Hyperscalers’ AI spending puts SaaS models under pressure

Binaya Dahal

Journalist

20 March 2026

Soaring investment in artificial intelligence (AI) is intensifying scrutiny on software-as-a-service (SaaS) business models, as investors assess whether subscription-based software can withstand the rise of AI-driven agents, according to Lonsec. 

The research house said as the hyperscaler’s expenditure is expected to exceed US$600 billion in 2026, with about three‑quarters of that spending going directly into AI, investors are increasingly focused on which parts of the value chain will capture returns.

Lonsec’s Manager of Global Equities, Hong Hon, said this shift has raised the question on strength and durability of subscription‑based SaaS offerings.

“After Anthropic unveiled its agentic ‘co‑work’ tools, which automate multi‑step workflows and threatens many subscription‑based SaaS offerings, the S&P 500 Software & Services cohort sold off hard and has remained lower year‑to‑date,” he said.

“Investors are now questioning pricing power, margin durability and whether AI agents may cannibalise the traditional ‘per seat’ subscription model.”

Some analysts have labelled the sector’s drawdown as “SaaS‑mageddon” or “SaaS‑pocalypse”, arguing that AI agents flatten the costly software middle layer by shifting execution into APIs and outcome‑based compute.

However, Hon said it does not appear to be a fundamental collapse yet, with many SaaS firms still seeing solid adoption, though several have hinted at margin pressure as AI spending ramps.

He said the opening months of 2026 effectively stress‑tested software business models and showed which companies can adapt pricing structures to an agent‑driven world.

“2026 has opened with more capital intensity, more leverage and a sharper focus on who monetises AI workflows,” Hon said. “The big-picture thesis – AI as a durable productivity wave, still stands.”

“Being invested now means being selective – watching upstream cash flows, downstream unit economics and increasingly, how software pricing models evolve in an agent‑driven world.”