The debate about whether enterprises would actually pay for AI tooling has been running for two years. Every quarter, some analyst says we're still in the proof-of-concept phase. Every quarter, some other analyst says the spending is real and compounding. Snowflake's Q2 FY2027 earnings, reported September 2, settled it in the most concrete terms available: $1.49 billion in product revenue, up 37% year-over-year, with AI products driving roughly half that acceleration. The company raised its full-year guidance to $6.07 billion.

That's not a proof-of-concept. That's compounding.

What CoCo actually is

Cortex Code — rebranded CoCo at Snowflake Summit in June 2026 — is Snowflake's AI coding agent for data teams. It is not a general-purpose IDE assistant. It lives inside your Snowflake environment and targets specifically the work that data engineers and analysts do there: SQL, Snowpark Python, pipeline scaffolding, multi-step agentic tasks across the platform. Under the hood it pulls from multiple frontier models — Anthropic, OpenAI, and others — but you interact through a single interface baked into Snowflake.

In Q2, CoCo passed 9,100 accounts, adding more than 2,000 new accounts in the single quarter. CEO Sridhar Ramaswamy on the earnings call: "AI continues to compound our advantages, creating a flywheel effect across the business." Management attributed roughly half the quarter's growth acceleration to AI products — CoCo and CoWork, their enterprise collaboration layer.

SNOW climbed 23% in after-hours trading on the print.

Source spread

• Bloomberg — Snowflake Jumps on Revenue Outlook, Uptake of AI Assistant [builder] — straight coverage of the Q2 beat and stock reaction; emphasizes the AI flywheel narrative management is pushing.
• Seeking Alpha — FY27 guidance at $6.07B, 14.5% non-GAAP operating margin [builder] — the guidance raise and operating margin picture in detail.
• The New Stack — Snowflake CoCo coding agent [skeptic] — good technical critique of the portability-vs-governance tradeoff and the stateless CLI limitation.
• NAI 500 — SNOW, CoCo, and the stock market's AI fever dream [skeptic] — questions whether the growth is genuinely AI-driven or warehouse adoption masquerading as AI uplift.

Pros & cons

What's real:

• A $1.49B quarter at 37% growth is hard to argue with. This is not speculative revenue. It is paying customers, paying more, quarter over quarter.
• The multi-model architecture (Anthropic + OpenAI + others) means CoCo is model-agnostic at the capability layer. Snowflake is not betting on one horse. That matters for longevity.
• AI attribution at "roughly half" of growth is unusually high and unusually specific for management to say publicly. Companies get vague on AI attribution when they're not confident in the number.
• CoWork + CoCo as a combined platform is a coherent enterprise bet. You get AI coding assistance and AI collaboration in the same workspace, both pulling from your data.

What deserves a side-eye:

• CoCo is warehouse-native, which means if your data stack is anything other than Snowflake-first, you get reduced value or none at all. The New Stack's technical review is direct about this: portability and governance trade off against each other.
• The "flywheel" framing is not yet validated over multiple cycles. One quarter of AI-attributed acceleration is evidence. It is not proof of self-reinforcing compounding. The next two quarters are the actual test.
• Stock up 23% on the print suggests the market had priced in worse. The 37% growth rate is excellent; whether it continues depends on whether CoCo's adoption curve is early-adopter or mainstream.

What builders need to know

• If your stack is Snowflake-first, evaluate CoCo now. The product is past proof-of-concept — 9,100 enterprise accounts is sufficient signal that it works in production. The governance-native advantage is real if you're already inside the Snowflake ecosystem.
• If your stack is multi-cloud or multi-warehouse, CoCo is a partial solution. The portability limitation is not incidental — CoCo's value is tied to Snowflake. You'd need to evaluate whether Snowflake-specific coverage is worth the cost.
• The multi-model backend is good for durability. Snowflake is not locked to one AI provider. When model costs drop or a better model ships, they can swap. That's a reasonable architecture for a platform betting on AI longevity.
• Watch the Q3 earnings print in December. The 37% growth rate is real. Whether it's a trend or a launch bump resolves in one quarter.
• Don't conflate CoCo (coding agent) with CoWork (collaboration). They're bundled in management narrative but distinct products. If you're evaluating one, don't automatically sign up for both.

Further reading

• Bloomberg — Snowflake Jumps on Revenue Outlook, Uptake of AI Assistant — Q2 earnings and stock reaction
• Seeking Alpha — FY27 guidance at $6.07B — the financial details
• Snowflake — CoCo product page — what the product actually does
• The New Stack — Snowflake CoCo coding agent — independent technical analysis with the portability critique
• NAI 500 — SNOW, CoCo, and the stock market's AI fever dream — skeptical read worth having