Meta has acquired Sundial, the AI-powered data analytics startup founded by former Meta leaders Julie Zhuo and Chandra Narayanan, according to Axios, which said Meta confirmed the deal. Sundial’s founders and other staff are joining Meta’s newly formed Data Analytics Tooling & Agents organization, or DATA, led by chief data officer Alex Schultz. Narayanan is expected to lead analytics tooling and agents work. Sundial will stop serving external customers once the transaction closes, though neither company has disclosed a price or closing timeline.
Why this matters: Meta is not buying another public-facing AI app. It is buying a way to let more employees ask questions of data, generate analysis, and potentially move faster from idea to experiment without routing every request through a central analytics team. That makes the deal a useful marker for where the AI market is getting more strategic: not just at the model layer, but at the operating layer that sits between people, metrics, warehouses, permissions, and decisions.
The real question for buyers and founders is whether this is mainly an acquihire for scarce analytics talent or a sign that natural-language, self-service data systems are becoming core infrastructure inside big companies. Based on what is public, the stronger reading is both — but the infrastructure story is the more important one.
More than a talent grab, but not proof of a category winner
Sundial was founded in 2021 and raised about $23 million in reported funding, including a $16 million Series A in 2025. Its pitch was straightforward and timely: let business users ask natural-language questions through a chatbot-style interface, then turn those questions into governed analysis using a stack that combines data cleaning, semantic layers, SQL generation, dashboards, and notebooks.
On its face, that makes Sundial an attractive acquihire target. Its founders know Meta’s culture and data complexity. Narayanan previously led analytics work across Facebook, Instagram, and Sequoia Capital. Zhuo was Meta’s vice president of product design. In a market where AI product talent is scarce and experienced analytics operators are scarcer, those credentials matter.
But the placement of the team matters just as much. Schultz became Meta’s first chief data officer in July 2026, with responsibility for analytics, experimentation, monitoring, and tools for Meta’s AI systems. Folding Sundial into that new DATA organization points to a deliberate internal capability build, not a side bet. If Meta simply wanted a handful of engineers, it could disperse them across existing AI teams. Instead, it is adding them to a centralized function responsible for the machinery around measurement and decision-making.
That does not make the deal proof that AI self-service analytics works safely at Meta scale. The public record does not show how widely Sundial’s product will be deployed internally, which pieces of its stack Meta values most, or whether the acquisition will improve decision speed, model quality, ad performance, or reliability. For now, the deal is best read as a strategic move on the control plane around data and agents, not as a demonstrated operating breakthrough.
Why the analytics layer is becoming strategic
Natural-language analytics promises a simple benefit: shorter queues. Product managers, marketers, operations teams, and engineers can ask questions directly rather than waiting for a specialist to write SQL, reconcile a metric definition, and build a chart. In a company that runs constant experiments, compressing that loop can be valuable.
That is especially relevant for Meta because the company now has a formal organization around analytics and tooling for AI systems. In that setting, the question is not merely who can query a warehouse. It is who can do so using the right definitions, the right permissions, and the right business logic, with enough guardrails that a fast answer does not become a misleading one.
This is where the Sundial-type product matters. A natural-language analytics layer is not just a nicer interface on top of SQL. It is a translation system that decides how plain-English questions map to semantic models, tables, joins, quality checks, notebooks, dashboards, and sometimes actions. When it works well, more employees can interrogate data and move from question to experiment faster. When it works poorly, the company gets metric drift, silent changes in logic, permission inheritance problems, and confident but flawed queries.
That tradeoff is why the deal lands beyond one startup’s exit. Enterprise AI vendors and their largest customers are increasingly fighting to own the deployment layer, not just rent access to foundation models. The company that controls the semantic layer, lineage, approvals, audit logs, and agent behavior can shape how decisions get made. Meta’s move suggests it sees that layer as too important to leave outside the company.
What changes for customers, buyers, and founders
For Sundial’s outside customers, the immediate issue is operational. Service will continue until the transaction closes, but the companies have not said when that will be, how customer migrations will work, or how warehouse credentials and related access will be handled during the wind-down. That uncertainty matters because analytics products sit deep inside a company’s data plumbing. Replacing them is rarely as simple as swapping a chat interface.
For enterprise buyers evaluating other AI analytics vendors, the lesson is not that every startup will vanish into a platform company. It is that durability and portability now deserve the same scrutiny as demo quality. Buyers should want export paths, semantic-layer portability, audit logs, support commitments, and a clear shutdown plan. A strong natural-language interface is helpful; an exit path is essential.
For data leaders, the deal sharpens a governance question that many teams are only beginning to confront: who owns definitions and approvals when agents can write SQL, assemble dashboards, or trigger downstream actions? In many organizations, the bottleneck is not raw query writing. It is deciding which metric is authoritative, who can see what, and how changes are reviewed. AI analytics can reduce dependence on a centralized reporting queue, but it can also make an organization faster at producing inconsistent answers if those controls are weak.
For startup founders and investors, Sundial offers a mixed signal. The positive reading is that a focused AI-native analytics company can become strategically important to one of the largest platform companies in the world. The harder reading is that the same outcome may end the standalone product for customers. That is the core tension in the economics of AI acquihires: a startup can prove the value of its team and product approach precisely by becoming too strategically useful to remain neutral.
The best interpretation of Meta’s move, then, is not that it has already solved enterprise analytics or that every business should rush to replace dashboards with chat. It is that large companies increasingly view the data-and-agent layer as operating infrastructure. Meta appears to want that capability in-house, under its own governance and at its own scale. What remains unproven is whether internalizing that layer will actually deliver better decisions, faster experimentation, or safer AI operations once the hard work of deployment begins.
By
By
By
By
By
By
