VAST Data has launched VAST Amplify, a program selling up to 6x or more effective capacity out of flash a customer already owns rather than flash nobody can currently buy.
The pitch from VAST Data is timed to the NAND supply crunch. Lead times have stretched and allocation windows have tightened, and the company's argument is that scarcity is exposing architectural waste that a loose supply market used to hide, specifically replication-heavy protection models, fragmented data stacks, and performance designs that lean on over-provisioned flash.
On inspection, Amplify is not a product at all. It is a customer engagement in three phases: estate intelligence, in which VAST audits an environment for underutilized SSDs and silo fragmentation; rapid platform qualification, in which it certifies a customer's existing servers and drives for reuse; and capacity reclamation and pooling, which consolidates that hardware into the VAST AI Operating System on its Disaggregated Shared Everything architecture.
The machinery underneath is not new. Platform-level durability handles protection once with erasure coding instead of replication. Global similarity-based data reduction runs across the whole namespace instead of per volume. An SCM-optimized write path absorbs random and bursty writes in Storage Class Memory and destages to SSD in larger sequential segments, cutting write amplification without a fat overprovisioning pool. Every one of those is existing DASE behaviour. What VAST has built here is a sales motion and a qualification fast lane, not a new data path.
"With VAST Amplify, we're giving customers a practical alternative: reclaim the flash you already have, consolidate it into a modern architecture, and materially increase the usable capacity and performance you can deliver to the business," said Phil Manez, vice president, GTM execution at VAST Data.
Decoded, that is an offer to displace a competitor's array using the competitor's own drives, at a moment when the customer cannot get replacement drives on any reasonable schedule. It is also a migration, and the release does not say what a reclamation engagement costs, how long qualification takes, or which server and SSD models are covered.
The 6x figure deserves the same scrutiny. It is not a measurement. It is the product of three separate multipliers stacked together, erasure coding against replication, global reduction against per-volume reduction, and reduced overprovisioning, and VAST qualifies it as up to 6x or more depending on workload characteristics and existing environment. No baseline platform is named, no dataset is described, and no reduction ratio is broken out. A three-way replicated estate holding highly similar data would sit at the top of that range by arithmetic alone, while a customer already running efficient erasure coding would not come close.
One line in the release does real work and carries no evidence: customers are already taking advantage of VAST Amplify today. None is named, and no capacity figure is attached.
The sharper technical thread is what happens next to flash demand. VAST notes that as AI scales from training into higher-volume inference, approaches that persist and reuse inference state, including key-value cache, are starting to add demand for fast storage. That is a second claim on the same scarce SSDs, which would make reclaimed capacity relief rather than a fix.
Bootnote: similarity-based reduction is not deduplication with a better name. Dedupe matches identical blocks and stops there. Similarity matching finds near-identical blocks and stores the delta, which is why VAST can quote reduction across a global namespace rather than within a volume, and also why the ratio swings so violently by data type. The number that settles a purchase is the one measured on a customer's own corpus, not the one on the vendor's slide.
Leave a Reply