Should Snowflake acquire Coalesce.io or dbt Labs?

Direct Answer

Neither acquisition is a layup, but if Snowflake has to pick one, Coalesce.io at an estimated $1.5-2B is dramatically cleaner than dbt Labs at an estimated $4-6B. The dbt Labs deal looks attractive on a whiteboard — buy the de facto transformation standard, lock in the SQL-modeling layer, capture the Cloud ARR — but in practice it walks Snowflake into an open-source community that will react badly, a Databricks competitor who will hard-fork dbt Core inside 90 days, and a multi-warehouse positioning premise that *requires* dbt to keep cross-engine credibility to be worth the price paid. Coalesce, by contrast, is already Snowflake-native, GUI-first, governance-friendly, and structured exactly the way an enterprise CFO wants the transformation layer to look in the Cortex-agent era. The strategic alternative — and the one most M&A operators with platform-CEO scars will quietly recommend — is to acquire Coalesce, deepen the dbt partnership at the commercial layer, and build native transformation primitives into the Cortex Cookbook rather than try to absorb either of dbt's two warring constituencies. The trap with dbt isn't the price tag, it's the post-close swamp.

Why Snowflake Should Care About The Transformation Layer

• The transformation layer is where Snowflake credit consumption gets locked in or leaks out — every dbt run, every Coalesce job, every Fivetran-then-transform pipeline drives compute that either stays on Snowflake warehouses or migrates to Databricks SQL warehouses depending on which tool sits in the middle.
• Existing partner relationships are deep on both sides — Snowflake co-sells with dbt Labs, runs joint Summit content with Coalesce, and both vendors carry premier-tier Snowflake partner status; the partner-vs-acquire calculus is genuinely close.
• Databricks' acquisition pattern is the forcing function — the Tabular acquisition for an estimated ~$1-2B (Iceberg control) and the Lakebase / MosaicML moves show Databricks will pay up to own the layers Snowflake treats as partner surface area.
• The partner-vs-acquire debate inside Snowflake is real — Mike Scarpelli has historically preferred sub-$1B tuck-ins (Streamlit at ~$800M was the high-water mark publicly disclosed), and any deal above that bar requires a different internal narrative than the standard tech-and-team playbook.
• The agent era changes the math — when Cortex Agents start writing transformation logic autonomously, the GUI-first / metadata-first vendor wins versus the code-first / Git-PR-first vendor, which structurally favors Coalesce's design over dbt's.

The Coalesce.io Case

• Funding context — Coalesce raised a Series B in 2022 (publicly reported ~$26M, led by Emergence) and a subsequent ~$50M round in 2024 (publicly reported coverage), implying a current private valuation in the several-hundred-million-dollar range; an acquisition price in the $1.5-2B band based on public funding rounds and analyst commentary would be a reasonable strategic premium.
• GUI-first positioning — Coalesce's column-aware visual transformation IDE is structurally different from dbt's YAML + SQL + Jinja paradigm; it's the layer non-engineering analysts can actually use without a dbt-Cloud onboarding cycle.
• Snowflake-native architecture — Coalesce was built Snowflake-first and only later added other-warehouse support; the integration surface, the metadata model, and the column-lineage features all assume Snowflake primitives.
• What Snowflake gets — a governance-friendly transformation layer that maps cleanly into Horizon Catalog, a credible answer to "how does my analyst build pipelines without Python," and a Cortex-agent-friendly metadata graph the platform can reason over.
• What it costs — the deal is small enough to absorb in 2-3 quarters, low enough community-risk to avoid press cycles, and structurally additive rather than disruptive to the existing partner ecosystem.

The dbt Labs Case

• Valuation context — dbt Labs was valued at $4.2B in its 2022 Series D (publicly reported), then took a 2023 down-round per TechCrunch coverage that reset expectations; an acquisition today would likely clear in the $4-6B range based on public funding rounds and analyst commentary, which puts it in mega-deal territory by Snowflake standards.
• dbt Core open-source community + governance risk — dbt Core is Apache 2.0 with tens of thousands of practitioners and a passionate community; any acquisition triggers a governance debate that has historically gone badly for the acquirer (see the Microsoft + GitHub friction window, the Elastic license drama, the HashiCorp + IBM community reaction).
• dbt Cloud vs. dbt Core split friction — dbt Labs has spent two years navigating the open-core tension and the recent Fusion engine + license updates; an acquisition forces that split to resolve faster than the community will tolerate.
• Databricks-friendly positioning — dbt's multi-warehouse credibility (Snowflake + Databricks + BigQuery + Redshift + Fabric) is core to its enterprise value proposition; the moment Snowflake owns dbt, every Databricks-shop dbt customer starts evaluating SQLMesh, Coalesce, or a hard-fork.
• What acquisition would unlock + what it would break — Snowflake would own the SQL transformation standard (huge), but break the multi-platform neutrality that *makes* dbt the standard (also huge), and inherit the Tristan Handy founder-key-person risk that anyone who has run a developer-tools acquisition knows is the deal's load-bearing wall.

Why dbt Is The Trap

• Open-source community blowup risk — the dbt Slack, dbt Discourse, and the dbt-core GitHub repo are public theaters where a Snowflake acquisition would be litigated in real time; the comms cost alone is a 2-3 quarter distraction.
• Databricks would respond by hard-forking dbt Core inside 90 days — they have the engineering bandwidth, the OSS-community goodwill from MLflow / Delta / Unity Catalog, and the strategic incentive; the fork would carry whatever name they choose and would ship with first-class Databricks SQL warehouse support.
• Microsoft + GitHub is the named precedent that worked but only because Microsoft committed publicly to open governance, kept the brand and team independent for years, and absorbed billions in dilution to do so; Snowflake's operating culture is leaner and the precedent is harder to replicate.
• Multi-platform positioning loss — the moment dbt becomes a Snowflake-owned product, every CIO running a multi-warehouse strategy (and that's most F500 data orgs) re-evaluates; the asset Snowflake bought decays the day the deal closes.
• Tristan Handy founder-key-person risk — Handy is the cultural and product center of gravity at dbt Labs; any acquisition has to retain him for 3-5 years post-close or the community follows him out the door, which puts the entire deal value at the mercy of one person's earn-out clock.
• Price-to-strategic-value ratio is upside-down — paying $4-6B for an asset whose strategic value depends on staying neutral toward your largest competitor is the textbook definition of a deal that should not happen.

Why Coalesce Is The Better Bet

• Already Snowflake-native — no architectural refactor, no metadata-model rewrite, no warehouse-abstraction-layer to maintain; the integration lift is measured in months, not years.
• Low community-risk profile — Coalesce is commercial software with a partner ecosystem, not an OSS movement with a Slack uprising waiting to happen; the deal can close quietly.
• GUI-first matches Cortex agent direction — when Cortex Agents start auto-generating transformation pipelines, a metadata-rich visual graph is a better substrate for an LLM to reason over than a directory of .sql + .yml files.
• Smaller integration lift — Coalesce's headcount, ARR base, and customer concentration are all sized for a sub-2-quarter integration playbook of the Streamlit / Neeva / Datavolo class.
• Faster ARR contribution — a Coalesce deal contributes meaningful ARR inside FY27 because the cross-sell motion (every Snowflake $1M+ account is a Coalesce qualifying lead) is mechanical rather than strategic.

What Snowflake Should Actually Do

• Acquire Coalesce.io at an estimated $1.5-2B based on public funding rounds and analyst commentary — clear deal, clean integration, additive to the partner ecosystem rather than disruptive.
• Deepen the dbt Labs partnership at the commercial layer without acquiring — co-sell harder, fund joint Cortex + dbt reference architectures, write the joint Snowflake Summit + Coalesce-the-conference keynotes, lock in the partnership economics.
• Build native transformation primitives into the Cortex Cookbook — Cortex Analyst already generates SQL; extend that into a first-class Snowflake-native transformation surface so the platform has organic optionality regardless of which third-party tool wins.
• Named-launch-customer co-sell — pick 10-20 marquee F500 logos, run joint Snowflake + Coalesce co-sell campaigns, prove the integration thesis with logos before the deal closes so the post-close narrative writes itself.
• Protect Marketplace partner economics — keep dbt, SQLMesh, Datafold, and the rest of the transformation ecosystem economically aligned with Snowflake even after a Coalesce close; the Marketplace lever from q1605 only works if partners trust the platform.
• Keep $2-3B of M&A budget powder dry for the agent-runtime layer — the more strategically important acquisition over the next 24 months is whoever owns the enterprise agent orchestration tier, not the transformation tier.

Target Comparison Table

M&A Decision Flow

Bottom Line

Buy Coalesce, partner harder with dbt, build native in Cortex, and keep the powder for the agent-runtime acquisition that actually matters in 2027. The dbt Labs deal is the kind of transaction that looks bold in the announcement press release and disastrous 18 months later when Databricks ships dbt-Fork-Spark-Edition and the OSS community migrates en masse. The Coalesce deal is the kind of transaction that looks small in the press release and quietly compounds into the transformation-layer lock-in Snowflake actually needs. The discipline that defined Snowflake's M&A history — Streamlit at ~$800M, Neeva, Truera, Datavolo as sub-$1B tuck-ins — is the same discipline that should govern this decision. *(see also: q1571, q1584, q1585, q1605)*