A disclosure before anything else: I work in presales for an infrastructure vendor that competes in this market. Read the following accordingly, and check it against your own sources. I have tried to describe dynamics rather than pick a favourite, and where I am probably biased is in what I find interesting rather than in what I claim is true.
The category split
What used to be one market is now two conversations that share a name.
Purpose-built backup storage is a target: a deduplicating appliance that your backup software writes to. Dell’s PowerProtect Data Domain has led this for a long time, with HPE StoreOnce and ExaGrid as the persistent alternatives, plus Quantum and others in specific segments.
Data protection platforms are the software: Veeam, Commvault, Rubrik, Cohesity following its Veritas acquisition. These sell the whole function, increasingly framed as cyber resilience rather than backup.
The second category is where the growth and the valuations are. The first is a mature market being fought over on price and efficiency.
What the numbers suggest
Independent tracking puts Cohesity-Veritas at the largest share of the data protection market after that merger, with Veeam second and growing fastest, and Rubrik prominent among the rest. Rubrik is widely credited with driving the repositioning of backup as cyber resilience, which changed who signs the purchase order — from infrastructure to security.
On the hardware side, Data Domain remains the leader by a wide margin while its mindshare has been reported as declining year on year. ExaGrid has reported sustained growth in the upper mid-market, explicitly on the basis of replacing incumbent appliances, and its architecture is the reason: a non-deduplicated landing zone that makes recent restores fast, which is a genuine architectural difference rather than a marketing one.
Meanwhile several software vendors have stepped back from selling their own hardware, having concluded that competing on boxes against the large infrastructure vendors is not where their margin is.
What that means if you are buying
The appliance decision is now downstream of the software decision for most organisations. Choose the data protection platform, then choose what it writes to. The reverse order was normal a decade ago and is now unusual.
Architecture differences are real, not marketing. Inline deduplication everywhere versus a fast landing zone is a genuine trade between capacity efficiency and restore speed. Ask which one you are buying and match it to whether your worst day is a capacity problem or a recovery-time problem.
Integration is the practical constraint. The appliance has to support your backup software properly, including the vendor-specific fast-path protocols that make ingest efficient. Check the version matrix for your specific combination rather than the general compatibility claim.
The flash question is unsettled. All-flash backup targets exist and make restores fast. Recent NAND pricing has made the economics harder, and reporting suggests that has cooled demand for flash-based backup targets generally. Worth watching rather than deciding on today’s price.
What I would actually do
Write down what you are protecting against, in priority order. If it is ransomware recovery time, weight restore architecture and immutability heavily and accept a worse capacity ratio. If it is capacity cost on a large slow-changing estate, weight deduplication. If it is operational simplicity for a small team, the platform vendors’ integrated offerings are genuinely easier and you pay for that.
Then get two vendors to run an assessment against your real data. The deduplication assumption and the restore rate are the two numbers that decide this, and both are measurable rather than debatable.
The thing that has not changed
Whoever you buy from, the appliance is a component of a recovery capability, not the capability itself. The organisations that recover well are the ones that have tested it. That is independent of the logo, and it is the only part of this market where the difference between vendors is smaller than the difference between customers.