Software isn't dead: AI's $688B investment gap creates 5-year runway for GTM teams

The Gist

• Hyperscalers spend $688B on AI capex vs $110B in revenue in 2026
• Revenue won't catch spend until 2031-2032 at $1T annual run rate
• Founders have 5+ years of cheap AI models and capital influx ahead

• Software isn't dead; it's just harder to win due to massive AI investments creating a 5-6 year runway for GTM teams.
• The AI industry is spending $688B annually on capex but generating only $110B in revenue, creating a $500B+ investment gap.
• AI must capture 15-17% of knowledge worker wages ($50K per $200K developer) to justify the $1T revenue target by 2031-2032.
• The 'harness' layer (software on top of raw models) will be critical for differentiating AI applications, similar to the LAMP stack for SaaS.
• Founders must choose between hypergrowth (5-10x) or capital efficiency, as public software multiples have collapsed due to slowed growth.
• AI can replace sales and marketing if the product sells itself, but companies can't afford high compute costs AND traditional GTM expenses.

• Plan for a 5-6 year investment runway in AI, but prepare for potential market pullbacks before 2032.
• Differentiate your AI product by building defensible 'harness' layers tailored to specific industries (e.g., legal, customer support).
• Optimize for capital efficiency if not pursuing hypergrowth, focusing on category leadership and differentiation.
• Audit whether your AI product can truly sell itself to justify high compute costs, or if traditional GTM is still needed.

• $688B (Hyperscalers' AI capex spending in 2026)
• $110B (Projected AI revenue in 2026 ($89B from foundation model leaders))
• 2031-2032 (Estimated crossover year when AI revenue surpasses cumulative capex)
• 15-17% (Share of US knowledge worker wages AI must capture to hit $1T revenue)
• 25% (Share of developer wages ($50K per $200K salary) going to AI tokens)
• 10% growth (Current average software company growth rate (down from ~30%))