Snowflake Bets the Company on Cortex Analyst and Cowork — and Commits to GAAP Profitability by Q4 FY2028
Snowflake Analyst Day, June 2, 2026 — Summit Conference, San Francisco
Snowflake's annual Investor Day delivered two genuinely new signals that investors needed to hear: a concrete GAAP profitability commitment for Q4 fiscal year 2028, and a clearer-than-ever articulation of how Cortex Cowork — the company's personal work agent — is being positioned not merely as a data query tool but as an enterprise control plane that competes directly with the likes of Microsoft Copilot and Salesforce Agentforce for the center of how knowledge workers operate. Neither of these was fully on the table before Tuesday.
GAAP Profitability Target Is the Headline Financial Commitment
CFO Brian Robins announced that Snowflake will reach GAAP profitability in Q4 FY2028, driven by three levers — revenue growth, operating expense discipline, and stock-based compensation reduction — with the emphasis explicitly on the latter two. Robins was careful to frame this as not a revenue assumption change: "This is not a discussion about FY28 revenue." The SBC trajectory is the key modeling input he handed analysts, noting that SBC has already dropped from 41% of revenue to 34% to a guided 27% this fiscal year, implying a continued glide path. Non-GAAP operating margin, which stood at 6.4% in FY25, was guided to 13.5% in the most recent quarter — more than doubling in two years while keeping non-GAAP product gross margin roughly flat at 75%. The company also confirmed it does not expect to execute any large M&A, and has approximately $800 million remaining under its $4.5 billion buyback authorization.
The GAAP profitability commitment matters symbolically as much as financially. CEO Sridhar Ramaswamy framed it as the culmination of three years of internal reinvention: "I see this as the culmination of the work that we have done over the past three years to reinvent ourselves to be more driven, more product-focused, more quality-obsessed." The message to investors is that the company believes it can grow and expand margins simultaneously — not by headcount suppression alone, but by fundamentally restructuring how work gets done using its own AI tools.
Cocoa and Cowork: Two Sides of the Same Strategic Coin
The most substantive product signal from Investor Day was the clearest delineation yet between Coco — Snowflake's coding and data agent — and Cowork, its personal work agent. Ramaswamy described them as "two sides of the same coin" sharing enormous infrastructure, but tailored for different personas. Coco is the technical practitioner's tool, already in broad deployment across Snowflake's 14,000-plus customer base, with measurable proof points: a Global 2000 hospitality firm completed a migration 60-plus percent faster using Coco, and a financial services firm saved over 500 hours on a recurring job. The Coco desktop went generally available at Summit after going into public preview just weeks prior.
Cowork is the bigger and more speculative bet. Originally launched as Snowflake Intelligence in November 2024 and conceived as an analytics aggregation layer, the product has since evolved into something considerably more ambitious — a personal work agent that integrates structured and unstructured data, connects to enterprise applications including Salesforce, Gmail, and Workday, and supports automation and scheduling. Robins described using it himself every morning: "I go into Coco and I type 'good morning,' and it takes all this data from structured to unstructured sources and puts it into an easy-to-read format within minutes." Through the Natoma acquisition — a newly disclosed tuck-in that adds MCP connectivity to 100-plus business systems — Cowork can now authenticate once and reach across an enterprise's application stack with governance policies controlling what agents can and cannot do externally.
Natoma Acquisition Adds the Enterprise Governance Layer Cowork Needed
The Natoma acquisition, announced at Summit, addresses one of the more pointed criticisms of agentic AI deployments: the lack of guardrails when agents push data outward. Christian Kleinerman, EVP of Product, walked through Snowflake's own internal configuration as an example: "The way we configured the email connector of Natoma was — you can ask your agent, Coco, to send an email. If the email is going to an internal recipient, it sends it. If the email is going to an external recipient, it puts it in your drafts." The product gives administrators the ability to set their own policies, audit all agent activity, and enable single-sign-on across all connected systems. For enterprises paralyzed by the governance implications of deploying agents at scale, this is a meaningful addition to the platform.
Snowflake-Managed Iceberg Storage Removes the Last Objection to Open Formats
One of the more technically important announcements, and one likely to be underappreciated by generalist investors, was the general availability of Snowflake-managed storage for Iceberg tables. The longstanding trade-off when customers moved to Iceberg was that they had to manage their own object storage, surrendering some of Snowflake's operational convenience. That friction is now eliminated. Kleinerman was direct: "You can still be interoperable, but we'll do the management of the storage, we'll give you the economics." He characterized this as "an even better tailwind relative to at least how we thought and modeled the adoption of Iceberg." The interoperability story is further reinforced by Snowflake's integration of REST Catalog APIs into its Horizon catalog, enabling read/write access to data sitting in Databricks, AWS Glue, and other engines. Kleinerman's assertion that Snowflake's Iceberg implementation is "second to nobody" was made, he noted, "based on facts" — the company is currently the broadest implementer of the V3 spec and is steering the V4 specification.
Migration Acceleration Is Becoming a Real Revenue Driver
Robins quantified migration momentum more explicitly than in prior quarters: migrations grew 1.9x from FY25 to FY26, and use cases grew 1.7x. Time-to-first-consumption has been cut from 10 months to 7 months, with further reduction targeted. The Datometry acquisition — now productized as part of Snowflake's migration suite — adds Teradata virtualization, allowing enterprises to present a Teradata interface to legacy applications while running on Snowflake underneath, potentially collapsing multi-year Teradata migrations into far shorter timelines. The Spark migration story is also improving organically: Kleinerman noted a customer who abandoned a legacy Spark API migration halfway through and simply asked Coco to convert the workload to Snowpark — completing the migration and achieving a 5x performance improvement in the process.
Sales Metrics Are Moving in the Right Direction
Ramaswamy disclosed two internal productivity metrics that suggest the go-to-market transformation is working. Use cases won per account executive increased 86% year-over-year in the most recent quarter. Use case go-lives per solutions engineer increased 58% year-over-year. Both metrics are measured on a per-person basis, not in aggregate, eliminating team size as a confounding factor. Sales cycle duration hit its lowest point in four quarters, which Ramaswamy acknowledged runs counter to intuition in a more competitive environment. New logo volume and new logo ACV both increased meaningfully year-over-year. The company's new CRO, referred to internally as JB, has been with Snowflake for over a decade and is leading a shift toward outcome-based pricing, with account executives now using Coco and Cowork in customer-facing settings with synthetic data demonstrations.
The OpenAI Data Analytics Announcement Is Not the Threat It Appears
When Deutsche Bank's Karl Keirstead raised OpenAI's concurrent data analytics product announcement, Ramaswamy looked it up in real time and concluded: "These look like lightweight skills that run on top of Snowflake. I think it's more about how you use these from core — and this is something they've actually talked to us about." He added that the announcement "is actually calling into our MCP connector — we gave them a quote on Friday." The broader question of whether frontier model companies will vertically integrate into the data layer drew a more measured response: Ramaswamy drew an explicit parallel to cloud providers, noting that hyperscalers who built competing data platforms ultimately became more complementary than competitive at most customer sites. He also flagged a growing dynamic that benefits Snowflake: "We're starting to hear customers tell us, 'Oh, I made a big commitment to this AI model company, but now I want to use the other one.' That dynamic we saw with the cloud providers, and it's starting to benefit us."
Cowork Is Still Early — Management Says So Explicitly
The most important caveat from Investor Day came from Ramaswamy himself, without prompting. On Cowork specifically: "I'm the first person to say that is early and we have to prove the scale use cases to you." The progression he outlined — create world-class products, get marquee customers to adopt, prove scaled adoption, then drive revenue — places Cowork firmly at stage two or three. When Brent Thill of UBS pressed on whether low-30s growth is consistent with the opportunity being described, particularly given that Snowflake's primary competitor is growing at roughly twice the pace, Ramaswamy offered no rebuttal: "Absolutely, we aspire for more, but showing is better than doing." Robins noted that guidance is rooted in observed behavior and that conservative upward revisions will follow as trends materialize. The honest read is that Cowork's contribution to revenue is not yet in numbers — it is in customer conversations, pipeline signals, and internal proof points, of which there are many but none yet at the scale that would move the growth rate in a visible way.
Workforce Reinvention Is the Unstated Operating Leverage Story
The most underappreciated theme from the day may be what Snowflake is doing to its own cost structure through AI-driven workforce transformation. The company disbanded its technical writing team, having concluded that coding agents produce superior documentation. Its site reliability engineering team has restructured around Coco-based operational tooling, reducing so-called "keep the lights on" overhead materially. Headcount growth has been near zero — 17 net adds in the most recent quarter excluding the Datometry acquisition, 37 the quarter before. When Raymond James's Adam Tindle asked why GAAP profitability is a priority given the growth opportunity, Ramaswamy's answer was revealing: "It isn't clear that simply throwing more humans at problems gets more things done. Scale no longer needs to be driven by the number of humans you have working on a problem." The implication is that Snowflake believes AI-augmented productivity is now structurally sufficient to grow the business without proportional headcount expansion — a thesis that, if correct, makes the margin trajectory durable rather than cyclical.