SaaS renewals in H2 2026: what CFOs are actually cutting

The H1 2026 renewal cycle just finished. We ran scope calls with 40+ mid-market CFOs and heads of procurement over the last 90 days, and the pattern is sharp enough to write down.

The overall shape: SaaS spend as a % of ARR peaked in Q3 2024, plateaued through 2025, and is actively falling in H1 2026 for the first time since ~2017. This isn’t a headline story because the fall is uneven — some line items are up, some are down, some are being replaced. Here’s the sort-by-what-CFOs-are-actually-cutting.

Category 1: Per-seat wrappers on commodity primitives (being cut aggressively)

Any SaaS that charges $150–$400 per seat per month for what is fundamentally a UI on top of a commodity primitive (Postgres, S3, Twilio, LLM API) is getting cut, not just renegotiated. What’s in this bucket by our count:

• Sales engagement tools (Outreach, Salesloft) — average cut: 40–60% of seats. Kept for SDRs, dropped for AE overlays.
• Product analytics wrappers (Mixpanel, Amplitude at Enterprise tier) — average cut: full replacement with warehouse-native (dbt + Postgres/Snowflake + Metabase). Payback: 8–14 months.
• Marketing automation for D2C (Klaviyo, Braze at enterprise tier) — rebuilds accelerating; SES-direct + Postgres economics are now 12–20% of the wrapper cost.
• Observability at consumption tier (Datadog, New Relic overages) — the tail (log ingest above the plan) is being aggressively cut. See our Datadog bill shock post.

What’s common to this list: the vendor priced when the underlying LLM / infra was more expensive, priced didn’t move down as unit costs improved by 3–5×, and the wrapper margin ballooned. CFOs finally noticed.

Category 2: Enterprise sales-driven contracts (being renegotiated hard)

Salesforce, Workday, ServiceNow, HubSpot Enterprise: the top-tier enterprise SaaS lines that account for 60% of most mid-market SaaS bills. These aren’t being cut wholesale (yet) — they’re being renegotiated with a “walk-away is credible now” posture.

What’s happening tactically:

• Seat rationalisation. The “120 seat expansion” that got signed in 2023 without a real audit is being wound back to actual usage. Typical outcome: 25–40% seat count reduction at renewal, negotiated on threat of walk.
• Feature tier downgrade. Einstein 1 / Enterprise Analytics / Premium Support that were sold as “bundled value” are being unbundled and cut. Base tier + no add-ons, or drop to Professional tier.
• Term flexibility. Multi-year commits with escalators are being replaced with 1-year renewals so the walk-away option stays open annually.

What is not happening: most of these accounts aren’t moving to a rebuild yet. The switching cost is real and the political capital is finite. But every scope call ends with “send us the rebuild math anyway, we’re building the case for 2027.” The mid-market Salesforce base in particular is being underwritten to walk in 2027 renewals.

Category 3: Point tools that solved a 2022 problem (being sunset)

The category CFOs are least loud about, but the biggest count by SKU. Every mid-market org accumulated 40–120 point-tool SaaS subscriptions between 2021 and 2024. In 2026 the discipline of “does anyone still use this” is finally landing.

Typical outcomes from a stack audit:

• 25–35% of SaaS subscriptions have no active user in the last 60 days. Full cancel.
• Another 10–15% have < 5 active users at a bulk-license tier. Downgrade or cancel.
• Another 5–10% are duplicates. Different teams bought different tools for the same job. Rationalise to one.

Combined, a typical stack audit surfaces 30–50% cancellable SaaS spend at zero business impact. The audit itself pays back in less than one billing cycle.

Category 4: What CFOs are NOT cutting (mostly)

Three categories that are surprisingly sticky:

• Security / compliance tooling. Vanta, Drata, cloud security posture management. These are being kept and often expanded because the alternative — an audit failure — costs 100× the SaaS bill.
• Foundational SaaS. Google Workspace, Microsoft 365, Slack, Zoom. Renegotiated at the edges but not replaced.
• Genuinely differentiated AI tooling. Anthropic and OpenAI direct APIs (both up in usage), plus a small number of AI product tools that provide real leverage. This is where SaaS spend is growing, even as the wrapper tier is being cut.

What the CFO conversation actually sounds like now

The three-question filter we’re seeing procurement leaders apply at every renewal:

1. “Is this vendor priced from a time when their unit costs were 3–5× higher? If so, why hasn’t their price come down?”
2. “What percentage of the features are we using? If under 30%, why is the price the enterprise tier?”
3. “If we cancelled, what’s the rebuild math?”

The third question is the one that’s new. It didn’t used to be part of the frame because “engineers are expensive and it’ll take a year” was true enough to shut down the conversation. Now the rebuild math (see what Opus 4.8 changed) actually competes, and the SaaS vendor has to defend against it.

What this looks like for the Rebuild engagement pipeline

Practical concentration in H2 2026:

• Marketing automation rebuilds (Klaviyo / HubSpot Marketing / Pardot replacement) — heavy pipeline. Payback typically 6–10 months.
• Observability rebuilds (Datadog / New Relic replacement) — steady flow, mostly larger clients where the bill has crossed $500K/yr.
• Sales ops bundles (Sales Cloud + Outreach + Gong + Apollo) — growing pipeline; multi-tool bundles are winning over single-tool at this size.
• Legacy modernization (see the legacy modernization inflection post) — growing, but a different economic story: not wrapper-margin, but license + contractor scarcity.

The through-line: what worked economically as a per-seat wrapper for the vendor when their input costs were high has ceased to work. CFOs finally noticed. Renewal cycles from Q3 onwards are going to be uglier than any since ~2017.

---

Read more: /rebuild/ · SaaS Spend Audit · Per-seat SaaS math is broken · Stack rationalisation executive brief