Palantir’s Data Stack Is Reshaping Performance Marketing

TL;DR: Palantir has moved from defense contracts into media and marketing data infrastructure, signing major deals with Fox News Digital, Zeta Global, Stagwell, and USA Today. Its forward-deployed engineers build alongside client teams rather than just advising, compressing timelines from months to weeks. For performance marketing operators, this signals a structural shift in how commercial data is organized, queried, and monetized.

From Government Intel to Commercial Data Stack

Palantir built its name on classified government work — contracts with the Defense Department, ICE, and intelligence agencies. That reputation created a polarizing public image. But the business underneath that image has been quietly executing a pivot into commercial data for the past several years, and the results are showing up in marketing infrastructure at scale.

In December 2025, Fox News Digital signed a deal with Palantir to build an internal production platform that now handles 95% of its published articles. In June 2026, Zeta Global locked in a seven-year partnership and rebuilt its entire data cloud on Palantir’s commercial platform, Foundry. Stagwell built two marketing products on Foundry in 16 weeks. USA Today announced a deal in August to analyze and monetize its audience data using Palantir’s systems.

These are not pilot programs. These are core infrastructure decisions made by companies managing hundreds of millions of consumer records. And they all point to the same conclusion: Palantir is positioning itself as the connective tissue between fragmented marketing data and the AI systems that need to act on it.

The Ontology Problem Nobody Else Solved

Media and marketing companies have accumulated enormous data estates that are nearly impossible to use at speed. Zeta Global works with 550 million opted-in individuals and processes more than 7,500 computations per millisecond at peak. Stagwell’s data spans 1 billion touchpoints across 140 million U.S. households.

The core problem is not volume — it is structure. Palantir pitches what it calls an “ontology”: a single, organized model of a company’s data and operations that both people and AI systems can query and act on. Stagwell’s chief AI officer John Kahan, who previously led data and analytics at Microsoft, frames the shift clearly: “The world spent probably $1 trillion trying to create structured data. When LLMs came out, all of a sudden you could reason with unstructured data. Now you can go against all of that: campaign briefs, branding data, brand insights.”

For operators who have spent years trying to reconcile CRM exports, ad platform data, and first-party behavioral signals, this is a direct answer to a problem that BI tools and data warehouses have failed to solve. The ontology layer does not replace your models — it sits between your data and whichever model works best for a given task, whether that is Azure, Google, or something newer.

This model-agnostic architecture is what separates Palantir from AI vendors selling proprietary model lock-in. Kahan’s framing is direct: “We want to be agnostic to the LLM. If someone comes out with a new model, we switch the model.”

Speed Is the Product

Every executive interviewed for this story returned to the same point: Palantir moves faster than internal teams can. Stagwell built a working agentic marketing system in 16 weeks. USA Today’s CEO Mike Reed told investors that timelines requiring considerable internal effort can compress to weeks with Palantir embedded. Zeta’s test was so convincing it skipped a competitive review entirely.

The mechanism behind that speed is Palantir’s forward-deployed engineers — technical staff who embed with client teams and build alongside them rather than delivering specifications and walking away. At Fox News Digital, those engineers shadowed the newsroom, sitting with reporters and editors before writing a single line of code. The resulting platform, called Newsroom, now handles the dozen post-writing tasks every article requires: video, links, photos, tagging. It produces 95% of Fox News Digital’s articles. Nothing reader-facing is AI-generated, but the operational infrastructure is entirely Palantir-built.

Fox News Digital president Porter Berry’s description of the approach is worth quoting directly: “They didn’t just come in and advise. They came in and built something with us that will endure over time.”

For operators evaluating paid media program management, that distinction matters. A vendor who embeds and builds is a different category of partner than one who delivers a dashboard and a Slack channel.

What This Means for High-CAC Vertical Operators

Operators in forex, iGaming, crypto, and legal run campaigns where a single qualified lead can be worth $300 to $3,000+. At that cost-per-acquisition, the quality of audience data is not a marginal variable — it determines whether a campaign is profitable or not.

Palantir’s model points toward a future where the data layer is unified, queryable in plain language, and connected directly to campaign execution. Stagwell CEO Mark Penn describes the goal as “the holy grail of marketing” — the ability to answer specific, real-time questions: which customers want a specific product, where they live, which brands they respond to, and when to reach them. That capability is not theoretical. Stagwell built a working version of it in four months.

For iGaming acquisition programs, this maps directly to geo-segmented deposit campaigns where audience overlap with sports content or financial news is a strong purchase-intent signal. For forex lead generation, it means connecting trading behavior signals to media consumption patterns at a scale that current DSP targeting cannot reach. For law firm intake campaigns, it means identifying anonymous site visitors and converting them to known relationships before they scroll to a competitor.

USA Today’s strategy around falling search traffic makes the same point from the publisher side: “Every visit, every session, and every moment of attention creates a signal. When we connect those signals, they become actionable intelligence.” Operators sitting on first-party data from lead forms, app installs, and CRM records should be reading that sentence as a description of their own untapped asset.

A structured data and channel audit is often the first step to understanding what you actually have before investing in infrastructure to activate it. Operators who have not mapped their own data estate will find that any new tooling — Palantir or otherwise — underperforms without that foundation.

Palantir as a Sales Channel

The commercial model Palantir is building goes beyond software licensing. For Zeta and Stagwell, Palantir is also a distribution channel. Palantir takes a revenue share on every deal Zeta closes with clients sourced through Palantir’s network of 600+ commercial customers. Stagwell goes to market alongside Palantir, with its products driving Foundry usage.

This creates a mutualism that is worth watching. Palantir gains vertical expertise — advertising, media, audience monetization — that it openly admits it does not have internally. Partners gain access to a pre-qualified client network that spans defense contractors, healthcare systems, shipping logistics firms, and financial services operators. Zeta’s CEO David Steinberg framed the term length of their deal as a negotiation: “We ended up with seven [years] because we wanted 10 and they wanted five.”

That kind of commitment from a company managing 550 million consumer records is not a vendor endorsement — it is a structural bet on where commercial data infrastructure is heading.

For operators running crypto acquisition campaigns or scaling precision audience targeting across multiple markets, the relevant question is not whether Palantir itself is the right tool. The relevant question is whether your current data stack can answer the same kinds of specific, real-time questions that Stagwell and Zeta are now running as standard operations. If it cannot, that gap is widening every quarter.

What to Watch Next

More than 800 USA Today employees across 31 unionized newsrooms pushed back on the partnership, citing data security, reader trust, and editorial independence. Those concerns are legitimate and reflect real tensions in how surveillance-adjacent technology gets embedded into media operations. They did not stop the deals — Stagwell’s Mark Penn said “there was no concern about reputation risk” — but they will shape how these partnerships are disclosed and governed going forward.

For operators outside the media industry, the governance question translates to data compliance: GDPR, CCPA, and financial services data handling rules all apply to how audience signals get ingested, stored, and actioned. Any infrastructure investment at this scale requires legal and compliance review before deployment, not after.

Palantir’s expansion into commercial data is real, accelerating, and backed by seven- and ten-year commitments from sophisticated buyers. Whether you use Palantir or a competing stack, the underlying shift — toward unified, queryable, AI-ready data infrastructure — is the operating environment performance marketing operators will be competing in for the next decade.

Originally reported by Adweek, September 2026.