Two-Thirds of MSPs Are Silent
Here’s a change in how small businesses buy technology — and a matching gap in how their providers sell it — that’s happening quietly, on its own, right now.
We start with Business Insider, which looked at how small companies are actually using AI today. The picture isn’t a pilot program or a boardroom strategy — it’s survival. Small businesses are reaching for AI directly, on their own, to cut costs and stay afloat. The reporting profiles a company called Sparkles Homes, where the founder describes leaning on AI automation to keep the business lean — handling work that would otherwise have meant hiring. Nobody sold him a managed service. He did it himself, because the math of his business demanded it.
And Sparkles Homes is not the exception — it’s the pattern. Channel E2E, reporting across the managed-services channel, put a number on how far this behavior has already spread: roughly ninety percent of companies now have people using personal AI tools at work. Ninety percent — brought in by employees themselves, one login at a time, usually without anyone in charge ever signing off. It’s already inside nearly every business you can name, and it arrived through the side door.
Now flip the camera around to the providers, because we went and measured them. The Business of Tech analyzed roughly three thousand MSP websites — every AI claim in their public marketing copy, machine-graded and validated against human labeling. And to be clear about what this is: it measures what MSPs say they sell, not what they’re capable of. It’s preliminary, and every number I’m about to give you is a lower bound.
Here’s the first one. Two-thirds of MSPs don’t mention AI at all. Across roughly twenty-nine hundred websites, sixty-eight percent have no AI mention anywhere on the site. Even among the larger, more visible MSPs — the ones you’d expect to be out front — it’s about half. The AI gold rush you keep hearing about has not reached most MSP marketing.
And of the ones that do show up, most are talking, not selling. Only about one in seven MSPs market-wide has a real, scoped AI offer — a named service or a defined engagement. About one in four among the established tier. Everyone else is either silent or just commenting on AI in a blog post.
Then there’s what the sellers are naming. Of the MSPs that name a platform in their AI offer, about eighty-five percent name Microsoft Copilot. Google, OpenAI, Anthropic — rounding errors. The providers who’ve entered the conversation are, overwhelmingly, pointing at somebody else’s product.
So hold it all together. Buyers moving on their own, from the bottom up, AI already inside ninety percent of companies through the side door. Two-thirds of providers not in the conversation at all. And the ones who are, mostly naming the same single platform. That’s the picture — before we’ve said a word about why.
The Memory Shock Splitting the Market
So far we’ve watched buyers and providers drift apart on their own. What turns that drift into a hard split — a genuine fault line — isn’t AI at all. It’s sitting underneath every laptop, every server, every device in these businesses: the price of memory.
Start with what just happened at Apple. The Verge reported that Apple raised prices across its Macs, iPads, and more — not by a few dollars, but by hundreds — and the reason it gave was blunt: a shortage of the memory and storage that goes inside the machines. When the most powerful hardware company on earth says it can no longer absorb a component cost and has to pass it to the buyer, that’s not a blip. That’s the floor moving.
And the size of the move is the part that should land. CNBC, describing a memory crunch it called existential for the small guys, put a number on it: the DRAM in these devices — take a common eight-gigabyte module — has gone from around thirty-five dollars to roughly three hundred. Nearly nine times over.
Now here’s why this doesn’t clear next quarter. Business Insider reports that the AI data-center build-out is on track to pull fifteen to twenty percent of all consumer-memory manufacturing capacity away from consumer devices and into data centers in 2027. So the same appetite driving today’s shortage has a second, larger wave already scheduled. The chips are being bid away by someone with deeper pockets, and that bidding isn’t stopping.
And notice the loop closing here. The same AI wave that sent those small businesses reaching for their own cheap tools is the very thing bidding the memory away from their next laptop. AI is manufacturing the demand at one end and inflating the cost of the hardware at the other — one force, pressing both walls of the middle inward at once.
And here’s the mechanism that matters: a rising box price doesn’t land on both kinds of buyer the same way. For the cost-driven owner — the one already reaching for cheap tools to stay afloat — a pricier machine is one more reason to defer the refresh and do it himself. For the buyer who only ever wanted an outcome, the box was never the point, so its price just accelerates the move to whoever sells the result instead of the hardware. Same force, opposite reactions.
So if the memory market is the force doing the sorting, the question is who gets caught standing in between it — and that’s a mirror the MSP has to hold up to itself.
No Buyer Left in the Middle
So put the MSP in that picture, because the spot most providers are standing in is the one spot the ground is opening under.
Techaisle went and studied this directly, and named it without flinching: the small-business technology market has split into two markets — a high-touch, managed-outcome segment and a low-cost, commoditized one — and, in their words, most partners’ go-to-market has not caught up. Read what that actually says about the provider. There is a buyer who wants an outcome owned for them, and a buyer who wants the cheapest possible tool and will run it himself. What there is no longer is a buyer for the thing in between — the undifferentiated bundle of boxes plus support at a blended monthly price. And the middle is exactly where the largest share of MSPs still sells. Remember what our own marketing scan found: two-thirds not in the AI conversation at all, and most of the rest just pointing at Copilot. That’s not a provider positioned for the top tier or the bottom one. That’s a provider standing in the gap, at the precise moment the gap is widening.
Here’s the part that should change the response, because it says the top tier is real and hungry. Babble, in its own Technology Performance Index — and this is Babble’s survey, so weigh it accordingly — found that eighty-eight percent of small and mid-sized businesses call technology critical to their success, while only forty-seven percent feel confident enough to actually deploy it. Sit in that gap. Nearly every business believes this matters, and fewer than half trust themselves to do it. That difference is the managed-outcome tier, waving its arms. Choosing to serve it isn’t a retreat from the commoditized buyer — it’s walking toward the buyer who’s telling you, out loud, that they can’t do this alone.
And the market is already starting to price the split. MSP360 — in its own announcement — just locked in pricing for its MSP partners, no automatic increases, and made the promise itself the product. When a vendor turns “your cost won’t move” into a selling point, that’s the tell: cost-risk is becoming a thing you either own on purpose or eat by accident.
So here’s the choice, and it’s one about which business you’re actually in. You can commit to the managed-outcome tier — sell the result, own it, and pull hardware out of your flat monthly number so it becomes a transparent line the client sees and owns, with the component-price risk sitting where it belongs instead of quietly bleeding your margin. Or you can keep defending the all-in, box-inclusive middle — and get ground down between a specialist reaching down from above you and a self-serve buyer walking away below you, while the memory market keeps pulling the two of them further apart.
And the honest place to start isn’t with your client’s business at all — it’s with your own.
Why Do We Care?
Because before you can sell a side of this split, you have to pick one for your own shop — and your marketing probably hasn’t. Go read your own website the way our scan read three thousand others: does it name a scoped outcome you own, or does it hedge in the undifferentiated middle? And sit with the absence we found across the whole channel — almost nobody markets that they use AI to run their clients’ IT better, which means the first provider who can prove that about their own operation has already claimed the outcome tier before saying a word about price.
What to Consider
- Grade your own site the way the scan graded the market. Pull up your homepage and services pages and mark, honestly, whether a stranger could tell in ten seconds which tier you sell — a named, scoped outcome, or a blended bundle of boxes-plus-support. If you land in the middle, you’ve just found the same gap two-thirds of the market is sitting in, and you found it before a client did. That audit is free, it takes an afternoon, and it’s the first thing that has to change, because you cannot sell a position you haven’t declared.
- Externalize your own hardware risk before you pitch it as a service. If you’re choosing the outcome tier, restructure your own agreements first — pull hardware out of your flat monthly number internally, so the memory-price shock stops landing on your margin and starts sitting where it belongs. You’ll deliver the client version of this far more convincingly once you’ve already run the move on your own P&L, and the reprice is easier to defend at renewal than to explain after a quarter of quiet bleeding.
- Close the absence — make “we use AI to run your IT better” true, then say it. The scan found essentially nobody marketing AI applied to their own delivery, only AI sold as the client’s revenue line. That’s an open lane: instrument one internal workflow with AI, measure what it saved, and turn that into the outcome-tier proof no competitor is making. It’s the credential that separates a provider who owns results from one who resells Copilot, and right now it’s unclaimed.
If this trend continues: within the next twelve to eighteen months, the MSPs whose own marketing still sits in the undifferentiated middle will be the ones a split market literally can’t place — while the providers who picked a side, and can point to AI running their own delivery as proof, will be taking the confident-gap buyers before price is ever discussed.