Snowflake's Native Apps Turn Data Warehouse Into SaaS Platform, Shifts $5.5B in Buyer Spend

Snowflake repositions as SaaS platform, not just data warehouse

Snowflake expanded its Native Apps and AI Data Cloud capabilities at its June Summit, positioning itself as a full SaaS platform substrate. Vendors can now build, distribute, and monetize complete SaaS applications that run inside customers' Snowflake accounts with direct access to governed data and built-in AI services. This architectural shift matters because it inverts the traditional SaaS model: instead of customers piping data to vendor-controlled multi-tenant environments, the application executes where the data already lives.

The numbers indicate enterprise adoption is accelerating. Snowflake reported $3.4 billion in product revenue for FY2025, up 33% year-over-year. Remaining performance obligations reached $5.5 billion, up 41%, signaling committed future consumption from both SaaS builders and direct enterprise customers. The Snowflake Marketplace grew to more than 700 listings, expanding over 50% year-over-year, with many of those being SaaS-native applications rather than static datasets.

What changes for enterprise architecture

The embedded SaaS model fundamentally alters data gravity economics. ISVs ship Snowflake Native Apps that execute using Snowpark, container services, and Snowflake's governance layer inside the customer's tenant. Data never leaves the customer's Snowflake environment, eliminating cross-cloud egress costs, reducing latency for data-intensive workflows, and simplifying compliance for regulated industries. A financial services firm running fraud detection or a healthcare provider doing clinical analytics can consume vendor AI models without transferring protected data to third-party infrastructure.

Apps can call Snowflake Cortex directly—its fully managed LLM and vector search service—letting vendors deliver AI features without provisioning separate LLM infrastructure. This removes a layer of operational complexity and a separate line item from the vendor's cost base, which can translate to more competitive pricing or faster feature velocity.

The competitive frame shifts from Snowflake-versus-Databricks in data warehousing to Snowflake-versus-AWS Marketplace plus serverless for SaaS distribution. Databricks offers a similar pattern with Lakehouse Apps running inside customer Data Intelligence Platform environments. Hyperscaler platforms—BigQuery, Redshift, Synapse—also enable embedded analytics and marketplace distribution, but Snowflake's cross-cloud neutrality and installed base give it a distinct channel for SaaS vendors targeting multi-cloud buyers.

MongoDB Atlas consolidates SaaS backends around AI-native database

MongoDB expanded Atlas Vector Search, Stream Processing, and serverless instances to GA, positioning Atlas as a turnkey AI-ready backend for SaaS platforms. Atlas now represents 70% of MongoDB's $1.92 billion in total revenue for FY2025, growing 37% year-over-year. The company added approximately 7,000 customers in the last fiscal year, reaching over 47,000 total.

The architectural angle: Atlas integrates vector search and RAG-style retrieval into the same operational database handling transactional workloads. SaaS vendors can avoid deploying separate vector databases like Pinecone or Qdrant, collapsing the stack and reducing the number of systems to operate and pay for. Atlas Stream Processing adds real-time event handling natively, supporting in-product analytics, personalization, and security monitoring without external stream processors.

For enterprise buyers, this creates a vendor consolidation opportunity. Standardizing on Atlas as the backend for internal and external SaaS applications eliminates separate vector database contracts, lowering integration overhead and vendor risk. The trade-off: serverless and autoscaling reduce over-provisioning but introduce per-request pricing exposure. Spiky SaaS traffic patterns can generate unpredictable costs if workload modeling is weak. Buyers need detailed usage forecasts and cost ceilings built into Atlas contracts.

Backblaze storage pricing pressures hyperscaler margins

Backblaze updated B2 Cloud Storage pricing and performance tiers, maintaining its position as a low-cost S3-compatible backend for SaaS storage. B2 lists at $6 per TB per month for storage and $10 per TB for egress, compared to AWS S3 Standard at roughly $23 per TB per month in many regions—three to four times higher. Backblaze reported $109 million in revenue for 2024, up 22% year-over-year, with B2 as the primary growth driver. The platform now processes more than 3 billion objects.

SaaS vendors building multi-cloud architectures or cost-sensitive storage tiers increasingly use B2 as a secondary or primary object store. For media, backup, or archival SaaS products where egress and storage costs dominate the bill of materials, B2's pricing can materially improve gross margins. The architectural implication: B2 becomes a credible alternative to hyperscaler lock-in for workloads that do not require tight integration with compute services in the same cloud.

Enterprise buyers evaluating SaaS vendors should ask whether the vendor's storage backend locks them into hyperscaler pricing or uses lower-cost tiers like B2 or Wasabi. Vendors passing through hyperscaler storage costs at markup create budget exposure as data volumes grow. Vendors architecting for cost efficiency signal better long-term pricing discipline.

What to watch

Snowflake's Native Apps model and MongoDB's AI-native database consolidation both push toward fewer, thicker platform dependencies. Enterprises should evaluate whether concentrating SaaS workloads into a single platform (Snowflake for data apps, MongoDB for operational AI) reduces vendor sprawl or creates new lock-in risk. Multi-cloud strategies require clear exit paths and data portability guarantees.

For SaaS vendors, the shift to embedded apps and consolidated backends accelerates. Vendors not architecting for data gravity—running models where enterprise data lives rather than forcing data movement—will face increasing buyer resistance. Storage cost discipline also matters: SaaS vendors absorbing hyperscaler egress fees without architectural alternatives signal weak margin control, which eventually surfaces as pricing pressure or reduced feature investment.

The broader pattern: platform vendors are pulling more of the SaaS stack into their managed services, and buyers need to decide whether that consolidation lowers total cost of ownership or simply shifts spend from multiple invoices to one larger one.