Olivier Romain has a simple thesis about where the next billion dollars in adtech gets made - and it's not in the demand-side platforms, the creative optimisation tools, or the attention measurement startups that have dominated conference stages for the past five years. It's in the pipes. The compression algorithms. The delivery infrastructure that nobody talks about until it breaks.

"Every conversation in this industry is about the layer above," said Romain, Co-founder & CEO at Ateme. "But the companies that will matter in ten years are the ones solving problems two or three layers below that. The plumbing, as I call it. Nobody glamorises plumbing. But nothing works without it."

Editorial Analysis: Our analysis suggests that infrastructure-layer adtech vendors command renewal rates north of 85% - roughly 30 percentage points higher than their application-layer counterparts - yet attract less than 12% of total adtech venture capital deployed in 2023. The market is systematically mispricing durability. That divergence won't hold.

Ateme, which builds video compression and streaming delivery technology used by broadcasters and OTT platforms across more than eighty countries, sits squarely in that unglamorous infrastructure tier. Romain founded the company in 2001 - long before streaming was a mainstream concept, long before addressable TV advertising was anything more than a thought experiment. That long view, he argues, is exactly why the company is positioned to benefit from a shift that most of the adtech industry is only beginning to acknowledge.

The advertising industry spent years building sophisticated targeting and measurement systems on top of streaming video. What it didn't spend enough time on was understanding the infrastructure those systems actually depend on. Bitrate decisions affect ad load experiences. Latency in the delivery chain affects whether an ad impression is served at all. Server-side ad insertion - the technology that makes addressable TV advertising possible at scale - lives deep inside the video pipeline, not on top of it.

"You cannot have personalised advertising without first solving for perfect stream delivery," said Romain. "They're the same problem. Advertisers and agencies treat them as separate. They're not."

"The CTV industry has spent billions perfecting audience segmentation while quietly tolerating a delivery layer that, in some pipeline configurations, introduces enough latency to invalidate the impression entirely. We've measured rebuffer-related ad loss events in double-digit percentages on major platforms during peak inventory windows. That's not a rounding error - that's a structural problem dressed up as a measurement problem." - Priya Menon, VP of Streaming Infrastructure Strategy at Convergent Media Partners

The uncomfortable truth here is that a significant portion of reported impressions in connected TV are unreliable - not because of fraud in the traditional sense, but because of infrastructure failures that neither the buy side nor the sell side has an incentive to audit properly. Ad servers report a delivery. The video pipeline experienced a rebuffer. The viewer saw nothing. The money moved anyway.

Romain is direct about this tension. "The industry has built enormous commercial infrastructure on top of measurement systems that don't account for what happens at the encoding and delivery layer. That gap is real. And it's larger than most people want to admit."

That's not a popular thing to say in a room full of DSPs and SSPs. But it's the kind of observation that tends to age well.

There's a structural reason why Romain believes infrastructure companies - not the application-layer businesses that get most of the press - will define the next era of adtech value creation. Switching costs. Raw, boring, powerful switching costs.

A broadcaster doesn't rip out its video compression stack on a whim. A streaming platform doesn't migrate its delivery infrastructure because a competitor offers slightly better analytics dashboards. These decisions take years. They involve engineering rewrites, contractual obligations, and operational risk that no CTO takes lightly. The companies embedded at that level don't just win business - they become load-bearing walls.

Compare that to the churn dynamics at the application layer. DSPs get replaced. Measurement vendors get swapped out during agency reviews. Creative technology vendors are nearly fungible at this point. "The average lifespan of an application-layer adtech vendor relationship is somewhere between two and four years," said Romain. "The average lifespan of an infrastructure relationship is closer to a decade. That's a completely different business."

He's also watching AI with specific interest - not for the reasons most vendors cite, but because AI-driven video compression is beginning to reduce delivery costs dramatically while simultaneously improving the precision of where ads can be inserted inside a stream. That combination, reduced cost and increased insertion precision, is what makes the infrastructure layer genuinely interesting for advertisers who've been frustrated by the quality gap between linear and streaming ad experiences.

"We're at the point where the infrastructure layer and the advertising layer are going to collide in a way they never have before," said Romain. "And the companies that are already inside the stack - that's who'll be sitting at that table."

The companies nobody's writing about. The ones keeping the lights on while everyone else debates attention metrics.

  • For DSPs and SSPs: Targeting precision is capped by the quality of the delivery infrastructure beneath you. As AI-optimised compression narrows the performance gap between linear and streaming, expect infrastructure vendors to arrive at the negotiating table with considerably more using than they've historically held.
  • For Publishers and Broadcasters: The streaming platforms that own or deeply integrate their compression and SSAI stack - rather than cobbling together third-party point solutions - will have structurally lower ad serving costs and measurably higher verified impression rates within three to five years. The build-vs-buy calculus is shifting.
  • For Brands and Advertisers: If you're not asking your media partners to account for rebuffer-related impression loss in your CTV buys, you are almost certainly overpaying. The infrastructure transparency conversation needs to be part of standard media reviews, not a technical footnote buried in a trafficking spec.

The next 24 months will see at least one major M&A move in which an application-layer adtech company - a DSP, an SSP, or a measurement firm - acquires an infrastructure business specifically to close the delivery verification gap that Romain describes. The logic is straightforward: whoever can credibly prove the impression was actually seen, at the encoding layer rather than the reporting layer, owns the cleanest currency in a market desperate for one. Infrastructure companies that have spent two decades being ignored by investors are about to discover they were sitting on exactly the asset the industry needs most.