AI M&A AdvisoryFounder-led AI & software companiesApplied AI, data & MLOps, infrastructure, and model teamsUS & CanadaSenior MD-led
In brief
Windsor Drake runs confidential, competitive sale processes for founder-led AI and AI-native software companies in the US and Canada, led by a senior MD. Pricing splits by layer of the stack: foundation models and generative AI at 12x to 20x+ revenue, AI infrastructure at 6x to 10x, data and MLOps platforms at 6x to 9x, and applied and vertical AI applications at 3x to 6x. Proprietary data, unit economics after inference cost and retention move the number.
Who We Advise
We work with founder-led companies across the AI stack. Our clients are operators with real customers and real revenue who want senior, conflict-free advice on whether, when, and how to sell. Every engagement is led personally by the firm’s principal, not handed to a junior team.
Operators with real customers and revenue using AI to win a specific industry or workflow.
The tooling and data infrastructure that trains, deploys, observes, and governs models in production.
Compute, inference, orchestration, and the infrastructure-adjacent software AI workloads depend on.
Foundation-model and applied-research teams whose data, weights, and talent a buyer cannot easily assemble.
What AI Companies Are Worth
AI pricing splits sharply by layer. The premium goes to proprietary data, proven unit economics, and retention you can evidence, not to the word “AI” in a deck.
| Foundation models & generative AI | 12x to 20x+ | Proprietary data, durable inference advantage, earliest-stage scarcity |
| AI infrastructure & infra-adjacent software | 6x to 10x | Compute economics, ecosystem position |
| Data & MLOps platforms | 6x to 9x | Workflow lock-in, data gravity |
| Applied, enterprise & vertical AI applications | 3x to 6x | Retention, unit economics after inference cost |
Indicative ranges; the earliest-stage model assets clear materially higher. For a fuller picture, see our breakdown of who buys AI companies and the current ranges in our AI software valuation report.
How We Run the Process
The work is hands-on from first call to close, and the founder is never handed off to a junior team.
01
We frame your data, model, and revenue story for the specific buyers who pay for it, and prepare materials that survive diligence.
02
We identify the short list of strategics and sponsors whose roadmap or thesis fits your layer of the stack.
03
We approach buyers under NDA and run them in parallel, so price reflects competition rather than a single bid.
04
We negotiate cash at close, rollover, retention, and earnout terms, then manage diligence through signing.
Who Buys AI Companies
The right buyer depends on your layer of the stack. A process that reaches all three pools, and creates tension between them, is what separates a premium from a clearing price.
Hyperscalers, enterprise-software and security platforms, and data and chip incumbents acquiring capability they cannot build fast enough internally. They price the time and cost of building the same thing, and most offers include retention packages for the technical team.
Sponsors buying applied-AI and vertical software with real revenue, retention, and a path to margin expansion. They underwrite cash flow, so gross margin after inference cost and net revenue retention carry most of the price, and offers often include rollover equity.
Investors backing model and infrastructure leaders, often the bridge between a private process and a larger strategic outcome. They pay for growth and market position, and many prefer a majority or minority recapitalization to a full sale.
What Buyers Underwrite
Buyers price an AI company on what diligence can verify. Six questions come first, and each one moves the price or the structure of the offer.
Whether the edge is proprietary and durable, or replicable with capital. Buyers test it by asking what a well-funded competitor would need to rebuild the product, and how long that would take.
Where the training and fine-tuning data came from, under what license, and whether customer contracts allow their data to improve the model. Gaps here become specific indemnities and escrows in the purchase agreement.
How much of the product runs on a model the company does not control, and what happens to cost and performance if that provider changes price or terms. Products that can switch models, or run their own, carry less risk.
Gross margin after inference and hosting cost, cohort by cohort, and whether it improves as usage scales. Buyers rebuild the margin from usage data and do not accept a blended figure.
Whether customers stay once the pilot budget ends. Buyers separate contracted recurring revenue from usage spikes and paid pilots, and give less weight to revenue that has not yet renewed.
The engineers and researchers a buyer would struggle to hire, and clean ownership of the code, models and weights. Invention assignments and open-source license compliance are checked line by line.
Before Going to Market
The work that protects price starts about a year before the first buyer call. Most of it is documentation that diligence will ask for anyway.
01
Confirm invention assignments from every employee and contractor, audit open-source licenses in the code base, and record the source and license of every dataset used in training.
02
Rebuild gross margin by customer cohort after inference and hosting cost, and show the trend. Renegotiate any model or compute contract a buyer would read as a concentration risk.
03
Assemble the evidence that the model works, including benchmark results against the alternatives a customer could use and error rates in production, along with the security and privacy documentation enterprise buyers request.
04
A buyer list built by layer of the stack, with the strategic rationale for each buyer written down before outreach, so every conversation starts from why that buyer should pay more than the others.
Most founders sell once. The outcome depends on the judgment of the person running the process, their command of how AI buyers actually price and underwrite, and their willingness to protect your interest when it costs them convenience. Windsor Drake takes a limited number of engagements so each receives senior attention, and we advise only in sectors we know well, so buyers know a business we bring has been vetted and positioned by people who understand the category.
Related reading: AI Software Valuation Report · Who Buys AI Companies · How to Sell to Private Equity
FAQ
It depends on your layer of the stack. LLM and generative-AI assets clear 12 to 20x revenue, data and MLOps platforms 6 to 9x, infrastructure-adjacent software 6 to 10x, and enterprise AI applications 3 to 6x. Proprietary data, unit economics, and retention move your number most.
Strategic acquirers such as hyperscalers, enterprise-software and security platforms, and data and chip incumbents; private equity platforms buying applied-AI vertical software; and growth and crossover sponsors backing model and infrastructure leaders. See our guide on who buys AI companies for the full buyer list.
AI is the most mispriced category in technology M&A. The same revenue can be valued at mature-software multiples or at double-digit revenue multiples depending on layer, data advantage, and which buyer is at the table. General technology M&A advisory tends to anchor on software comps; AI requires an advisor who can evidence a data or model moat, defend unit economics after inference cost, and reach the strategic buyers paying for capability rather than cash flow.
Yes. We work across the AI stack, applied and vertical AI software, data and MLOps platforms, machine learning and AI infrastructure, and model and research teams. Each layer has a distinct buyer pool and valuation logic, and the positioning is built to the layer.
When your unit economics and retention can be evidenced, your data or model advantage is clear, and strategic buyers are actively paying for capability in your category. Selling into live buyer appetite is the difference between a premium and a clearing price.
Most sell-side processes run several months from preparation to close. The preparation carries the value: positioning, buyer mapping, and clean diligence materials should be finished before launch, not improvised during the process.
If it wants competing bids, yes. Generative AI and model assets price at 12x to 20x+ revenue on this page's indicative ranges, but the same revenue can be valued at mature-software multiples depending on data advantage and which buyer is at the table. A sell-side advisor approaches buyers under NDA and runs them in parallel, so price reflects competition rather than a single bid, then negotiates cash at close, rollover, retention and earnout terms.
On this page's indicative ranges, AI infrastructure and infrastructure-adjacent software price at 6x to 10x revenue, with compute economics and ecosystem position driving the premium. Strategic buyers here include hyperscalers and data and chip incumbents acquiring capability they cannot build fast enough internally. Growth and crossover sponsors also back infrastructure leaders, often as the bridge between a private process and a larger strategic outcome.
Applied, enterprise and vertical AI applications price at 3x to 6x revenue on this page's indicative ranges, below data, infrastructure and model assets. Buyers underwrite retention and gross margin after inference cost first, including whether customers stay once the novelty passes. Private equity platforms buy applied-AI and vertical software with real revenue, retention and a path to margin expansion, and strategic acquirers buy capability they cannot build fast enough.
The headline price is one term among several. Windsor Drake negotiates cash at close, rollover, retention and earnout terms, then manages diligence through signing, with the aim of converting a flattering headline into the best achievable terms. Buyers also underwrite team and IP, the talent and technology they would struggle to assemble independently, so the structure has to account for the people a buyer is paying for.
Windsor Drake runs sell-side processes for founder-led companies. Conversations are confidential, and you will get a plain answer on whether a process makes sense now.
Windsor Drake is not a registered broker-dealer and does not offer, sell or place securities. When a transaction requires a registered broker-dealer, that work is handled by a registered broker-dealer partner.
