EBITDA multiple

Once a company is genuinely profitable, buyers usually prefer to price it on what it earns rather
than what it sells, because earnings are closer to what an owner actually gets to keep. An EBITDA
multiple is the standard way of doing that: it prices the business against earnings before
interest, tax, depreciation and amortisation, which strips out financing choices and accounting
conventions so that two companies can be compared on their operating performance alone.

For a founder, the appeal is that it rewards the discipline of running a genuinely profitable
business rather than a fast-growing one that has not yet proven it can be. The catch is that EBITDA
itself is not a fixed, objective figure — it is built from a set of add-backs and adjustments that
both sides will argue about, and that argument can move the underlying number as much as the
multiple applied to it.

The multiple is applied to an EBITDA figure, but almost nobody uses the raw accounting number.
Buyers and sellers negotiate "adjusted EBITDA" — the reported figure plus add-backs for one-off
costs, owner compensation above market rate, and expenses that will not recur under new ownership.
Every one of those add-backs is a negotiation in itself, because a generous add-back inflates the
base the multiple is applied to, and a skeptical buyer will challenge every item that looks
convenient rather than genuinely non-recurring.

For an AI company, this adjustment process carries extra weight because a meaningful share of
operating cost can be compute and model-inference spend, and the two sides frequently disagree
about how much of that is a genuine cost of running the business versus a cost that a new owner
could reduce through better infrastructure or pricing negotiated at their own scale. That
disagreement changes the EBITDA base itself, before the multiple is even discussed — which is why
two parties can agree on a multiple and still be far apart on price.

Adjusted EBITDA is a negotiation, not a fact. Every add-back — one-off legal costs, an owner's
above-market salary, a bad hire that will not be repeated — has to be justified individually, and a
buyer's advisors will challenge the ones that look self-serving. Be ready to document each
adjustment with evidence, not assertion.

Compute and inference costs are a live argument in AI deals specifically. If a large share of
your cost base is model usage, expect a buyer to ask whether that cost is structural or something
their own scale and infrastructure would reduce — and expect the answer to move your adjusted
EBITDA meaningfully in either direction.

A small EBITDA makes the multiple do outsized work. When earnings are thin, a modest change in
the EBITDA figure — one add-back accepted or rejected — swings the resulting valuation far more than
it would for a company with substantial earnings. Scrutinise the base as carefully as the multiple
when your margins are tight.

The multiple embeds the same judgement a revenue multiple does. Growth, quality of earnings, and
customer concentration all still matter here; profitability does not remove the need to explain
why your business deserves the multiple being proposed rather than a lower one.

Where a company has real earnings to price, the AI analysis report's valuation range reflects
that — it is built from what you uploaded plus public data across an eight-model pipeline, and it
comes with per-section confidence rather than a single confident-sounding figure, because the
adjusted-EBITDA argument described above genuinely produces a range of defensible outcomes. The
Valuation Up-Lift stage on the M&A track is where that range meets a real buyer's own view of your
adjustments.

Start with the free tool at /valuation to see where your earnings-based valuation
currently sits.