Zeta's AthenaOS Land Grab: The Marketing Cloud Grew a Second Act
A 21st straight beat-and-raise was the headline; the real story is a headless-architecture reversal, eight Palantir yeses, and Google search budgets finally moving.
ZETA · Earnings Call · 2026-11-09
The Beat Was Loud; the Narrative Changed
Zeta's Q3 was a blowout by any measure: revenue $493 million, up 46% (30% excluding Marigold and political candidate revenue), adjusted EBITDA $121 million at a 24.6% margin, free cash flow $82 million, and GAAP EPS of $0.14. Management called it the 21st consecutive beat-and-raise and raised full-year revenue to $1.855 billion. But the numbers were not the surprise. The surprise was how far CEO David Steinberg pushed the story beyond marketing software. return on marketing spend remains the engine, but the new pitch is enterprise intelligence — using Zeta's Data Cloud and AthenaOS to solve problems in inventory, supply chain, fraud detection, even cybersecurity. The Q3 call introduced an AI scorecard that is now central to how management wants investors to value the business. AI scorecard metrics showed over 50% of super-scaled customers were monthly active Athena users, up from just over 40% in Q2. Athena adopters accounted for more than 70% of revenue. The top 25 Athena users grew revenue more than 70% in Q3 versus 46% company-wide. And net revenue retention for Athena adopters ran over 500 basis points above overall Zeta. This is not boilerplate AI enthusiasm; it is a disclosed usage funnel that management says it also uses internally as a forecasting tool.Athena Is No Longer a Demo
A year ago Athena was a promise. In Q2 2026, management was still selling the early innings. On that call, Chris Greiner noted that “heavy users had a 400 basis point higher net retention rate than our other clients for Athena today.” — David A. Steinberg, Chief Executive Officer (CEO) · 2026-08-04 Three months later the premium is 500 basis points. That progression matters. It suggests Athena is not just a UI layer but a consumption expander. Steinberg described the dynamic bluntly: clients who adopt Athena are “moving budget faster than at any point in our corporate history” — David Steinberg, Chief Executive Officer · 2026-11-09. Client campaign development more than doubled quarter over quarter, which management treats as a top-of-funnel indicator for activation revenue. The financials confirm the operating leverage. In the last filed quarter (Q2), revenue was $443 million, up 44% year over year, while operating margin reached 3.8% and net income turned positive at $8 million. The Q3 call pushes those trends further: adjusted EBITDA grew 55%, free cash flow grew 73%, and GAAP EPS was $0.14. The old knock on Zeta was that it grew fast but not profitably. That knock is getting stale.The Headless U-Turn
The most revealing exchange came near the end, when Scott Berg asked about Zeta's new headless MCP server. In Q2, Steinberg had dismissed headless architectures: “I don't think it becomes a competitive force for us, and I don't think it is something that we would move into quickly from our front.” — David A. Steinberg, Chief Executive Officer (CEO) · 2026-08-04 On this call, he reversed himself.That is a genuine strategic change, not a messaging tweak. Athena will plug into ChatGPT, Claude, Gemini and other frontier models, while Zeta keeps its data cloud and activation layer proprietary. The UI becomes someone else's; the intelligence and execution stay Zeta's. Why does it matter? Because it turns the large language models from a threat into a distribution channel. Zeta does not need every enterprise to live inside its interface. It can meet customers inside the tools they already use. That is a bigger addressable market and a lower-friction land motion, but it also puts Zeta in a more crowded interface layer. The moat now has to be the data and the activation, not the dashboard. This is Zeta's enterprise version of the market's Personal AI agents wave.yes, this is a pivot for us. I have always been we need to control everything, and now I'm saying the UI can actually exist in a third-party platform.