Why are SaaS gross margins under pressure in 2027?

Published June 14, 2026 · Updated June 14, 2026

Direct Answer

SaaS gross margins are under pressure in 2027 because the foundational assumption of the software business — that software has near-zero marginal cost — is breaking as AI features carry real, recurring compute cost. For two decades, software companies enjoyed 75–85% gross margins because serving one more customer cost almost nothing.

AI changes that: every AI inference (a generated summary, a drafted email, an agent action) consumes GPU time or a paid model-API call, so AI is effectively the first software with a genuine, usage-scaling cost of goods sold. As companies embed AI to stay competitive, their COGS rises and gross margin falls — and because software valuations are tied to growth *and* margins, that compression has real consequences for Rule of 40, burn efficiency, and how much the company is worth.

For RevOps and finance, this is not an abstract concern — it reshapes pricing, unit economics, and reporting. The practical response: track gross margin net of AI compute cost, instrument cost-per-customer and cost-per-AI-feature, price AI features to a real margin floor, manage the power users who blow up COGS, and work with engineering on inference efficiency. The companies that ignore the AI cost line grow revenue into compressed or even negative-margin AI usage; the ones that treat AI COGS as a first-class number protect both margin and valuation.

This guide explains the pressure and what to do about it.

The Old Assumption: Software Has No Marginal Cost

Classic SaaS economics rested on a simple truth: once the software is built, serving an additional user costs almost nothing — a bit of hosting and support. That near-zero marginal cost is why software commanded 75–85% gross margins and premium valuations versus hardware or services businesses.

Investors priced software on revenue multiples precisely because those high, scalable margins meant most revenue dropped toward profit as the company grew.

That assumption shaped everything — pricing (per seat, since cost did not scale with usage), the Rule of 40, and how the market valued growth. In 2027, AI is quietly dismantling it, and the businesses that still assume "software has no COGS" are mis-pricing and mis-forecasting their own economics.

Why AI Breaks SaaS Gross Margins

AI features have a cost structure unlike anything in classic software: they cost real money every time they run. A summary, a generated reply, an agent completing a task — each consumes GPU compute or a paid call to a model provider like OpenAI, Anthropic, or Google. That cost scales with usage, so the more your customers use the AI features you built to differentiate, the more your COGS rises.

The result is a structural shift: a product that was 82% gross margin as pure software might run materially lower once AI usage is heavy and unpriced. AI is the first software feature with a true variable cost of goods, which is why finance leaders now talk about "AI-adjusted gross margin" as a distinct number.

A heavy AI user can be a revenue-positive but margin-thin — or even margin-negative — customer, exactly the kind of economics software never had to worry about before.

The Other Cost Pressures

AI is the headline, but several forces compound the margin squeeze. Cloud and infrastructure costs have risen, and AI workloads are especially compute-hungry. The shift to usage-based pricing — increasingly common — can compress margin if usage is not priced above its delivery cost.

Data and storage costs grow as products handle more data for AI. And dependence on third-party model providers means a chunk of your COGS is set by another company's pricing, exposing you to their changes. Together these turn the cost side of software from an afterthought into a number RevOps and finance must actively manage.

What Falling Margins Mean for Valuation and Rule of 40

Margins are not just an operational concern — they directly drive value. The Rule of 40 (growth rate plus profit margin should exceed 40%) weakens when gross margin falls, because there is less profit to contribute. Software is valued on revenue multiples that assume high, durable gross margins; when those margins compress, the multiple the market is willing to pay compresses with them.

In 2027, investors increasingly scrutinize AI-adjusted gross margin and ask whether a company's AI revenue is actually profitable or just impressive-looking top line.

For a RevOps leader, this means the gross-margin line is now a board-level metric tied to the company's worth, not a finance footnote. A growth story built on heavily-used, under-priced AI features is a weaker story than the revenue alone suggests.

How RevOps and Finance Should Respond

The practical playbook for protecting margin:

• Track gross margin net of AI compute cost as a first-class metric, and break out AI COGS so you can see it.
• Instrument cost-per-customer and cost-per-AI-feature, so you know which features and which customers drive the most compute cost.
• Price AI features to a real margin floor — usage tiers, consumption pricing, or premium tiers — rather than giving unlimited AI on a flat fee that ignores cost.
• Manage power users whose heavy AI usage erodes margin, through pricing, limits, or upsell to a tier that covers the cost.
• Partner with engineering on inference efficiency — caching, smaller or routed models, and prompt optimization materially cut COGS, so model choice is now a margin decision RevOps should see.

RevOps and finance jointly own the AI-margin model, because protecting gross margin is now as much a go-to-market and pricing decision as an engineering one.

The Offsets: Model-Cost Deflation and Efficiency

It is not all downward pressure. Model costs are falling fast — the price per token from major providers has dropped steeply and continues to, so the same AI feature gets cheaper to run over time. Inference efficiency is improving — smaller fine-tuned models, caching, and smart routing cut cost dramatically.

And AI features command premium pricing that, if set correctly, more than covers their cost and can even lift margin. The companies that win pair these offsets with discipline: they ride model-cost deflation, invest in efficiency, and price AI to capture value — turning a potential margin threat into a managed, even accretive, cost line.

The threat is real, but it is manageable for operators who treat AI COGS as a number to optimize rather than ignore.

FAQ

Why are SaaS gross margins falling when software used to be 75–85% margin? Because AI features have a real, recurring cost of goods that classic software did not. Every AI inference consumes compute or a paid model-API call, so COGS now scales with usage. As companies embed AI to compete, that cost rises and compresses the historically high software gross margin — AI is effectively the first software with a genuine marginal cost.

What is AI-adjusted gross margin? Gross margin calculated with AI compute and model-API costs fully counted in COGS. Finance leaders track it separately because AI revenue can look strong on the top line while being thin or negative on margin once inference cost is included. It is becoming a key metric investors scrutinize to judge whether AI revenue is actually profitable.

How does margin compression affect company valuation? Software is valued on revenue multiples that assume high, durable gross margins, and the Rule of 40 weakens as margin falls. When AI compresses gross margin, both the Rule-of-40 score and the multiple the market will pay tend to compress.

A growth story built on heavily-used, under-priced AI features is worth less than the revenue alone implies.

What should RevOps do about AI costs eroding margin? Track gross margin net of AI compute as a first-class metric, instrument cost-per-customer and cost-per-AI-feature, price AI features to a real margin floor with usage or premium tiers, manage power users who blow up COGS, and partner with engineering on inference efficiency.

Protecting margin is now a pricing-and-go-to-market decision, not just an engineering one.

Is the AI margin pressure permanent? Not necessarily. Model costs are falling steeply, inference efficiency keeps improving, and AI features can command premium pricing that covers their cost. Operators who ride model-cost deflation, invest in efficiency, and price AI to capture value can manage or even improve margin.

The pressure is real but manageable for those who treat AI COGS as a number to optimize.

Sources

• Bessemer and a16z research on AI-era SaaS gross margins, inference cost as COGS, and AI-adjusted economics, 2026–2027.
• Public software-company disclosures and investor commentary on gross-margin trends as AI features scale.
• Model-provider (OpenAI, Anthropic, Google) pricing histories showing token-cost deflation over time.
• Analysis of the Rule of 40, revenue multiples, and how gross margin drives software valuation.
• Pulse RevOps operator analysis of AI-adjusted gross margin, cost-per-feature instrumentation, and AI pricing, 2026–2027.

*SaaS gross margin review / SaaS margin compression reviews / AI gross margin rating / SaaS gross margin review 2027 / review of why SaaS gross margins are under pressure in 2027.*