SaaStr warns of AI token spend crisis: 5x costs with no proven revenue lift

The Gist

• Companies quintupled AI token spend H1 2026 with no proven ROI
• Coinbase cut AI spend 50% while increasing usage via open-source
• Boards now demanding provable revenue impact for AI engineering spend

• Companies quintupled their token spend in the first half of the year with no proven revenue lift.
• AI spend on engineering should credibly produce revenue lift for software companies.
• The rise of open-source models may cannibalize revenue for frontier AI companies faster than expected.
• Microsoft's AI strategy is under scrutiny due to lack of a standalone frontier model and decelerating Azure growth.
• Anthropic's revenue growth may surpass all public software companies combined by year-end.
• Bending Spoons' success highlights a potential model for turning around decaying B2B products with AI-driven revenue.

• Tie AI spend directly to revenue lift or cost savings to justify investments.
• Evaluate open-source AI models as a cost-effective alternative to frontier models.
• Monitor Microsoft's AI strategy and Azure growth as a bellwether for the broader AI market.
• Explore AI-driven turnarounds for underperforming B2B products to unlock new revenue streams.

• 5x increase in token spend (Companies quintupled their token spend in the first half of the year with no proven revenue lift.)
• 50% cut in AI spend (Coinbase reduced AI spend by 50% this quarter while usage increased, driven by a shift to open-source models.)
• $44B revenue run rate (Anthropic's revenue run rate grew from $1B at the start of last year to $44B mid this year.)
• 16.5% drop in Microsoft's stock (Microsoft's worst month since 2000, attributed to concerns over its AI strategy and Azure deceleration.)
• $500M in 18 months (Higgsfield crossed $500M in revenue in under 18 months, highlighting outlier growth in AI.)