Gianluca Carrera
← The register
2026-06-30N/ANot relevant✓ Reviewed

Snowflake Marketplace partners earn $100M+ in gross bookings in first half of 2026, up 277% YoY

Snowflake Marketplace partners including Dun & Bradstreet, Sema4.ai and Dataiku booked over $100M gross in H1 2026, up 277% year on year across 1,700+ transactions. Snowflake runs the venue but owns none of the underlying data assets sold across it.

Where the rake sits

Off the configuration axis, which is the point of the case. Snowflake does none of the enrichment and owns none of the substrate, so it has no rake on these deals to own or to lend — the partners hold the buyer, the contracted revenue and the rake, exactly as the framework predicts for work they enriched themselves. Snowflake is instead paid by the transaction without being party to it, in consumption of the compute the products drive, which is a claim on the venue rather than on any deal crossing it. Two consequences worth holding separately: the price tracks compute rather than the value of the data transaction, so the take is decoupled from what the buyer actually bought; and the fee that is not being charged today is an asset only if the venue is becoming the default, which the scale does not yet show and the growth rate might.

Who is involved

Snowflakelisted

US cloud data platform (the AI Data Cloud) that hosts customer data and runs a Marketplace where third-party data providers, app builders and AI startups sell data products, Native Apps and agents to Snowflake customers; Snowflake itself owns no substrate in the Marketplace transactions.

NYSE: SNOW; FY2026 revenue US$4.68B; approximately 9,060 employees (2026).

The reading

Where the work is

The enrichment work sits with the Marketplace partners who ship the data products, Native Apps and agents; Snowflake operates the venue and the compute/distribution rails rather than owning or enriching the substrates listed. The headline figure - >$100M in partner gross bookings over 1,700+ transactions in H1 2026, +277% YoY - is partner revenue, not Snowflake's take on a data asset of its own.

Enrichability

Not read from Snowflake's seat: the substrates being enriched belong to the Marketplace partners, each pointed at a different paying buyer's decision (which campaign, which model to fine-tune, which trade). Snowflake's contribution is reach and in-place query access, not an asset whose enrichability can be named for a specific payer.

The boundary

Nothing of Snowflake's own data crossed; what moves is partner-owned data products that buyers consume inside Snowflake accounts. The relevant asset boundaries belong to each partner's listing, not to Snowflake.

Under-capture

Not determinable from Snowflake's seat in this event - the author's prior ruling on Snowflake Marketplace is that it is a venue, not an owner, so there is no owner-side rake here to call well- or under-extracted.

Why it matters

This is an example of a position the enrichment lens cannot see. Snowflake owns no substrate in these transactions, does none of the enrichment, and faces neither the data seller nor the data buyer in the contract that matters; under the framework's own definition it therefore holds no rake on the data and cannot hold one on loan. What it holds is the venue, and the venue is taxed on a different surface: value that scales with the aggregate of every deal on the platform rather than with any one of them, that is indifferent to which vendor wins, and that is charged on use repeatedly rather than on the sale once.

The reading that matters is what the absent take-rate means. A venue that declines to charge while supply accumulates is not under-capturing — it is buying the option to charge later, and cheap listing is what makes supply arrive, supply is what makes buyers arrive, and default status is what makes a fee collectable at all. Deferred extraction and under-capture look identical in a single year's accounts; the distinction is whether the venue is accumulating power while it forgoes the fee.

If Marketplace becomes the place data transactions happen, expect a disclosed take-rate, or listing terms that tighten toward one — the tell that the option was real and has been exercised. If growth flattens as open formats erode the switching cost, and no fee ever appears, then the indirect capture was the ceiling and not the strategy. The 277% is the number to watch; the $100M is not yet a number that means anything against a business earning $1.49bn a quarter.

The argument this deal tests: Platform Economics: Three Waves of Value Creation

Related deals

Sources

  1. snowflake.com — primary
  2. contentsquare.com — corroborating
  3. en.wikipedia.org — party background
  4. stockanalysis.com — party background

Announced 2026-06-30

How this was classifiedThe register