Publicis is Buying More Than People Realise

Quick disclaimer up front. Nothing shared in this post is proprietary or confidential information, and all views expressed here are entirely my own rather than those of my former employer. This also went quite long without me covering everything I had hoped to, so there may be a part 2 😅

Yesterday Publicis announced they’re acquiring LiveRamp for around $2.2 billion. $38.50 a share, all cash, roughly a 30% premium to the close. The deal is expected to close before the end of 2026.

Bias declaration up front. I joined LiveRamp in early 2024 via the Habu acquisition, and spent the next two years operating as a global specialist on the clean room product across UK, EU, APAC and LATAM. I had plenty of exposure to our agency deals and had the pleasure of working with Publicis teams across multiple markets during that period. So this is one of those rare deals where I actually have something to say from the inside, rather than the outside looking in.

Let me work through what I think the market’s getting right, what it’s getting wrong, and what the more interesting questions are.

The multiple is low, and that tells you something

LiveRamp was pushing on $850M of ARR. $2.2 billion on that is roughly 2.6x revenue. By any reasonable standard for a technology business with that kind of strategic position, the multiple is low.

There are two things going on. The first is that adtech as a category trades at a discount to the rest of enterprise SaaS, and has done for years. Even The Trade Desk, the cleanest pure-play in the space, sits at single-digit revenue multiples right now. The market has structurally undervalued anything with “ad” attached to it since around 2022, and that drag applies to LiveRamp whether it’s fair or not.

The second is more specific. LiveRamp carries an inherent regulatory risk premium because so much of the underlying value depends on the ability to resolve and onboard identity… which is exactly the area that GDPR, CCPA, state privacy laws, and Apple and Google platform changes have been chipping at for the best part of a decade. Investors haven’t been wrong to price risk in there. They’ve arguably been wrong about whether LiveRamp has actually adapted.

Also it is worth noting a personal data point on this one. When LiveRamp acquired Habu in early 2024, the deal converted at an effective LiveRamp share price of roughly $40. Two years on, with whatever theoretical value accretion and ARR growth has happened in between, the company is being sold at $38.50. Lower, not higher. Which says quite a lot about where the public market has had LiveRamp pegged across the hold period, and reinforces the point that the buy-side multiple here is meaningfully below what the underlying technology and network warrant.

The identity crisis paradox

Let me point out the obvious thing whilst we’re here… LiveRamp is a company whose entire brand is literally about identity, and it’s fair to say that it has spent years suffering from something of an identity crisis of its own.

That’s not a dig. It’s an observable fact about how the company is perceived in different markets and industries. In the US LiveRamp is still largely seen as THE deterministic onboarder, ID graph and data marketplace. In the UK and EU though it’s more strongly associated with Safe Haven and a measurement partner, and for the clean room product. In APAC maybe it’s some mix of all of the above plus the post-Habu cloud-native pivot as countries like Australia leaned in hard there. Different markets, different stories, different toolkits and therefore different perceptions of what the company actually is in 2026.

A clear illustration of this came in amongst the Whatsapp chatter when the deal broke yesterday, whereby one person’s take, more or less verbatim, was “$2bn for something that can’t see iOS.” This is an opinion rooted in a different decade. It’s the kind of thing that might have been a credible criticism in 2017 but it is not a credible criticism in 2026. And yet it’s still alive in the heads of senior people in the holding companies (who now more firmly compete with LR), and which frames exactly the perception problem I’m describing.

What I found when I got inside

I joined via the Habu acquisition in early 2024 with my own preconceptions. 2014 Tim sitting at Krux, watching this big US data onboarder grow on the back of CRM-to-cookie identity resolution, had a very particular view of what kind of company LiveRamp was. By 2024, and on my entry into the firm, that view was wildly out of date. The version of LiveRamp that I encounteres on the inside was unrecognisable from the one I’d carried in my head.

A few things genuinely surprised me:

Safe Haven was built EU-out, not US-first. It was designed in direct response to GDPR with European regulatory constraints as a first-class concern, then exported into the US later. That’s actually pretty rare for a US-headquartered platform of LiveRamp’s scale. Almost everyone else builds in the US, retrofits for Europe afterwards, and discovers the retrofit doesn’t quite work. The tail wagging the dog, in a healthy way.

The DataFleets acquisition was a bold early bet. LiveRamp paid $68 million for DataFleets in February 2021 for federated computation, differential privacy, and the broader privacy-enhancing technology stack. That was an early move on owning the architecture that’s now basically table stakes for clean rooms. Five years on, even though DF didn’t become the backbone of how LR’s tech operates, it looks like one of the more prescient acquisitions in the category.

Habu wasn’t a defensive purchase. There’s a version of the story where LiveRamp bought Habu in 2024 to neutralise a competitor and slow it down. That isn’t what happened. Habu was bought to be the next-generation, cloud-native, decentralised collaboration layer, and the team I worked alongside through 2024 and 2025 was being asked to actively inform the broader platform direction, not just sit there and get absorbed and shelved.

Privacy overruled commercial every time, as it should. There’s a persistent misconception in the market that LiveRamp skews heavily towards commercial imperatives at the expense of privacy ones. The reality I found on the inside was the opposite. The diligence and rigour applied to tech and data privacy review processes was second to none in the industry.

If you sat on the LiveRamp side for any length of time you’d notice the same thing I did. The company has been quietly upgrading and reinventing its core for years. The external brand perception and understanding of this just hasn’t quite kept pace which is a shame.

The accrued moat is the real prize

This is the bit I think is most underappreciated in the early coverage of the deal.

LiveRamp has spent roughly a decade building a vast network that would be genuinely difficult to replicate. Every major walled garden has signed onto the RampID framework and offers impression-level analytics and inbound targeting on it. Every major DSP and SSP is integrated. Premium web and CTV publishers all route their authenticated traffic signals through RampID. This is a network effect that’s taken a long time to build and would take even longer to rebuild from scratch in today’s world. You can’t vibecode this.

So Publicis isn’t buying ~$850M of ARR for $2.2B, they are buying the only piece of independent identity infrastructure in the market that has actually managed to sign every important counterparty. ARR is reallyt a lagging indicator of the value of this deal. Instead the contracts, the integrations, and the counterparty trust are one of the biggest assets, provided they all stick.

The AI angle

A significant portion of LiveRamp’s revenue over the last decade has come from the data marketplace. Prepackaged, taxonomized segments that belong to one party, licensed to another, intermediated by LiveRamp’s onboarding and resolution layer. That model has worked well for everyone involved. It’s also looking increasingly legacy.

The world is shifting from “which segment do I want to buy?” to “how do I connect my model to the data that would make it smarter?” Shipping a segment was always a fairly transactional commercial conversation. Connecting a model to data is a different shape entirely. It surfaces the same trust, commercial and privacy concerns that data owners have always had about who gets to see what, but at a much higher level of stakes, because the model itself becomes the asset of value rather than just the extract you pull from a query.

Data owners aren’t going to ship their data to whoever happens to have the model. They’re not going to license it raw to a holding company. What they will increasingly do is let a model come to them in a controlled environment, run on their data without anyone seeing the underlying rows, and ship only the resulting parameters, embeddings or predictions back out. That model-to-data architecture is exactly what the last decade of work on secure multi-party computation, confidential computing, and federated learning has been building towards. It’s also a category that has lacked a credible commercial owner at scale.

This is where former M&A investments and continued product enhancements start looking like the spine of LiveRamp’s next revenue model. Clean rooms are evolving from segment-sharing and audience overlap pipes into places where in theory models meet data without anyone having to hand over the underlying data itself. The infrastructure already inside LiveRamp from acquisitions, plus the regulatory grounding from Safe Haven, plus the network that already exists at scale… that’s a stack you could plausibly build a “data co-creation for smarter agents” business on. Which is, incidentally, exactly the language Publicis chose for the press release.

The leading edge in this deal might just be the privacy-preserving compute layer that connects models to data that data owners would never otherwise share. Most of the early coverage I’ve seen has missed this, and it will lbe critical given trust issues inherent when sharing data with agencies.

Does it raise a question about Lotame?

Publicis announced the Lotame acquisition in March 2025, part of a roughly $1.5 billion M&A run, and Lotame was positioned at the time as the international identity expansion of Epsilon. Fourteen months later they’re spending another $2.2 billion on LiveRamp, which is, at minimum, an overlapping set of capabilities.

I don’t have inside information on what Publicis is thinking here, but the question is hard to avoid. Has the Lotame integration delivered what the original deal thesis required, and is LiveRamp the answer to a gap that opened up? Or are these two complementary solutions / networks that, when taken together, cover something neither could cover alone? Or maybe the AI agent thesis (the press release language is “data co-creation for smarter agents”) genuinely shifted the goalposts to the point where the requirements changed mid-flight?

All three are plausible but the first is the most uncomfortable. The way the deal gets articulated in client meetings over the next Q will tell us all which one it actually is.

What I think the deal is actually about

If I strip out the press release language, the way this deal makes the most sense to me is roughly this.

Publicis is the holding company that has been most aggressive in positioning itself as the data and AI play among the big agencies, on the back of Epsilon, Sapient, Lotame and now LiveRamp. The strategic question at agency level for the next five years is whether the holding companies can own the infrastructure underneath all of this… identity, privacy-preserving compute, the place where models meet client data… or whether they end up renting it from someone else.

LiveRamp at $2.2 billion buys Publicis a very widely adopted and contracted identity network they otherwise have to lease (and which they almost certainly do pay for today - LR touts having commercial relationships with ALL the holdcos), and it puts the Habu and DataFleets engineering and product nous directly inside the business. That’s worth significantly more strategically than just ARR multiples, and it’s probably also worth more than the market has currently priced into the Publicis side of the trade.

This is the most strategically important acquisition in the agency holding company space since Publicis bought Epsilon back in 2019. And as someone who spent the last two years inside the company being acquired, I’d say the technology asset being purchased is meaningfully better than the public market thinks it is.

Last one out?

Or perhaps this deal is a bellwether for the last of the current generation of adtech and identity players exiting before we move full-fledged into the AI era… which of course could look really quite different really quite fast.

For Publicis, dropping $2.2 billion on LR right now is’t a hail mary. The risk is low and, as I have discussed, the deal comes with the right people, paper, and technology to make it something of a no-brainer.

With the crazy pace of change across all sectors due to AI, there are questions still to be answered and that could probably warrant posts of their own:

• What will the agency - client landscape look like in 6 months when the deal is scheduled to close?
• Does one of the other holdcos have it in them to try and entirely leapfrog this with a “next-gen” play? If so what does that actually look like?
• Will advertisers accept further centralisation of power around Publicis, or simply push back?

The AI-native advertising and marketing stack is developing rapidly, and is underpinned by an entirely different set of primitives. More thoughts on that in another post 👀