AI Token Costs, Principal Media, and Data Law: AdTech's Triple Squeeze

Three separate stories landed this week that, read together, describe the same underlying pressure: the cost and control architecture of digital advertising is being renegotiated simultaneously from three directions. AI infrastructure pricing, holdco margin engineering, and data regulation are not parallel tracks — they’re converging. If your media and data strategy was built for 2024’s operating environment, you’re already behind.

AI Token Costs Are Now a Media Budget Line Item

AdExchanger reports that large agencies are actively wrestling with how to offset the rising cost of token consumption as AI tooling embeds deeper into campaign operations — from creative generation to audience modelling to reporting synthesis. OpenAI has also begun outlining policies for how brands can apply ad credits toward campaigns, which signals a structural shift: AI platforms are not just vendors anymore, they’re becoming media inventory holders in their own right.

The strategic implication is underappreciated. Token costs scale with usage in a way that traditional SaaS licensing doesn’t. An agency running AI-assisted planning across dozens of clients can see compute costs compound faster than the efficiency gains justify — particularly when the AI outputs still require human review cycles. Brands in Southeast Asia running regional campaigns with multilingual creative variants face this acutely: each language permutation multiplies token load. The discipline of prompt engineering is becoming a legitimate cost-control function, not just a productivity trick.

The smarter holding groups are already building internal token budgeting frameworks — treating AI compute the same way they once treated data licensing: as a variable cost that needs governance, not just access.

Principal Media Is Holding Companies’ Most Honest Business Model

Digiday’s Michael Bürgi surfaces something the holdco earnings calls tend to obscure: while AI tools dominate the announcement cycle, principal media — where the agency buys inventory at wholesale and resells it at a margin — has become a quietly dominant profit driver. Holdcos are deploying increasingly sophisticated incentive structures to push their trading desks and client teams toward principal arrangements, precisely because the margin is more predictable than fee-based AI service revenue.

The tension is real and growing. Agencies are simultaneously positioning themselves as AI-powered performance partners while running a media arbitrage model that predates programmatic. These aren’t incompatible, but they create an alignment problem: the AI tools optimise for client outcomes, while the principal inventory model optimises for agency margin. When those two objectives diverge — and they do — the client is usually the last to know.

For marketing directors in Southeast Asia, this is particularly worth watching. Regional holdco desks often bundle principal inventory from platforms like Lazada’s DSP, Grab Media, or regional open-web supply into managed service packages where margin opacity is high. Asking your agency to disclose whether they hold principal positions on any recommended placements is no longer a paranoid question — it’s basic procurement hygiene.

---

New Jersey’s Data Broker Law Is Live Right Now — Not Eventually

The timeline here is genuinely unusual. New Jersey’s data broker legislation was introduced on June 28, signed into law on June 30, and — per AdExchanger’s Allison Schiff — took effect immediately, with only the registration deadline pushed to 2027 after trade group pressure. As Charlie Simon, VP of Private Advertising, noted, the speed made CCPA’s own fast-tracked passage look measured by comparison.

For AdTech infrastructure teams, the immediate operative question isn’t about New Jersey residents specifically. It’s about what this signals for the pace of regulatory action as US states move from deliberation to execution. The registration delay gives data brokers breathing room on the compliance paperwork, but the substantive provisions — governing how consumer data is collected, sold, and used — are already in force.

If your identity resolution stack, clean room configuration, or third-party data partnerships involve brokers who aggregate US consumer data, those relationships need legal review now. The Southeast Asia angle: regional brands running performance campaigns into US markets through programmatic channels are often surprised to discover that their data supply chain touches brokers operating under exactly these frameworks. Ignorance of the upstream data lineage is not a compliance defence.

The broader pattern matters more than any single state law. Regulators are no longer signalling intent — they’re acting. The window for leisurely compliance roadmaps is closing faster than most MarTech teams have planned for.

Reading the Three Stories Together

Strip away the individual headlines and you see one argument: the infrastructure layer of digital advertising — who controls the AI compute, who holds the inventory, who owns the data relationships — is being contested and repriced in real time. Brands that treat these as separate vendor management questions will find themselves reacting to each development individually. The ones that recognise them as a single question about where value and risk sit in their supply chain will be better positioned to negotiate.

The playbooks for this environment are genuinely still being written. Anyone telling you they have it figured out is selling something — probably on principal.

---

Key Takeaways

• Treat AI token consumption as a managed budget line with governance policies, not an open infrastructure cost — especially for multilingual, multi-market campaign operations.
• Ask your agency explicitly whether they hold principal inventory positions on any recommended placements; in Southeast Asia’s bundled managed-service environment, this question is overdue.
• New Jersey’s data broker law is operative today, not in 2027 — audit your third-party data supply chain for broker relationships before your legal team has to.

The deeper question this week’s developments raise isn’t tactical. It’s structural: as AI platforms become media owners, agencies become inventory principals, and regulators move at legislative speed, what does an independent, auditable media supply chain actually look like? That’s not a question with an easy answer yet — but the brands asking it now will have significantly more leverage than those who wait for the industry to settle.

---

At grzzly, we work with regional brands navigating exactly this kind of infrastructure uncertainty — from clean room architecture and identity resolution to supply chain audits and AI tooling governance. We’re comfortable in the contested territory where the frameworks are still being written. If any of this is landing close to conversations you’re already having internally, Let’s talk.