NAND flash: AI inference boom lifts eSSDs to 48% of shipments

The AI hardware story keeps moving down the stack. This week's data from Counterpoint Research shows the shift from training to inference is now reshaping the memory market — and NAND flash makers are posting record revenue because of it.

Server-grade enterprise SSDs hit 48% of all NAND flash bits shipped in Q2 2026 — nearly double their 26% share a year earlier — and the AI-driven demand surge pushed industry revenue to a record, roughly five times its level in Q2 2025.

According to Counterpoint Research's Q2 Memory & Storage Tracker, the cause is the industry's migration from training to inference. Serving AI models at scale demands storage that can hold KV caches and datasets at high speed and low power, so hyperscalers are buying eSSDs in volumes the market has never seen. The scramble left consumer supply short and pushed consumer SSD prices to record highs. Counterpoint expects server eSSDs to absorb more than half of all NAND bits by the end of the year.

The boom is also reshuffling the supplier pecking order. Samsung kept the shipment crown with a 25% share, down from 32% two years ago as it prioritizes higher-margin DRAM; SK hynix, boosted by subsidiary Solidigm's 40% quarter-over-quarter bit growth, took second at 22%. The milestone belongs to China's YMTC: a 14% share put it in the global top three for the first time, overtaking Kioxia on the back of 22% year-over-year shipment growth and 267-layer 3D NAND in mass production.

The catch — and the part worth watching — is that volume and profit are diverging. YMTC finished only fifth by revenue because its mix is still consumer-heavy, with little of the high-priced data-center eSSD business. Counterpoint's verdict frames the next year: with servers set to take most NAND bits, profitability through 2027 will be decided less by who ships the most, and more by who ships the right mix.

For anyone tracking AI economics, this is the storage layer's coming-out party. GPUs and power dominated the buildout narrative for two years; now the memory market's center of gravity is moving to the inference era, where the cost of keeping models answerable — caches, datasets, fast low-power storage — starts to rival the cost of training them.

What to watch: whether server eSSDs cross the half-market mark by year-end, and whether YMTC's shift toward data-center drives turns its top-three shipment rank into top-three revenue.

Storage is becoming the next AI bottleneck — are you seeing it in your cloud bills? Tell us in the comments.

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Read more

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone