NAND prices have risen 600 to 800% in the last six to seven months. Everpure's own product costs are up more than 300%. And Charlie Giancarlo, the company's Chairman and CEO, spent a meaningful chunk of his keynote and the press Q&A that followed talking about it directly, instead of letting it sit quietly in a footnote.
That kind of candor is unusual at a vendor conference built around new product announcements. But it set the tone for a day that mixed hard financial reality with Everpure's biggest architectural bet yet: convincing customers to rebuild how their enterprise data works, in the middle of a semiconductor pricing crisis nobody saw coming a year ago.
The Semiconductor Math
Giancarlo didn't soften the numbers. NAND flash costs have spiked 600 to 800% since roughly the end of last year — a swing he called breathtaking. Everpure's own product costs have risen more than 300% as a direct result.
His explanation traces back to fixed-cost economics. Semiconductor fabs take years and billions of dollars to build. They don't come online quickly, and when AI demand suddenly outstrips total global fab capacity, prices don't adjust gradually — they spike, because supply simply can't catch up on any reasonable timeline. It's not specific to storage, either. Giancarlo pointed out that the price shock is hitting every category of semiconductor, from ten-cent components to enterprise-grade chips, across the entire industry.
When a journalist directly challenged him with language used by competitors — accusing vendors broadly of profiteering off the shortage — Giancarlo's response was concrete rather than defensive. Everpure indicated on its fourth-quarter earnings call that it would operate at the lower end of its gross margin range, and he said the company has done exactly that. Prices have risen far less than the underlying 6-8x cost increase Everpure is absorbing, and he pointed to that margin compression as the actual proof point, not a talking point. He also noted that competitor price increases came sooner and were steeper.
He had sympathy for the position customers are in, beyond the higher bill itself. Storage pricing has deflated for decades. Customers have never had to develop the skill of evaluating whether a price increase is fair, because pricing only ever went one direction. Now that assumption is broken across the board, and there's no easy reference point for what a "good deal" even looks like anymore.
One side effect worth noting for infrastructure planners: Giancarlo said the price shock has given the hybrid hard disk and flash market a new lease on life, at least temporarily, as some workloads shift back toward disk economics that look more attractive again given flash pricing.
The Architectural Pitch, Mid-Crisis
It would be reasonable to expect a company under this kind of margin pressure to play it safe on messaging. Instead, Giancarlo used the keynote to make the most ambitious architectural argument Everpure has made in years: that enterprise IT needs to fully invert its relationship between applications and data.
He framed the urgency in concrete terms from Everpure's own environment. The company runs more than 750 internal applications, each with its own version of what the data actually means. Pulling together something as simple as a single invoice requires reconciling data across many of those systems, and they frequently don't agree with each other. That's not a hypothetical problem he's selling to customers — it's one Everpure is actively living through itself.
His framing for why this is different from data-centric pitches the industry has made before, including ones Giancarlo himself made years earlier under different language, comes down to two forces converging at once. AI has pushed enterprises into the same trap that analytics created — copying data yet again for every new AI use case, on top of every analytics copy that came before it. And SaaS vendors, in trying to make their own AI agents useful, are each independently asking customers for a copy of their data too. At some point, Giancarlo argued, enterprises have to ask how many times they're willing to replicate the same information across how many different vendors before the model breaks down entirely.
The technical distinction he drew between data primacy and traditional ETL is worth sitting with. ETL is fundamentally about copying data and transforming it into a new form for a specific purpose. Data primacy, in his framing, is about reading data in place and building a layer of relationships across systems without copying it at all — letting agents and analytics tools leverage those relationships directly. It's a smaller technical distinction than the marketing language suggests, but it's the distinction the entire pitch depends on.
Answering the Hard Question Directly
I asked Giancarlo a direct question during the Q&A: customers are already stretched thin by this year's storage cost increases, and data primacy is a multi-year architectural undertaking. What's the realistic business case for starting that journey now, rather than waiting for pricing to stabilize?
His answer was specific rather than deflective. Everpure's own internal transition — reconciling data sources across all of its application environments and building governance around the resulting systems of record — is expected to take roughly two and a half years. The payoff, in his telling, comes in three forms: meaningfully reduced human labor spent manually reconciling spreadsheets and one-off integrations, fewer copies of data circulating across the enterprise (which directly shrinks both attack surface and liability risk), and a reframed starting point for IT planning. His core argument was that enterprises re-architect some part of their IT environment every single year regardless. The real choice isn't whether to do re-architecture work this year — it's what target that work is aimed at. Customers don't necessarily need to change what they're building this year, just the direction they're aiming it.
It's a fair answer, and one that reframes the timing objection rather than dismissing it. Whether two and a half years holds up as Everpure's own case study unfolds is something worth tracking.
The Organizational Problem Nobody Wants to Own
A question from a separate journalist pushed Giancarlo on something that's easy to gloss over in a product-focused keynote: getting different functional groups inside an enterprise to agree on shared data definitions is arguably more of a political problem than a technical one.
Giancarlo agreed without hesitation, and described how Everpure has handled it internally. The company's own data primacy initiative — internally codenamed Mercury — has been running for about a year and a half, with a weekly coordination meeting Giancarlo has personally attended the entire time. His point wasn't that the CEO necessarily has to drive this kind of program at every company, but that it requires a senior leader with real authority, likely a chief data officer, because no single functional group can resolve these definitional disputes by themselves. Finance, sales, and operations all built their own systems around workflows that made sense locally. Aligning them requires a level of cross-functional negotiation that doesn't happen organically.
He drew a comparison to the first implementation of major platforms like Salesforce, Workday, or ServiceNow — pointing out those rollouts also required serious organizational change management. The difference with data primacy, in his view, is that instead of organizing the change around a single application's workflow, organizations now have to think about how every new initiative affects their underlying data architecture, regardless of which application happens to touch it.
A Contrarian Note on Tokens
In a smaller moment that's worth surfacing for engineering audiences, Giancarlo pushed back on a trend he's seen in recent industry announcements: vendors promoting "token maximization" as a selling point. His view is straightforward and a little contrarian relative to how that's often marketed. Business leaders don't actually want to maximize token usage — tokens cost money. They want token minimization for a given quality of answer. He argued that well-architected storage, by enabling AI systems to retrieve precise, contextual results rather than reprocessing the same broad context repeatedly, directly supports that goal. Storage, in his framing, is simply cheaper than memory and cheaper than redundant token generation.
It's a useful reminder for anyone evaluating AI infrastructure vendors that "more tokens" and "more capability" aren't the same claim, and that the economics often favor whichever architecture gets you a good answer with less computational overhead, not more.
What This Means Going Into the Rest of 2026
Giancarlo's NAND commentary suggests storage buyers should expect continued pricing pressure for the foreseeable future — fab capacity constraints don't resolve in months. His direct, numbers-based response to the profiteering accusation is worth keeping in mind as a reference point if other vendors' pricing moves come up for comparison later this year.
On the data primacy front, the honest answer is that this is a story still being written. Giancarlo is candid that Everpure's own internal transition will take roughly two and a half years, that the political and organizational challenges are real and unresolved, and that the company is making a multi-year bet during a period when its own margins are under unusual pressure. That combination — financial strain and architectural ambition arriving at the same time — is the real story coming out of Pure Accelerate 2026, more so than any single product announcement.
Everpure (NYSE: P) provided press access to Pure Accelerate 2026.
