VAST Data’s Flash Reclaim attacks competitors’ installed bases

VAST Data thinks the SSD and HDD supply shortage problem is so severe that it has a Flash Reclaim initiative to repurpose a customer’s installed SSDs for use in VAST Data storage with its claimed significantly higher data reduction ratios – so customers cram more data into less flash.

Co-founder Jeff Denworth briefed analysts, saying that that there is a significant storage media supply shortage problem encompassing both disk drives and SSDs, due to existing HDD manufacturing and and flash fabrication capacity being more-or-less used up and new capacity being 12 to 18 months away before it comes on stream.

There is almost a duopoly in HDD supply, with Seagate and Western Digital having the bulk of the market. They came out of an over-supply problem in 2023, resolving to maintain price discipline and match product capacity better to demand. Unfortunately they matched it to relatively short-term demand, 12 months say, when it takes 18 months or more to add significant new HDD manufacturing capacity. They can currently sell every disk drive they make, make a nice profit, and have no public plans to add HDD manufacturing capacity.

This is happening at the same time as an increase in AI data storage needs. The alternative to HDDs is flash, but the five main flash fabricators also came out of an oversupply situation in 2023 and have been reluctant to add hugely expensive new fab capacity until they can see clear and enduring market demand for the SSDs that use their NAND chips.

Now they can, but new fabs take 2 years or more to come on stream so that lead times for both HDDs and SSDs are lengthening, meaning that customers’ AI and other IT initiatives needing storage capacity face delays.

Denworth said: “VAST is a very serious entity within the market now. If there’s roughly a hundred or so exabytes of enterprise flash that will be deployed from the flash manufacturers out into the market, VAST and our customers … will be deploying in excess of a few dozen exabytes of that total capacity that’s available in the marketplace. We’re now to the point where we’re representing a significant part of the market. The sales are booming … we’re growing at hundreds of percents this year.”

The point being that VAST’s growth will be slowed if customers can’t get the flash capacity that they need.

After talking to potential customers and media suppliers at CES 2026, VAST realized that this was not a minor supply blip. Denworth said: “We’re now in the worst, let’s call it storage rut, … the people that we were talking with have seen in their entire careers, some of which date back for 40 years.”

He thinks that “there’s roughly about a 200 exabyte deficit of capacity in the market driven by hard drive manufacturers. That then leads the market to go and source enterprise SSDs as an alternative to hard drives for deep capacity. So, in particular here, QLC becomes the target for people that have large scale capacity problems and QLC drives have just been under a tremendous amount of demand and allocation, which then results in a situation that we’re now in where there’s just not enough supply to go around, and people call this a dry year.”

“Lead times are now basically pushing out to nearly COVID levels. … there’s aggressive buying that’s happening as a result of people realising that everybody’s under allocation. And what’s happening now is that lead times are now spiking to 52 weeks. You’re talking about a year of delivery lag in certain cases, and it’s contributing to a major, major issue in the market.” In particular: “hard drive lead times are now in excess of 18 months.”

Denworth concludes: “People just have data that they need to store as power and cooling are no longer critical buying considerations, and availability is the number one buying determinant.”

What can be done if the supply of new flash capacity is restricted? Re-purpose existing SSDs.

VAST says its erasure-coded data protection needs less overhead than competing RAID systems, again meaning more capacity for data. Denworth said this was provided up to 50 percent more capacity utilization for companies building big data lakes: “Big data products that typically triplicate their data, and here there’s roughly a 70 percent error correction overhead efficiency advantage. That means that, in some cases, we get roughly 50 percent more capacity utilization. In ex extreme cases, we can deliver 200 percent capacity utilisation.”

Also, VAST claims, its data reduction technology is so much better than its competitors that you need less flash to hold the same amount of data. Denworth again: “Our average data reduction ratio value today is 3.4 to one, and that’s the average across roughly a thousand or so customers. Our median DRR value is 1.75 for one. So there’s a few numbers that are pulling that down versus the average, but then there’s also some extreme numbers on the other hand where customers are just seeing an exceptional amount of data reduction at the highest end. It’s currently 724 to one,” citing VAST’s call home customer telemetry data.

So “this, coupled with the efficiencies that we bring from a RAID perspective, deliver on average an outsized level of benefit. … with VAST, you just need one third less than what you would need with any other technology.”

The VAST idea is to re-purpose a customer’s existing SSD capacity where possible: “We’re rolling out a new programme. It’s currently titled the Flash Reclamation Programme. This is a programme that our sales team and our sales engineering team will be rolling out that basically goes into a customer estate and says, okay, if you can’t get new flash drives, how can we make the drives that you have much more capacity-rich?”

“How can we take our software, remove the software that you’ve got deployed on it from a storage or database perspective and essentially give you two to three times more available capacity than what you had previously planned on? So this is obviously a very disruptive approach to selling, but desperate times call for desperate measures.”

It’s pretty much a naked attack on its competitors’ all-flash array installed bases. Denworth again: “If you made an investment in some sort of legacy storage array and we can give you four times some capacity back from that SSD that sits inside of it, customers will have to make some hard decisions.”

He added “We already have a few customers that are looking at what they’ve acquired over the last year saying, I haven’t necessarily deployed this into production, and if VAST can give me three times the [capacity], then this is the only practical way that I can keep up with data generation within our environment. And so it involves us coming in with our engineers to do a full estate survey. Our R&D team has committed to doing rapid hardware qualifications, and then we can basically extend NVMe investments by a factor of two to three.”

In practise: “That includes us going into customers’ data lakes and ripping out the different types of SSDs they may have in their storage systems. In certain cases, we’ll be able to use their servers as well. That includes us going into different event streaming environments. And in particular, if you look at what contributed to a lot of the recent data creation, it’s AI. If we can get a 1.71 data reduction and then some additional error correction efficiency on top of that, I can take a system that’s not running VAST software for AI workloads and I can extend the capacity by a factor of two.”

There’s a cost aspect to this as well: “The interesting thing though is, if you look at this programme, this Flash Reclamation programme, not only does it free up customer’s capacity, it also pins the price point to 2025, which is roughly 50 percent less than what customers today have to pay.”

Denworth thinks the flash supply situation is going to get worse because of a growing need for AI context memory storage (to hold KV cache data): “What we expect is that just with the forecast that Nvidia has over the next 12 to 24 months, coupled with some other financial analyst forecasts, there’s going to be a need for another 100 exabytes of additional capacity in the market.”

Overall: “we’re going to do our damnedest to compress the net capacity impact that this has on the market.

Finally, if you run your AI inferencing using VAST’s AI OS, running on BlueField-3 and 4, and its storage, he says you get a lot of power savings as well.

We can expect to hear a lot more about this at VAST’s Forward 2026 conference, Feb 24-26, in Salt Lake City.

Bootnote

Asked about potential System Integrator involvement, Denworth said: “There’s going to be no systems integrator effort. Basically, we take whatever SSDs are available [and] we give them to our platform partners. We actually had to do this with a very large AI lab last year. It’s not the first time that we’ve encountered this, where they bought a new product for another project that they were working on. That product failed miserably, and so they were left with a few hundred petabytes of available flash and they wanted to put it into their estate. And so we just said, okay, give us the drives, we’ll qualify them. And within a span of three weeks, I think, we actually had 300 petabytes of some other vendor’s appliance flash deployed in the customer’s production environment.”