Seagate says SSDs now cost 6.5 times more per terabyte than disk drives, and the company has built its first 2026 Data Infrastructure Readiness Report around that gap. The message to AI-building organizations is blunt: don't put all your AI data eggs in one flash basket.

The report, commissioned by Seagate, argues that AI is turning storage from a back-office cost line into strategic business infrastructure. Data volumes are growing, and with them the value organizations can extract — but only if the data stays preserved, accessible and affordable to reuse over years, not quarters.

Melyssa Banda, SVP of Edge Storage at Seagate, put it this way: "AI is reshaping the way organizations plan, build and operate infrastructure. As data volumes grow, so does the value organizations can derive from the data. They need data infrastructure that helps them preserve, access and use more of that data over time."

She frames the pitch under a term the report leans on heavily: Sustainable Scaling.

What is 'Sustainable Scaling'?

Seagate defines it as "increasing AI capacity and business value while continuously improving the efficiency of the infrastructure that supports it." In practice, that means scaling infrastructure decisions with rising AI data demands while improving efficiency, sustainability and long-term data value.

Strip the marketing varnish and it means one thing: buy more disk.

The report itself is more diplomatic. "Organizations must plan for both more data and more value embedded in that data," it states. "A short-term approach focused only on immediate capacity can leave information fragmented, inaccessible or expensive to reuse. A longer-ranging view can connect retention, access and governance decisions to the future value the organization expects AI to create."

The gap it's describing is real. AI pipelines generate enormous intermediate datasets — training corpora, embeddings, checkpoints, inference logs — that must be retained for compliance, retraining or audit long after the hot workload moves on. Where that data sits determines whether reuse is cheap or punishing.

Why one architecture doesn't fit the AI lifecycle

The report's sharpest point is architectural. "Rather than applying a single architecture across the entire data lifecycle, organizations can design systems that balance performance, capacity, efficiency and long-term value according to workload requirements," it says. "This creates a more practical framework for scaling AI while controlling infrastructure costs and resource demands."

That "single architecture" is a stand-in for SSDs. The report's implication, never quite spelled out, is that all-flash can't do it all — and that disk drives improve efficiency, sustainability and long-term data value for everything that doesn't need nanosecond latency.

With flash carrying a 6.5x price premium per terabyte, Seagate is knocking at a fairly open door. Most organizations have already converged on a tiered consensus: hot AI data lives on flash, warm data lives on disk, and cold data heads to tape — either on-premises or in the public cloud.

Is this self-serving or just obvious?

Both, frankly. Seagate sells disk drives, and a vendor-commissioned report recommending disk drives should be read with that in mind. But the economics do the arguing anyway.

At 6.5x the cost per terabyte, storing petabyte-scale AI datasets entirely on flash would consume budgets that could otherwise fund compute — the resource AI teams actually fight over. The report's contribution is less the recommendation than the framing: treat retention, access and governance decisions as part of AI strategy, not as janitorial work bolted on afterward.

Banda's second point carries the argument: "Sustainable Scaling is about making those infrastructure decisions more efficient, more durable and more connected to long-term data value."

The report is available through Seagate's website. Whether the industry adopts the "Sustainable Scaling" vocabulary or not, the tiering logic behind it — flash for speed, disk for scale, tape for archive — is already the de facto blueprint, and Seagate will keep pressing the 6.5x number as AI data lakes grow faster than flash prices fall.