The Future of the Open Web With Aly Nurmohamed and Christer Ljones

Full analysis

The pitch on this episode is clean: open-source AI turns the model layer into a commodity, so the only durable advantage left in advertising is first-party data owned by publishers with direct consumer relationships. Aly Nurmohamed of Nodals AI and Christer Ljones of Schibsted want you to believe premium publishers can build a Meta-style outcome engine on their own logged-in audiences and stop renting the pipes. It's a real argument, and it's also a founder selling the shovels. Both things are true at once.

What's actually being decided here, for an operator, is where you place your bets over the next five years: keep leaning on programmatic intermediaries for premium inventory, or invest to own the AI-plus-first-party-data stack yourself. That's a hard-to-undo call. Building authentication, data infrastructure, and model capability takes years and real headcount. Sitting it out is easy to reverse. That asymmetry matters.

The Market Analyst. The through-line connecting this episode to the reading list is consolidation. Ljones predicts publishers grouping around definable segments, geography, vertical, passion category, with scale as the price of entry. The Infillion/Foursquare deal on the same day says the ad-tech side is aggregating data assets for exactly the same reason. Both sides of the chain are concentrating. For an informed generalist: the middle of the market is being squeezed from both ends. The winners are the very large and the very specialized. If you run a mid-tier SSP or DSP whose value is moving premium-publisher inventory, this thesis, if it scales, shortens the supply path right past you. That's the structural read worth holding.

The Skeptic. Steelman the case against, and it's not hard. Schibsted is a Nordic login-first media group with a decade head start on authenticated audiences. Nurmohamed's headline evidence, that Criteo's conversions "dropped off a cliff" after Schibsted offboarded, is a proprietary anecdote from the man who now sells the publisher-empowerment story. Unauditable, and directly self-serving. The $138 billion Meta capex figure Ljones cites is off by roughly double against Meta's reported guidance. When your case rests on one un-checkable conversion story and one wrong capex number, be careful. The thesis needs premium context to beat retargeting AND publishers to actually build the AI. Nurmohamed is quietly conceding the second half is hard, which is why Nodals exists.

The Operator. Try to run this Tuesday morning at a normal publisher. You need logged-in users at scale, a data pipeline that feeds a model, and direct advertiser relationships that will transact on outcomes instead of CPMs. Most publishers have none of the three. Schibsted spent since 2019 cutting from thousands of vendors to 30-50, and they had login behavior baked into the product for years. The part that breaks first is authentication: no logins, no linkable first-party data, no model worth training. The part that breaks at 90 days is sales. Your direct team knows how to sell inventory, not guaranteed outcomes. Meta can promise "give us the budget, we'll do the rest" because it controls creative, targeting, and environment. A single publisher controls one environment.

The Customer / End User. From the buyer's chair, outcome-based buying from a premium publisher is appealing only if the outcome clears. Ljones frames the opening as Meta going from three times better to maybe 25-30% better than a good publisher. Fine, but 25-30% worse at scale is still worse, and buyers chase efficiency. The genuine draw for advertisers is Nurmohamed's placement point: a premium editorial environment converts better than the same user found cheaply somewhere else. If that holds, buyers will pay for context. But they will demand proof per campaign, not an anecdote from 2019. No publisher gets outcome budgets on trust.

The CFO. The economics only work above a scale line most publishers sit below. Ljones's own out is the evidence on the math: shed the long-tail and problematic inventory, keep 85-90% of revenue with far less operational and regulatory risk. That's a margin play, not a growth play. You're trading messy programmatic dollars for cleaner, fewer, higher-quality ones. The upfront cost is model infrastructure and authentication you build before the revenue shows. EMARKETER's Nate Elliott puts AI advertising at $32 billion in the US this year, so the pool is real. The payback question is whether your slice of authenticated audience is big enough to justify building versus buying the capability from a vendor like Nodals.

Where they part ways. The first split: is the moat first-party data, or is it scale? Nurmohamed says data. Ljones, in the same breath, says you need dominant scale in a segment or you consolidate. Those aren't the same claim, and the gap matters, because most publishers have neither. The second split: is Meta a template or a wall? Both Nurmohamed and Ljones call it inspiration. But Meta controls the whole loop, creative, targeting, environment, and spends more on infrastructure in a year than most publishers earn in a decade. A single publisher can copy the pitch. The machine is out of reach. The third: build or rent. The whole episode argues models are commoditized, then Nurmohamed sells you the reason you still need a vendor to use them.

What it hinges on. Whether premium context actually delivers enough outcome lift to command outcome-based budgets, and whether enough publishers have authenticated scale to build on. The council leans skeptical on breadth and sympathetic on direction. For a handful of login-first, large-segment publishers, this is a genuine path. For the mid-tier that is most of the open web, it's aspirational, and the intermediaries they'd disintermediate aren't going anywhere soon. Before committing real capex, verify one thing the episode never proves: run a controlled test of your own premium inventory against retargeted reach and measure post-click outcomes. If context lift is real in your data, build toward authentication. If it isn't, you're funding a founder's slide. Prediction: At least two more data-asset acquisitions of the Infillion/Foursquare type will be announced by Q2 2027 earnings season, while no non-login premium publisher outside the Schibsted tier launches a Meta-style outcome-based direct ad product at scale in the same window.

Confidence: Medium — the consolidation half is well-supported by deal flow already in motion; the publisher-inertia half depends on a build most publishers won't finance before the revenue is proven.

Why: The Infillion/Foursquare deal and the episode's thesis point to the same pressure: scale in first-party and location data is now the price of competing, and players who lack it are acquiring rather than building. That pressure does not stop after one deal, because distressed and strategic ad-tech data assets keep coming to market and aggregators have signaled appetite. On the publisher side, Schibsted's path required a decade of login-first product investment and a deliberate vendor cull that most mid-tier publishers have not started. Without authenticated scale already in place, no publisher can close the outcome-budget gap against Meta before mid-2027, regardless of how cheap the model layer gets.

Revisit by 2027-06-30: We're right if at least two more acquisitions centering on first-party or location data assets close by Q2 2027 earnings reports, and no non-Schibsted-tier open-web publisher publicly reports outcome-based direct budgets at material scale. We're wrong if a mid-tier publisher outside the Nordic login-first model launches an audited outcome product that attracts meaningful direct advertiser spend, or if the M&A pace stalls at the current single deal.