There is a clean separation happening in enterprise software right now, and most people are still trying to explain it away.

Wall Street has been aggressively unwinding positions in software companies that cannot demonstrate a straight line between AI product announcements and actual billings growth. The iShares Expanded Tech-Software Sector ETF (IGV) peaked in September 2025 and has since fallen roughly 30%. Between mid-January and mid-February 2026 alone, approximately $1 trillion was wiped from the collective value of software stocks – the worst monthly decline for the S&P North American Software Index since 2008. The trigger was a series of AI agent launches that raised an uncomfortable question for per-seat SaaS models: if an agent can do the work, why pay for the seat? That fear, whether fully justified or not, has compressed software price-to-sales ratios from roughly 9x to 6x – levels not seen since the mid-2010s.

Oracle (ORCL) is operating in a different category entirely.

In its fiscal Q2 2026 results, Oracle reported total revenue of $16.1 billion, up 14% year-over-year. Cloud Infrastructure (IaaS) revenue came in at $4.1 billion, up 68%. Total cloud revenue reached $8.0 billion, up 34%. These are not soft growth figures dressed up with AI branding – they are contracted, multi-year revenue obligations that show up in the backlog. Remaining Performance Obligations (RPO) stood at $523 billion as of Q2, up 438% year-over-year. Oracle signed four multi-billion-dollar contracts with three separate customers in Q1 alone, and management expects RPO to exceed half a trillion dollars in the near term. For context, that backlog already includes most of the revenue in a five-year infrastructure forecast that projects OCI revenue reaching $144 billion annually by fiscal 2030.

What matters here is not the AI branding. It is the stickiness of the underlying workload. Enterprise database migrations – the kind Oracle is processing through its sovereign cloud contracts – are not discretionary IT purchases. They do not get cut in a budget rationalization cycle. That is the structural difference between a platform running mission-critical operations and a SaaS vendor still trying to prove its AI layer generates incremental revenue beyond a product announcement.

The broader SaaS picture is more complicated than the selloff implies. Global SaaS spending is projected to grow from $318 billion in 2025 to $576 billion by 2029. Enterprises are not abandoning software – they are reallocating budgets toward AI infrastructure and away from marginal seat additions. That distinction matters for stock selection. The companies that own the infrastructure layer, the data layer, or genuinely mission-critical workflows are in a different risk bucket than those monetizing headcount.

Oracle’s position is not without risk. Execution on a $144 billion five-year revenue target is a substantial operational challenge, and any slowdown in hyperscaler demand or enterprise cloud migration cycles would pressure those RPO conversion rates. Non-GAAP EPS came in at $2.26 for Q2, up 54%, but GAAP EPS was $2.10 – a gap worth watching as capital expenditure scales to meet contracted demand.

The split in enterprise software is not a short-term sentiment trade. It is a structural repricing of what counts as infrastructure versus what counts as tooling. Oracle is firmly in the first category. Where that line falls for the rest of the sector is the question that will define software investing through 2026 and well beyond.