How AI is Accelerating the Shift to Enterprise SSDs and Redefining Data Center Economics

AI applications are causing a surge in global compute power and data generation, especially video and image data used for model training. Huawei predicts that for every additional 1 YB of data created, the volume used for training could increase a thousand‑fold to 400 EB by 2030, shifting the hot‑data proportion in data centers from a 20%‑30%‑50% split (hot‑warm‑cold) to roughly 30%‑70% (hot‑warm‑cold combined), thereby activating the value of previously cold data.

These trends raise performance, power, and cost demands on storage systems, accelerating the migration from traditional mechanical HDDs to enterprise‑grade SSDs. Using Solidigm’s 10 PBAI data‑flow solution as an example, SSD deployments are reduced by 5.2 ×, server counts drop ninefold to 15 units, rack numbers shrink from nine to one, power density rises 2.7 ×, five‑year energy expenses fall 4.4 ×, and total cost of ownership is cut by about 40 %.

IDC forecasts reinforce this shift: by 2028, all‑flash array deployments are expected to exceed 50 % of enterprise storage systems. Global enterprise SSD shipment capacity is projected to grow from 219 EB in 2024 to 517.6 EB in 2028, with unit shipments rising from 262 000 to 324 000 and market size expanding from $26.2 billion to $32.4 billion.