Put two numbers next to each other.
HVAC roll-ups are trading at roughly 8x EBITDA right now — that's what we've seen across current roll-up activity as buyers underwrite acquisitions in the trades. Meanwhile, Eric Siu, in a widely-viewed X post, has argued that AI-native service firms are commanding multiples closer to 30x — and that this is precisely why a16z, Sequoia, and YC are, in his words, "chasing services, not SaaS."
Same word — "services." Almost four times the multiple. That gap is worth sitting with, because it's easy to misread and expensive to misread wrong.
Two facts, two sources, don't blur them
Start with what's actually verifiable and where it comes from.
The ~8x figure is Advisy's own read on current roll-up activity in the trades — it's what operators and consolidators are paying for HVAC and adjacent home-service businesses today. It's not a single deal, not a universal law, just the observable center of gravity in that market right now.
The ~30x figure is Eric Siu's claim, made in an X post, about a different pool entirely: AI-native service businesses attracting venture and growth capital. That's a different buyer, a different thesis, and — this matters — a different kind of company. Nobody is paying 30x for a plumbing business. Nobody is paying 8x EBITDA for a two-year-old AI services startup with no EBITDA to speak of. These are not two prices on the same shelf. They're two different shelves.
That distinction is the whole point of this post. Not "AI gets you a better multiple." Something more specific, and more useful.
What Siu is actually arguing
Siu's thesis isn't "add AI, multiple goes up." It's sharper than that, and worth restating carefully because most people will round it down to something simpler.
His argument: most agencies and services businesses are about to misread this moment. The obvious move — the one nearly everyone will make — is bolting AI onto an existing billable-hours model to cut headcount and pad margin. Same service, same delivery model, same client relationship, just fewer people doing the work and a fatter number at the bottom. Siu calls this "playing the small game." It's real, it's available to almost anyone, and it's not where the re-rating is happening.
The bigger opportunity, in his framing, is something he calls "services-as-software" — closing the gap between what gets spent on software (roughly $1) and what gets spent on services (roughly $6) for the same category of outcome. Not by replacing the service with a SaaS tool, and not by quietly automating headcount out of the existing model, but by rebuilding the delivery model itself around AI agents plus human judgment, priced and sold on the outcome rather than the hours. That's a structurally different business than "the same agency, now with fewer employees." It's why, in his view, VC is circling services at all — historically not their territory.
What this actually means if you own a trades business
Here's where it's tempting to draw a straight line: "adopt AI, move from 8x toward 30x." Don't draw that line. It isn't true, and claiming it would be the kind of promise this post is explicitly not going to make.
The 30x multiple belongs to a different buyer pool — venture and growth investors underwriting AI-native software-shaped businesses — and the 8x multiple belongs to strategic and PE buyers underwriting trades roll-ups on fleet age, customer concentration, technician retention, CRM maturity, and owner-dependency. Those are separate markets with separate rules. A garage door company or an HVAC shop doesn't migrate from one to the other by installing a chatbot, and nobody serious would tell you it does. AI doesn't fix a concentrated customer base or an owner who's the only one who can close a big deal. Those remain the larger levers in how a trades business gets valued today.
What does travel across the gap is the thesis, not the number. The distinction Siu is drawing — restructuring how a service actually gets delivered and priced around outcomes, versus quietly automating the existing model to protect margin — applies inside the trades just as much as it applies to agencies. Inside your own industry, there's still a meaningful difference between a services business run the way it's always been run, even with AI quietly trimming a few line items, and a services business rebuilt around an AI operating layer that changes what you can promise a customer and how you price it. One of those is optimization. The other is a different kind of business, running inside the same industry, priced by a different logic — even if today's trades buyers haven't fully caught up to pricing it that way yet.
That's the gap worth paying attention to. Not "8x becomes 30x." "Which kind of business am I actually building, inside the industry I'm already in."
Where this fits into what we build
This is the exact seam Advisy sits in: not a marketing agency bolting AI onto lead-gen, not a legacy M&A broker pricing off multiples that haven't updated yet. We build the AI operating layer — the Citadel Council agents running strategy, ops, sales, and finance day to day — with an eye on the same underwriting criteria buyers actually screen for, whichever market eventually prices your business.
If you're trying to figure out which business you're actually building, not just which tools you're bolting on, talk to us: advisy.com/#apply.