ZETA — Zeta Global Holdings

Zeta Global has built a genuine AI-driven marketing platform — 2.4 billion deterministic identities, 120% net revenue retention, eighteen consecutive quarters of beating and raising guidance — and at roughly 19 times forward free cash flow after a 34% stock decline, the price does not reflect the operating reality. The cloud is real: a securities class action filed in December 2024 alleging fraudulent data collection practices remains pending, and no quarterly earnings report fully dissolves litigation risk that turns on a court determination rather than revenue retention rates. Interesting but requires a specific catalyst to be actionable.

---

The enterprise marketing technology industry in 2026 is living through a structural disruption created by the collapse of third-party cookies. Google's deprecation of cookie-based tracking in Chrome, combined with privacy legislation across U.S. states, the EU, and increasingly APAC, has dissolved the principal architecture on which the global marketing technology industry was built. For two decades, the ability to track individual consumers across the web rested on a browser-level identifier that consumers never explicitly consented to share and that regulators are now systematically eliminating. The deprecation of that identifier does not eliminate the demand to understand and target consumers — it eliminates the tool that most of the industry relied on to do so. Companies that built their data assets around first-party, consent-based, deterministic identity have a structural advantage in this environment. Companies that relied on cookie infrastructure are scrambling to retrofit identity resolution onto architectures that were not designed for it.

This upheaval is arriving simultaneously with an AI-driven restructuring of the marketing workflow. The marketing operations of a Fortune 500 company in 2026 look fundamentally different from 2020: systems that once required data science teams to configure now generate campaign recommendations, audience segmentation, creative variations, and attribution analysis with minimal human involvement. The winners in this environment are not necessarily the incumbents — it is the platforms whose underlying data assets and AI models are suited to the post-cookie, machine-to-machine marketing world. The legacy platforms assembled their capabilities through acquisition over decades: Salesforce buying ExactTarget and Pardot, Adobe buying Marketo, Oracle buying Eloqua and BlueKai and a half-dozen others. The result is what enterprise buyers now call "Franken-stacks" — portfolios of acquired technologies with separate data models, separate APIs, and integration overhead that consumes a meaningful portion of the marketing operations budget. Oracle has largely ceded the marketing application layer as it pivots toward cloud infrastructure, leaving a specific gap that challengers have been quick to exploit.

The global marketing technology market is estimated at approximately $670 billion in 2026, a figure that reflects the breadth of what "marketing technology" encompasses more than a serviceable addressable market for any single vendor. The relevant competitive arena for Zeta is the enterprise marketing cloud segment — the platform layer that large organizations use to acquire, retain, and re-engage customers across digital channels. That market is dominated by Salesforce, Adobe, and to a declining extent Oracle, each generating billions in marketing-related software revenue. All three assembled those capabilities through acquisition rather than building them natively, creating structural integration problems that become more expensive as the cookie crisis forces enterprise marketing teams to audit their identity data. A platform that cannot resolve a single consumer identity across email, display, paid search, and Connected TV cannot measure attribution accurately or target efficiently. The incumbents are attempting to add identity resolution to architectures not designed for it. The integration complexity of that retrofit is the opening that purpose-built challengers are competing to capture.

Zeta Global operates what it calls an AI Marketing Cloud — a platform built on three integrated components that were designed to work together from the beginning. The first is a proprietary identity graph of over 2.4 billion deterministic global identities: records linked to real individuals through first-party, consent-based data collection rather than probabilistic inference. The second is an AI engine named Athena that processes first-party and third-party signals to predict purchase intent, churn probability, and audience receptivity at the individual level. The third is an omnichannel activation layer that orchestrates delivery across email, display advertising, Connected TV, mobile, and the open web. The company was founded in 2007, went public on the NYSE in 2021, and has grown revenue at 20% or better in each of the seven years since the business reached meaningful scale.

Revenue is primarily subscription-based for platform access, supplemented by usage-based fees tied to media activation volume. This structure creates a natural expansion mechanism within existing accounts: a customer who adopts the platform for email marketing and observes improved deliverability and open rates discovers that adding Zeta's identity targeting to their paid media spend increases their return on ad spend. That additional use case increases their annual contract value without requiring Zeta to acquire a new customer. The Marigold acquisition, completed in November 2025 for approximately $325 million, added over 100 global enterprise brands including more than 40 Fortune 500 companies — primarily European businesses using the acquired platforms for loyalty and retention marketing. These customers have not yet received Zeta's customer acquisition or re-engagement capabilities, making the acquired customer base a three-to-five-year cross-sell opportunity from the acquisition date.

The competitive moat at Zeta rests on the identity graph. Building 2.4 billion deterministic identities requires years of partnerships with publishers, retailers, financial institutions, and data-rich platforms — each partnership adding signals, each signal improving the model's ability to resolve identity across device and context, better identity resolution producing better targeting outcomes for clients, better outcomes retaining and attracting clients who contribute more first-party signal. The compound effect of this flywheel has created a data asset that competitors cannot quickly replicate. It is not the data technology that is the moat — technology diffuses — it is the accumulated scale of consented, verified identity relationships that took a decade to build and would take another decade to replicate from zero.

The strongest evidence that this moat is real and delivering value is the net revenue retention rate. NRR at 120% means that the cohort of enterprise customers who spent $100 with Zeta last year spent $120 this year — 20% expansion without adding a single new customer. Enterprise marketing teams operate with detailed attribution systems and quarterly ROI reviews. They are not expanding their platform spend by 20% annually because they enjoy the relationship; they are doing so because the platform produces measurable improvements in marketing efficiency. A business built on fraudulent data profiles would generate campaigns that reach fabricated consumers, produce zero conversions, and surface immediately in attribution analysis. The opposite is happening: NRR expanded from 114% in 2024 to 120% in 2025, accelerating as the identity graph and AI model have grown larger. Eighteen consecutive quarters of beating guidance — including four quarters after Culper Research published its short-seller report in November 2024 — is a financial track record that is not compatible with a core business built on consent farms.

FY2025 revenue reached $1.305 billion, a 30% increase year-over-year following 38% growth in FY2024. Free cash flow grew 78% to $165 million, at a 13% margin — meaningful cash generation from a business that is still in a growth phase. Q4 2025 produced $6.5 million in GAAP net income, the first quarter of GAAP profitability in the company's public history, at a 1.7% margin on $395 million in revenue. Management guided FY2026 to positive GAAP net income for the full year. The divergence between GAAP and non-GAAP reporting is material: the FY2026 guidance projects adjusted EBITDA of approximately $391 million at a 22% margin, while GAAP net income is expected to be positive but thin. The reconciling items are primarily stock-based compensation — which is a real cost to shareholders — and acquisition-related intangible amortization. The non-GAAP presentation captures the operating leverage of the underlying platform business; the GAAP presentation is what shareholders actually own. The convergence of both measures toward positive territory in 2026 is the most significant financial development in Zeta's public life.

Management has been transparent about where the numbers diverge from the narrative. The guidance for FY2026 revenue of $1.749 to $1.762 billion (34-35% total growth) includes approximately $190 million from the Marigold enterprise software business and revenue from political advertising in a midterm election year. Organic platform growth, excluding Marigold and political candidate revenue, is guided at 20-21%. This distinction matters: investors evaluating the quality of Zeta's growth should anchor to the 20-21% organic rate rather than the 34-35% headline, which benefits from the non-recurring structure of a single large acquisition and the quadrennial cycle of political spending. The organic rate remains strong — but the headline flatters the underlying trajectory by approximately 14 percentage points.

David Steinberg co-founded Zeta in 2007 and has served as its CEO since founding. His personal ownership of approximately 11% of the company — combined with other insiders, total insider ownership reaches 18% — is the alignment mechanism that carries more weight than any compensation structure. Steinberg purchased $1.03 million in personal shares at market prices during the post-short-seller decline, a signal that carries weight when it comes from a founder who has built the business from zero and has no information advantage over the market regarding the fraud allegations — he either knows them to be false or he is committing his personal capital to a burning ship. The former is the more rational interpretation. The share repurchase program has been actively deployed: Zeta repurchased 7.9 million shares for $120 million in 2025 at an average of approximately $15.19, and an additional 1.5 million shares for $25 million in the first six weeks of 2026, with $139 million remaining under the current authorization. Steinberg characterized buybacks as the "single best use" of capital at current prices — a statement economically rational only if he believes the stock materially understates the business's value. The Marigold acquisition, at approximately 1.7 times the acquired business's annualized revenue, was disciplined relative to historical enterprise software transaction multiples and was funded primarily from cash, preserving a net cash position of approximately $177 million entering 2026.

The growth trajectory across four years makes the runway argument concrete:

Year	Revenue ($M)	YoY Growth	Scaled Customers	Super-Scaled	Net Revenue Retention	FCF ($M)
2022	$591	29%	403	~103	—	—
2023	$729	23%	~451	~131	—	—
2024	$1,006	38%	~528	~148	114%	$93
2025	$1,305	30%	602	184	120%	$165
2026E	$1,755	35%	—	—	—	$231

The table shows an operation running as a platform flywheel predicts: scaled customers growing consistently, super-scaled customers growing faster as the customer base matures toward multi-channel adoption, net revenue retention expanding as the platform delivers compounding value, and free cash flow growing materially faster than revenue as operating leverage accumulates. The 38% revenue growth in 2024 benefited from a strong presidential election advertising cycle; 30% growth in 2025 maintained momentum against that comparison. The acceleration in FCF — from $93 million in 2024 to $165 million in 2025, a 78% increase on 30% revenue growth — is the operating leverage story made visible. Nine of Zeta's top ten verticals grew more than 20% year-over-year in 2025, suggesting the growth is broad-based rather than concentrated in one or two end markets that could reverse.

The penetration argument begins with the current customer base. Zeta serves 602 scaled customers — enterprise accounts spending more than $100,000 annually on the platform — against a primary addressable universe of Fortune 1000 and Global 500 companies that, non-overlapping, comprises approximately 1,200 to 1,500 accounts. At 602 scaled customers, the company has captured roughly 40 to 50% of its most directly addressable enterprise population. This implies that meaningful growth in the scaled customer count will be more difficult to sustain than in prior years, and that the expansion opportunity increasingly lives within the existing customer base rather than in new logo acquisition. The NRR of 120% transforms this limitation into a structural advantage: existing customers expanding by 20% annually compound the revenue base faster than equivalent investment in new customer acquisition. Marigold's 100+ enterprise accounts, which have not yet received the full Zeta platform offering, represent three to five years of cross-sell opportunity — applying Zeta's customer acquisition and re-engagement capabilities to a customer base that is already bought into the platform for loyalty and retention marketing.

The international runway is earlier in development. Marigold's EMEA presence — established relationships with European enterprise brands in financial services, retail, and media — gives Zeta an international footprint that it previously lacked. European markets are meaningfully earlier in their data-driven marketing adoption curves than the U.S., and APAC is earlier still. The penetration of Zeta's platform into those geographies is a decade-long opportunity rather than a near-term growth driver, but the Marigold acquisition has opened the door to a geographic expansion that was previously inaccessible.

On November 13, 2024, Culper Research published a short-seller report asserting that Zeta had assembled its identity graph through a network of consent farms — sham websites designed to gather consumer data under false pretenses — and that these practices allegedly represented 56% of adjusted EBITDA. The report additionally alleged round-trip revenue transactions with Kubient, an ad-tech company, designed to inflate reported revenue without genuine economic substance. The stock fell 37% in a single trading session. In December 2024, a securities class action was filed; Labaton Keller Sucharow was appointed co-lead counsel in February 2025. As of April 2026, the litigation is pending.

Zeta denied both allegations completely. Management stated that the company does not operate consent farms, that all data collection is consent-based and subject to regular audit by enterprise clients and external partners, and that the relationship with Kubient was legitimate and represents an immaterial portion of revenue. Steinberg publicly characterized Zeta as an ethical company that could not serve 40% of the Fortune 100 while engaging in the practices described. The operational evidence aligns with the denial. A business generating revenue from fraudulent data profiles would produce marketing campaigns that reach non-existent or misrepresented consumers, producing zero measurable conversion outcomes that enterprise attribution teams would immediately identify. The NRR expansion from 114% to 120% in 2025 — meaning existing clients are collectively increasing their spend by 20% annually — is the precise opposite outcome of a platform whose data quality is compromised. Eighteen consecutive quarters of beating guidance, including four after the Culper report, and Steinberg's personal million-dollar share purchase at depressed prices, are financial commitments that are difficult to reconcile with knowledge of fraud.

The most credible residual risk is not that the Culper allegations will be proven — it is that Zeta's data collection practices, even if not fraudulent under current legal standards, exist in a regulatory environment that is actively tightening its definition of appropriate consent. The FTC and state attorneys general have been expanding their interpretation of consent requirements in digital data collection. A company operating at the frontier of permissible consent practices could face regulatory scrutiny that does not require the existence of fraud, only a determination that the consent mechanisms were insufficient by evolving standards. This is an unquantifiable risk that requires legal resolution rather than operational evidence to dismiss. Until the class action has been dismissed or settled, and no regulatory inquiry has materialized, this risk remains imperfectly priced regardless of how strong the quarterly results are.

At approximately $15.75 per share, Zeta's market capitalization is approximately $4.5 billion. Against net cash of approximately $177 million, the enterprise value is approximately $4.3 billion. Against FY2026 guided free cash flow of $231 million, the forward EV/FCF multiple is approximately 19 times. Against guided adjusted EBITDA of $391 million, the forward EV/EBITDA is approximately 11 times. Against guided revenue of $1.755 billion, the forward revenue multiple is approximately 2.5 times.

These multiples are not expensive for a software business growing at 35% with 120% NRR and a path to positive GAAP net income. A comparable enterprise data platform growing at 20% organically with 110-115% NRR would typically trade at 25 to 30 times forward free cash flow in the current market. At 19 times, Zeta is priced as if either the growth rate will decelerate materially faster than the NRR evidence suggests, or the litigation represents a certain significant financial loss rather than an uncertain probability. Neither assumption is obviously correct. The most legitimate bear objection on valuation is that the $231 million in guided free cash flow rests on non-GAAP adjusted EBITDA that excludes $150 to $180 million in annual stock-based compensation — a real cost to existing shareholders. If SBC is treated as the economic dilution it represents, the effective multiple on cash generation attributable to equity holders is materially higher than 19 times. The partial answer is that the buyback program — $145 million deployed in 2025 repurchasing diluted shares — offsets a portion of that dilution, and that the convergence toward GAAP profitability in 2026 suggests the gap between non-GAAP and economic reality is narrowing.

For the thesis to become outright compelling — not merely interesting — one of two conditions needs to materialize. The first is legal resolution: dismissal of the class action on a finding that the Culper Research allegations fail to meet the materiality or scienter standards for securities fraud, or a settlement on terms that represent an immaterial fraction of the business's annual free cash flow. The second is continued operating vindication: if Zeta produces strong Q1 and Q2 2026 results with organic growth sustaining 20%+ and Marigold integration proceeding cleanly, the unresolved litigation will feel progressively less relevant to an institutional owner who has watched the platform compound through eighteen quarterly reports. The more likely path is both conditions arriving simultaneously — quarterly results that make the litigation feel like a nuisance rather than a threat, followed eventually by legal resolution that removes the formal overhang.

The flywheel has been turning for four years, the data moat is real, and the CEO is buying stock with his own money. The only thing keeping the price at 19 times a 35% grower is the unresolved question of whether the data that powers the platform was collected ethically. The operating evidence says it was. The court system will eventually say the same or it will not — and that determination is the catalyst the stock is waiting for.