workloads. The capability boundaries and division of labor for SSDs are being redrawn.

This trend indicates that competition in AI infrastructure has expanded from pure GPU computing power to the storage layer. Whoever can provide storage solutions tailored to high-concurrency inference scenarios will occupy a more critical position in the AI computing supply chain. Collaboration between storage vendors and AI infrastructure providers will be the focal point of competition in the next phase.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.