When I was building a municipal network for a school district, a board member said "I have a router, I used it last weekend." I asked how it went. "Great. The cabinets came out beautiful."
That's how I feel every time a storage company tells me they're part of my AI stack.
What this means for the CIO, CTO, and CISO
Storage vendors claiming AI relevance are repackaging old concepts with new buzzwords. "Moving processing to data" is edge computing, and we've discussed it for a decade. It's not an AI innovation.
Adding an "AI layer" to storage doesn't make storage part of your AI stack. I don't know why storage companies think I'd select them as my AI provider. I wouldn't select my AI provider as my storage vendor either.
The root cause storage vendors ignore is data quality, not data speed. Garbage in, garbage out applies regardless of how fast your storage is. If data quality is low, AI projects fail, full stop.
The inside perspective
When I was managing infrastructure, storage was storage. It held data. Fast storage held data faster. The value was in what you did with the data, not where you stored it.
I never once sat in a planning meeting and said "our AI strategy depends on our storage vendor." I never evaluated storage arrays based on their AI capabilities. I never considered my storage provider to be part of my AI architecture.
Storage is infrastructure. Necessary infrastructure. Important infrastructure. But infrastructure nonetheless, like networking, like power, like cooling. You need it. It needs to work. It's not where the value creation happens.
The value happens in the models, the data quality, the use cases, the business outcomes. Storage is the plumbing. Good plumbing matters. But nobody ever said "our competitive advantage is our plumbing."
The outside observation
Now I watch storage vendors desperately try to insert themselves into AI conversations.
Every storage deck I see has AI slides. "AI-optimized storage." "Storage for the AI era." "Your AI infrastructure partner." The messaging is relentless.
I understand why. Storage vendors have been competing on cost-per-terabyte for decades. Flash solved the performance problem for most workloads years ago. Modern arrays are fast enough for the 90th-plus percentile of enterprise use cases. Storage has become a commodity.
When you're a commodity, you look for relevance. AI is the hottest thing in enterprise technology. If you can convince customers that storage is "AI infrastructure," you escape the commodity trap. You justify premium pricing. You get invited to strategic conversations instead of procurement negotiations.
The problem is, it's not true. Adding AI buzzwords to storage marketing doesn't make storage an AI platform. It makes it storage with better marketing.
The uncomfortable truth
The conversation about moving processing to data is not new. It's not an AI innovation. It's edge computing, and we've been discussing it for a decade.
Moving compute closer to data sources makes sense for certain workloads. Reducing latency. Minimizing data transfer costs. Processing at the edge where data is generated. These are legitimate architectural patterns.
But they're not new. They're not AI-specific. And storage vendors didn't invent them.
What storage vendors are really saying is: "We'd like to be relevant to AI, so we're claiming that the proximity of storage to compute is an AI innovation." It's not. It's infrastructure optimization that has existed for years, applied to a new workload type.
The baseball cards analogy
Let me be direct about what's happening here.
Some of this "AI storage" conversation feels like the nerdy kid saying "I play baseball too" while pulling out a collection of baseball cards. "I'm an athlete too, I play Magic: The Gathering." We're not talking about the same thing. We're not playing the same game.
Storage vendors storing data that AI uses doesn't make them AI companies. My refrigerator stores food that I cook with. That doesn't make it a cooking appliance.
While it's cute to say "we're an AI company too," the reality is you're not. You're a storage company. Adding AI, or adding an AI layer, isn't actually helpful to me.
I don't know why storage companies think I would select them as my AI provider. I don't know why they think they should be considered part of my AI stack. If AI is used for their benefit, within their closed-loop system, to optimize their storage operations, great. That's not part of my AI strategy. It shouldn't be conflated with my AI architecture.
The commoditization desperation
Let me explain why storage vendors are doing this, because understanding the motivation helps you evaluate the claims.
The performance era. Storage used to differentiate on speed. Faster IOPS. Higher throughput. Lower latency. Vendors commanded premiums for performance.
Flash disruption. SSDs changed everything. Flash is fast enough for the vast majority of enterprise workloads. Performance differentiation collapsed for most use cases.
Commodity pricing. With performance solved, competition shifted to cost-per-terabyte. Year after year of price decline. Margins compressed. Storage became infrastructure that buyers negotiate hard on price.
Feature parity. Deduplication, compression, replication, snapshots. Every vendor has similar capabilities. Differentiation is minimal. Switching costs are the main thing keeping customers in place.
The AI pivot. AI is hot. AI commands attention and budget. If storage can be "AI infrastructure," maybe it escapes commodity pricing. Maybe it gets invited to strategic conversations. Maybe margins recover.
That's the motivation. Evaluate vendor claims accordingly.
The market reality nobody discusses
Here's what the storage-AI narrative ignores.
The hardware adoption bottleneck. Only real tech disruptors and cloud infrastructure companies can onboard AI hardware fast enough. Most enterprises are struggling to get out of first gear with the GPU hardware they've already bought. Storage isn't the limiting factor. Organizational capability is.
Dormant infrastructure. Millions, maybe billions, of dollars in GPU infrastructure is sitting dormant right now. Waiting for workloads. Waiting for use cases. Waiting for organizations to figure out what to do with it. The constraint isn't storage throughput.
The non-existent problem. Very few companies are actually struggling to get data off their AI systems fast enough. Very few are bottlenecked by storage performance for AI workloads. Storage vendors are solving a problem most enterprises don't have yet.
The real problem. Data quality. Garbage in, garbage out. If your data quality is low, your AI projects fail, regardless of how fast your storage is. This is what's actually causing AI failures at enterprise scale.
Storage vendors want to sell you faster plumbing when your actual problem is that the water is contaminated.
The root cause they're ignoring
Look at the MIT reports on AI project failure. The Gartner surveys. The statistics about implementations that don't deliver value.
Dig into them. The root cause is almost never "storage wasn't fast enough." The research consistently shows poor data quality, unclear business objectives, lack of AI expertise, no data governance, unrealistic expectations, organizational resistance, and missing change management.
Storage performance appears nowhere in the top causes.
The MIT research on AI project failure consistently shows organizations did nothing for data quality. They flung garbage at models hoping diamonds would emerge. That's neither how garbage nor diamonds work.
If your data quality is low, it doesn't matter how fast your storage is. The garbage arrives faster. The wrong answers come back quicker. The project fails at higher throughput.
Storage vendors are positioning premium "AI storage" as a solution to problems that aren't storage problems. That's not helpful. It's distracting.
Signs you're buying the hype
If three or more of these apply, reconsider your approach.
- Your storage vendor is invited to AI strategy meetings. They sell infrastructure, not AI capability. Why are they at the table?
- You're evaluating storage based on "AI optimization" claims. Ask what specific AI problems this solves that standard modern storage doesn't. Demand specifics.
- You're paying premium pricing for "AI storage." What exactly are you getting beyond marketing? Can you quantify the AI-specific value?
- Your storage vendor can't explain how they improve AI outcomes. If they can't articulate specific, measurable benefits, the "AI" label is just marketing.
- You haven't addressed data quality but you're buying AI storage. You're optimizing the wrong layer of the stack. Fast garbage is still garbage.
- Your AI projects are failing and a storage upgrade is the proposed solution. Storage almost never appears in root cause analysis of AI failures. Look elsewhere.
What storage vendors actually offer
To be fair, let me acknowledge what storage vendors legitimately offer for AI workloads.
- High-throughput storage for large training datasets.
- Low-latency access for inference workloads.
- Reliable, durable data storage.
- Integration with data pipelines.
- Scale-out capacity for growing datasets.
These are real infrastructure capabilities that matter. I'm not saying storage is unimportant.
What's not legitimate is claiming to be part of your AI stack, positioning as a strategic AI partner, claiming AI-specific capabilities that change outcomes, charging premiums for "AI storage" that's just storage with marketing, or suggesting storage solves AI project failures.
Storage can be good infrastructure for AI workloads. That's different from being an AI company. Know the difference.
The 2026 prediction
Storage vendors claiming AI relevance will face a reckoning in 2026.
Enterprises will realize that fast data access doesn't fix their actual problems: data quality, organizational capability, unclear use cases, lack of expertise. The "AI storage" premium will become harder to justify when AI projects continue failing for reasons that have nothing to do with storage.
The vendors who positioned storage as AI infrastructure will find themselves back in commodity pricing negotiations. The enterprises who fell for the marketing will have wasted budget on storage premiums while their actual AI blockers remained unaddressed.
Storage will remain necessary infrastructure. It won't become strategic AI capability. The marketing might change. The underlying reality won't.
A note on analyst culture
The analyst industry has created "AI Storage" as a market category. Reports evaluate vendors on AI capabilities. Landscapes include anyone using the buzzword.
This validates vendor marketing rather than challenging it. It helps storage vendors justify premium pricing. It confuses enterprises about where to allocate AI budget.
We analysts are supposed to be the skeptics in the room. We're supposed to validate vendor claims, to distinguish between marketing and capability. When a commoditized infrastructure vendor suddenly claims strategic relevance to the hottest market in enterprise technology, we should be asking hard questions, not amplifying their messaging. When we create categories based on vendor positioning rather than genuine capability differentiation, we're failing at the job.
Storage is storage. AI is AI. Adding one word to the other doesn't change the underlying reality. Claiming to be "AI infrastructure" because you store data AI uses is like claiming to be a chef because you own a refrigerator. Customers are starting to notice.