App Economy Insights — Snowflake: AI Consumption Wins (May 2026)

One-line take: A two-model read (not stock advice). Snowflake — product revenue +34% to $1.33B (accelerating, beating its own guide by 7pp) and +38% after hours, erasing the YTD drawdown: a consumption business that gets paid more as AI agents run, with a 5-yr $6B AWS commitment (Graviton + custom accelerators supporting a 75% product gross margin), NRR 126%, the Natoma MCP acquisition to govern what agents do, and Cortex Code/Snowflake Intelligence reaching data inside Microsoft/Salesforce/SAP apps. Salesforce — the seat-based-SaaS dilemma ("the penalty box"): organic ~9%, cRPO +13% missed, a $25B debt-funded buyback that cut FCF-growth guide to 4-5%, Marketing/Commerce/Tableau dragging, and the "value-abstraction" pricing risk in Headless 360 (Salesforce becomes the data store behind someone else's agent) — Agentforce ARR $1.2B is real but small. Referenced: AMZN (the $6B AWS Graviton/accelerator deal), MSFT + SAP (data sources Snowflake now reaches). "Views" are App Economy's framing — SNOW constructive, CRM cautious.

1. Stocks & names mentioned

Ticker	Name	Research	View	What's said	Source
SNOW	Snowflake	QT · SA · STK · FA	Positive	The consumption winner. Product rev +34% to $1.33B (accelerating from +30%, beating its own projection by 7pp); FY product guide raised to $5.84B (+31%); a 5-yr $6B AWS commitment (Graviton + custom AI accelerators) supports a 75% product gross margin. Q rev +33% to $1.39B ($70M beat), non-GAAP EPS $0.39 ($0.07 beat), NRR 126%, 46 customers >$1M (vs 26 a year ago). Cortex Code >7,100 accounts; Natoma (MCP) acquired to govern what agents DO (acting agents consume far more compute). RPO +38% to $9.21B but missed $9.43B — backlog matters less for consumption. "Agents increase the need for governed data — exactly where Snowflake sits."	article ↗
CRM	Salesforce	QT · SA · STK · FA	Negative	The seat-based dilemma — "the penalty box." Q rev +13% (just ~9% organic ex-Informatica) to $11.1B ($70M beat), non-GAAP EPS +50% to $3.88 ($0.75 beat), but cRPO +14% to $33.6B missed $34B+. A $25B debt-funded accelerated buyback (largest ever) cut operating/FCF growth guide to 4-5% (from 9-10%). New 2-bucket disclosure hides Marketing/Commerce + Tableau softness. Agentforce ARR $1.2B (+205%) real but small; Headless 360 lets external agents (incl. Anthropic's Claude Code) reach Salesforce data — but pricing is unclear ("value abstraction" risk). Until cRPO re-accelerates and Headless 360 has a pricing model, it may stay in the penalty box.	article ↗
AMZN	Amazon (AWS)	QT · SA · STK · FA	Neutral	Referenced via AWS — Snowflake's 5-year $6B commitment uses AWS Graviton CPUs and custom AI accelerators to lower compute cost and support Snowflake's 75% product gross margin. Also a disclosed App Economy Portfolio holding.	article ↗
MSFT	Microsoft	QT · SA · STK · FA	Neutral	Named two ways — Snowflake's Cortex Code / Snowflake Intelligence reach data inside Microsoft apps (widening Snowflake's moat), and Microsoft Fabric is a competitor in the agentic data layer. A disclosed App Economy Portfolio holding.	article ↗
SAP	SAP SE	QT · SA · STK · FA	Neutral	Named as one of the enterprise application data sources Snowflake's Cortex Code / Snowflake Intelligence now reach (alongside Microsoft and Salesforce apps) — part of the "wider moat" of governed access to data wherever it lives.	article ↗

"View" is App Economy's analytical framing in this post — SNOW positive (the consumption model wins as agents need governed data), CRM negative (the seat-based dilemma + missed cRPO + value-abstraction pricing risk); AMZN/MSFT/SAP are referenced/neutral (the AWS deal + the data sources Snowflake reaches). App Economy Insights is financial-analysis journalism, not a buy/sell stance. Research: QT Qualtrim · SA Seeking Alpha · STK Stock Analysis. The "Source" links open the newsletter (no per-name timestamps — it's a written post). Note: the original post's "Mode Mobile / $MODE" block was a paid advertisement and is excluded.

2. Talking points

Software is splitting in two — consumption vs seat-based

AI is forcing a re-rating of which software model deserves a premium. Snowflake surged 38% after hours; Salesforce kept sliding. The new line: companies that get paid MORE when AI works harder (consumption) vs those built on seat licenses for humans who may need fewer seats (traditional SaaS).

Snowflake — product revenue re-accelerates to +34%

Product rev +34% to $1.33B, accelerating from +30% and beating its own projection by 7pp; FY product guide raised to $5.84B (+31%, from $5.66B). Total Q revenue +33% to $1.39B ($70M beat), non-GAAP op margin 12%, non-GAAP EPS $0.39 ($0.07 beat).

The AWS handshake — a 5-yr $6B commitment

Snowflake doubled down on AWS with a 5-year $6B commitment for Graviton CPUs + custom AI accelerators — lower compute cost that helps support a 75% product gross margin. The cloud-provider relationship is a margin lever, not just a hosting bill.

NRR 126% and the $1M-customer count nearly doubled

Net revenue retention 126%; 46 customers crossed $1M trailing revenue, nearly double the 26 a year ago. The land-and-expand engine is widening at the top end.

Cortex Code, Snowflake Intelligence — reaching data in other apps

Cortex Code (CoCo) >7,100 active accounts; Snowflake Intelligence accounts more than doubled Q/Q. Customers use these to access data inside Microsoft, Salesforce, and SAP apps — a wider moat that meets data wherever it lives.

RPO +38% but missed — why backlog matters less here

RPO +38% to $9.21B but missed the $9.43B bar. For a consumption business, backlog is a weaker signal than usage — revenue is recognized as compute is consumed, not as contracts are signed.

Natoma — governing what agents DO, not just read

Snowflake acquired Natoma (enterprise MCP — Model Context Protocol — implementation) to govern what agents are allowed to do, not just what they can read. Agents that act consume far more compute, so governance and consumption reinforce each other.

The competitive set for the agentic data layer

Snowflake competes with Databricks, Microsoft Fabric, and Google BigQuery on enterprise security and identity for the agentic data layer. The thesis: as agents proliferate, the need for governed data, secure workflows, and scalable compute grows — exactly where Snowflake sits.

Salesforce — AI metrics over a clear consumption line

CRM buried earnings under AI metrics (28.6T tokens +152% Q/Q, 3.8B "Agentic Work Units" +111%, 52T records into Data 360) — but with no clear consumption revenue line. Agentforce ARR crossed $1B in ~18 months (one of the fastest ramps ever) and reached $1.2B (+205% Y/Y), net-new ARR ~$400M — real, but small against the base.

The buyback and the disclosure that hides the soft spots

With the stock halved from its peak, CRM announced the largest accelerated buyback in its history ($25B, half of a $50B authorization, debt-funded) — which cut operating/FCF growth guide to 4-5% (from 9-10%). Strip out Informatica ($8B, Nov) and organic revenue grew ~9%; cRPO +14% to $33.6B missed $34B+. A new 2-bucket disclosure (Applications +9%; Infrastructure & Data +25%) is less granular just as investors need visibility — hiding Marketing/Commerce weakness and Tableau softness in commentary.

Headless 360 — the value-abstraction pricing risk

Headless 360 lets external agents (including Anthropic's Claude Code) access Salesforce data — Anthropic is a top customer, usage reportedly +5x in Q1. But pricing is unclear: the "value abstraction" risk is that Salesforce becomes the commodity data store behind someone else's agent, capturing little of the value the agent creates. Until cRPO re-accelerates and Headless 360 has a pricing model, the stock may stay in the penalty box.

3. In plain English

A jargon-free summary of the thesis behind each name — what it actually is and why that view. (Plain-language companion to the table above; renders on each ticker's consolidated page.)

SNOW — Snowflake Positive

Snowflake sells a cloud "data platform" — companies pour their data into it and pay based on how much computing they use to query and analyze that data. That last part is the whole thesis: it's a consumption business. You're not buying a fixed number of seats (logins for people); you're billed for usage. So when AI agents start running queries around the clock, Snowflake's revenue goes up — the harder the AI works, the more the customer pays. That's the opposite of seat-based software, where AI doing the work might mean a company needs fewer human logins.

Two details make the quarter strong. First, the AWS deal: Snowflake committed $6 billion over five years to Amazon's cloud, using Amazon's own efficient chips (Graviton CPUs and custom AI accelerators) to lower its computing costs — which is how it protects a fat 75% gross margin on its product. Think of it as locking in cheap raw materials. Second, the Natoma acquisition: "MCP" (Model Context Protocol) is a standard way to let AI agents safely connect to company systems, and Natoma helps govern what those agents are allowed to do, not just what they can read. Agents that take actions consume far more compute — so better governance literally drives more billable usage. App Economy's read: as AI agents multiply, every company needs governed, secure, scalable access to its data, and that's exactly the spot Snowflake occupies.

CRM — Salesforce Negative

Salesforce is the big "CRM" (customer-relationship-management) software company — sales, service, and marketing tools sold mostly as seats: you pay per user, per year. That's the problem in the AI era. If AI agents do work that humans used to do, companies may need fewer seats — so the core pricing model is exactly the kind AI threatens. App Economy calls this "the penalty box": the stock has already halved from its peak, and the latest print didn't break it out.

The numbers tell the squeeze. Revenue grew 13%, but a chunk of that is the Informatica acquisition; strip it out and organic growth was only ~9%. "cRPO" (current remaining performance obligations — basically contracted revenue due in the next year, a leading indicator of growth) missed expectations. To support the stock, Salesforce announced its largest-ever buyback ($25 billion, funded with debt) — but borrowing to buy back shares forced it to cut its cash-flow growth guidance to 4-5%. And a new, vaguer two-bucket way of reporting revenue conveniently hides softness in its Marketing, Commerce, and Tableau lines.

The most interesting risk is "Headless 360." Salesforce is letting outside AI agents (like Anthropic's Claude Code) reach into Salesforce data. Anthropic is a big customer and usage reportedly jumped 5x. Sounds great — but the pricing is unclear, and the danger ("value abstraction") is that Salesforce ends up as just the boring data warehouse behind someone else's smart agent, while that other company captures the value the AI creates. Until contracted revenue (cRPO) speeds back up and Headless 360 has a real pricing model, App Economy keeps it cautious.

Key points & figures extracted from the public App Economy Insights newsletter (in transcript.txt) for personal study. Not investment advice. © App Economy Insights for source material.