From SaaS to service as software: How AI is launching a 20-year enterprise super-cycle and what it means for investors
The “SaaS-pocalypse” narrative is getting the causality backwards. Public software valuations are being priced for obsolescence right as incumbents pivot to “service as software,” using agentic AI to sell outcomes, not seats, and expanding software’s addressable market from IT budgets to that of the labor market. As benchmarks like ARC-AGI-2 push past the 60% “labor replacement” milestone at around $1 to $10 per task, the economic logic flips: an annual charge of $1,200 per seat becomes an annual charge of $10,000 per automated workflow, with software spend increasingly competing with payroll spend.
Meanwhile, hyperscale AI leaders are laying the groundwork for deployment at scale, with AI infrastructure capital expenditure surging toward $1 trillion annually—the classic “infrastructure precedes applications” playbook. The winners will not be generic user-interface wrappers; they will be systems-of-record incumbents and AI-native operators that combine distribution, data context, workflow integration, and governance into durable moats, then monetize via outcome pricing. For investors, the call to action is simple: Reassess software exposure through a service-as-software lens (watch gross margins and workflow-wrapper risk), lean into defensible platforms (security, payments, and mission-critical infrastructure), and underweight seat-based value traps and billable-hours-heavy services models.
Table of contents
| Key takeaways | 1 |
| Executive summary | 2 |
| Enterprises wake up | 3 |
| Infrastructure precedes application | 4 |
| Productivity to replacement | 5 |
| Structural moats and the enemies at the gates | 6 |
| The future, today | 10 |
| Recommendations to investors and vendors | 11 |
| PE and private credit | 11 |
| VC and corporate VC | 12 |
| Public equity | 13 |
| SaaS founders | 14 |
| IT managers and chief information officers | 15 |
| Conclusion: Fear is the accelerator | 16 |