Will Snowflake's cloud infrastructure costs remain above 70% of product COGS for 2+ consecutive quarters by H1 FY2027?

Q4 FY2026 earnings provided direct evidence bearing on this market. FY2026 product gross margin came in at 75.8%, meaning product COGS was 24.2% of product revenue. Management guided FY2027 product gross margin DOWN to 75%, explicitly confirming margin compression. This is the first time in the analysis that management has guided gross margins lower rather than flat-to-higher. The CFO directly stated 'margin profile for AI products aren't as high as core business,' confirming the mechanism: AI workloads are structurally margin-dilutive. The 75% FY2027 guide implies product COGS of 25% of revenue — but this is at the product level, not the cloud infrastructure level specifically. However, the direction is unambiguous: margins are compressing, and AI is the stated driver. The question is whether cloud infrastructure specifically stays above 70% of product COGS. Given Q3 FY2026 was at 72%, and the gross margin compression guidance implies COGS pressure continues, the most likely path is persistence above 70%. The CFO's comment about 'finding efficiencies in core business' to offset AI margin dilution suggests the offset mechanism is in non-cloud cost lines (personnel, overhead), not in cloud infrastructure costs themselves. The 200-person RIF supports this — headcount reduction is a non-cloud cost efficiency. Cloud infrastructure costs themselves remain elevated due to AI GPU demand. Calibration note: applying the lesson from the guidance market, I resist excessive bearishness but in this case the direct evidence (margin compression guidance, CFO confirmation of AI margin dilution) is unambiguous and points HIGHER.

Updating from the prior 60% aggregate based on Q4 FY2026 earnings. The prior prediction was made with genuine uncertainty — committee said single-quarter COGS spike was 'insufficient to conclude structural.' Q4 earnings substantially resolved this uncertainty in the direction of persistence. Three key developments: (1) FY2026 product gross margin was 75.8%, confirming that the Q3 COGS spike persisted through Q4. If cloud infrastructure COGS had reverted below 70% in Q4, we'd expect full-year product GM above 76%. The 75.8% figure is consistent with cloud infrastructure COGS remaining in the 70-72% range. (2) Management guided FY2027 product gross margin DOWN to 75%, explicitly projecting further compression. This is forward-looking confirmation that the cost pressure continues. (3) CFO directly attributed margin dilution to AI workloads — 'margin profile for AI products aren't as high as core business.' This converts the prior 'unresolvable uncertainty' into a confirmed mechanism. The question requires BOTH Q4 FY2026 AND Q1 FY2027 to be above 70%. Q4 data appears likely to show persistence based on the full-year GM figure. Q1 FY2027 guidance at 75% GM implies continued compression. The 'efficiency in core business' offset (including the 200-person RIF) could marginally improve cloud infrastructure ratios, but the structural AI margin dilution is now confirmed. Moving probability from 60% to 68%, reflecting the conversion of uncertainty into confirmed direction.

Strong directional update from Q4 earnings. Prior prediction was 60% based on single-quarter uncertainty. New evidence resolves much of that uncertainty: FY2026 product GM at 75.8% (implying Q4 COGS stayed elevated), FY2027 GM guided down to 75% (margin compression continues), and CFO confirmed AI is margin-dilutive. The 200-person RIF and headcount discipline show management finding efficiencies, but these are personnel costs, not cloud infrastructure. Cloud commitments are non-cancelable and escalating. The 70% threshold needs to hold for 2 consecutive quarters (Q4 FY2026 + Q1 FY2027). Q4 data appears likely above 70% based on full-year figures. Q1 guidance suggests continued pressure. Remaining risk to YES: management could find cloud-specific efficiencies (renegotiated contracts, workload optimization) that bring infrastructure costs just below 70% even as overall margins compress. This is possible but unlikely given the confirmed structural mechanism. Moving from 60% to 66%.