Our Take
A vendor's quarterly results are not proof that pricing models need overhaul; they are proof that the vendor is winning under the current model.
Why it matters
Enterprise software pricing is under real pressure as AI workloads shift from predictable transactional patterns to variable, GPU-intensive compute. How vendors respond will determine whether AI adoption stays concentrated in well-capitalized firms or spreads.
Do this week
Finance teams: audit your AI infrastructure costs against contract terms this week so you can identify whether you are locked into mismatched unit economics.
Snowflake reports strong quarter, signals pricing shift
Snowflake delivered what the company described as a "monster quarter," and CEO Frank Slootman used the results to argue that traditional software pricing models are no longer sufficient for the AI era. The company's strong performance, according to the Fortune report, prompted Slootman to comment publicly on why vendors need to rethink how they charge for software as enterprises scale AI workloads.
The core argument: existing pricing tiers built on per-seat licensing or consumption-based models (gigabytes stored, queries executed) do not account for the variable, compute-intensive nature of AI inference and training. Snowflake itself offers consumption-based pricing, and the company's financial strength suggests the model is working for it today.
The real pressure: cost unpredictability in AI workloads
Slootman's public comments reflect a genuine market tension, even if his timing is self-serving. Traditional software pricing assumes predictable usage patterns. A sales team of 50 people buys 50 seats. A data warehouse stores X terabytes. Both forecasts hold stable month to month.
AI workloads break that assumption. A single fine-tuning job on a proprietary dataset can spike compute demand 10x in a day. A model serving batch inference at variable load consumes resources unpredictably. Enterprises budget for steady-state costs but hit surprise bills when demand spikes.
This is not new friction. Cloud providers (AWS, GCP, Azure) have wrestled with it for a decade. But enterprise software vendors (Snowflake, Databricks, others) now face direct pressure: customers deploying large-scale AI want either (1) predictable, capped pricing, or (2) transparent, real-time cost visibility so they can control spend without guessing.
Slootman is correct that the status quo is unsustainable at scale. Vendors who stick with opaque per-query or per-GB pricing will lose customers to competitors offering simpler tier structures or fixed-price plans. But that does not mean the market is broken—it means the market is sorting.
What to do now
If you run AI infrastructure at scale, your contract negotiations have shifted. Do not accept pricing that ties costs to metrics you cannot predict (queries, tokens, inference calls). Instead, push for either fixed-tier pricing with guaranteed overage terms or transparent unit costs with a monthly cap.
Request a cost audit from your vendor. Ask them to show you the p50 and p95 spend for your workload type over the past quarter. Compare it to your contract rate. If there is daylight, use it in the next renewal.
Do not assume a "strong quarter" from a vendor means their pricing is fair to you. It means the pricing is fair to them.