Snowflake's AI Data Cloud has a consumption math problem—and earnings will show if it can still win on volume

Snowflake’s AI Data Cloud thesis is simple: if customers use more—compute, storage, and data transfer—Snowflake’s “consumption” revenue should follow. The tricky part is timing. Because revenue is recognized as customers consume resources (not evenly across contract terms), Snowflake can be “right on demand” while still printing quarter-to-quarter results that move more than subscription software investors expect.

That’s why this upcoming earnings window matters as the market’s first real check on whether AI Data Cloud consumption accelerates fast enough to re-rate the business—especially as Databricks’ private-company valuation narrative keeps a high bar on how much AI data infrastructure can be monetized.

Snowflake’s filings describe an AI Data Cloud, delivered with a customer-centric consumption-based business model. The key mechanics for investors are:

• Revenue comes from fees tied to resources consumed: compute, storage, and data transfer.
• Customers can choose compute, storage, and data transfer independently.
• Compute consumption depends on compute type and duration of use (and certain features by processed volume); storage is priced on average terabytes per month; data transfer is priced based on terabytes transferred, public cloud provider, and region.
• Capacity arrangements (majority of revenue) are typically multi-year, with billing patterns that can differ from usage timing. On-demand is charged monthly in arrears.
• Crucially, Snowflake recognizes revenue as customers consume resources. It also notes it generally lacks visibility into the timing of revenue recognition for individual contracts.

The implication is direct: even with steady customer adoption, a quarter’s product revenue can be more driven by usage timing than by contract spend.

Snowflake reported Q2 FY2026 product revenue of $1.09B (+32% YoY) and guided Q3 FY2026 product revenue to $1.125–$1.130B (25–26% YoY), with full-year FY2026 product revenue guidance at $4.395B (+27% YoY). On paper, that’s clean software-like momentum.

But the consumption model matters: the company’s own disclosures emphasize that revenue recognition depends on consumption timing, which is inherently variable across customers and use-cases.

Because Databricks is private, investors can’t watch public revenue/earnings prints to calibrate valuation. Instead, the market often anchors on fundraising and reported run-rate claims. TechCrunch reported Databricks closed a $5B strategic round at a $190B valuation and cited a $7B annualized revenue run-rate.

This is where the “pricing vs. consumption” question becomes investment-relevant for Snowflake: if AI Data Cloud adoption ramps but does not translate into measurable resource usage within a quarter, Snowflake may struggle to deliver the consumption conversion the market expects.

AI Data Cloud consumption sits downstream of a broader supply chain: models, orchestration, data movement, and execution environments. Even if customers want to “use Snowflake data,” the unit economics depend on where computation actually happens and how frequently data is moved.

From Snowflake’s own pricing mechanics, consumption revenue depends on:

• How much compute time is actually run in Snowflake for relevant workloads.
• How much curated data is stored as models and applications iterate.
• How often data is transferred out (and between regions/clouds).

So the earnings question becomes: are AI teams running enough real workloads that match Snowflake’s priced resource meters? Or are they experimenting at the edges—where data landing happens, but repeated compute/storage/data-transfer consumption isn’t yet large enough to change the revenue curve?

Using listed-company fundamentals as guardrails, Snowflake is still not a near-term profitability story. Its net income has been negative in FY2024 and FY2025 in the income statement data provided, consistent with high reinvestment.

That increases the importance of how management frames product revenue quality (usage depth, retention, and timing) and how much of the AI Data Cloud cycle is showing up in current quarter consumption.

Short-term (this quarter and the next one): the market will focus on whether recognized product revenue tracks with AI Data Cloud adoption in a way that reduces uncertainty around timing. Because revenue depends on consumption, investors should expect management to highlight utilization and to give guidance that assumes believable workload ramp.

Long-term (1–3 years): investors should ask whether consumption-based economics become more predictable as AI workloads mature—more stable compute cycles, deeper data residency, and more repeatable data transfer patterns. If Snowflake can make consumption conversion smoother across quarters, it can justify a valuation more closely aligned with the “infrastructure winner” narrative—without needing to lean on pricing alone.