How Much Do US Companies Spend on SaaS in 2026?

In short: the median US company spends about $9,500 per employee, per year on software (Zylo's 2026 SaaS Management Index) — and roughly 46% of those licenses go unused or underused. Most of it earns its keep. But that's a lot of money to leave un-examined, and it raises a sharper question: which parts of the stack are worth owning instead of renting? This guide breaks down where the money goes, what it costs in real prices, and how to tell the difference.

On this page

• Where the money actually goes
• What this looks like in real prices
• Does spend change with company size?
• The real cost is the waste
• Buy the commodity, build the differentiator
• What I did in my own company
• How to audit your own stack
• The bottom line

Software rarely looks expensive one subscription at a time.

$20 here. $100 per user there. A CRM. An HR platform. A support tool. An analytics product. A dashboard. An automation platform. A few AI subscriptions. Then somebody in finance adds it all up.

According to Zylo's 2026 SaaS Management Index, the median US company now spends about $9,500 per employee, per year on software — and the fastest-growing line is AI, with ChatGPT now the single most-expensed app in their dataset.

That does not mean software is bad. Most businesses could not operate without it. But it is enough money to ask a question that got much more interesting in the age of AI-assisted development:

At what point should a company stop adding another SaaS product and build the part of the stack that is specific to how it actually operates?

Where the money actually goes

$9,500 per employee is a per-head figure, so it scales with the company. A rough sense of the totals:

Team size	~Annual software spend
10 people	~$95K
30 people	~$285K
100 people	~$950K

Per-employee spend actually falls as you scale (more on that below), so treat the top of that range as generous — but even the low end adds up faster than most founders expect.

And it is spread across dozens of tools. A company does not decide one morning to buy its whole stack. It accumulates it: sales picks a CRM, marketing adds automation, finance buys forecasting, HR adds employee management, ops needs its own system, product adds analytics, someone wires it together in an automation tool, another team adds an AI assistant.

Each purchase is reasonable in isolation. The complexity appears between them.

What this looks like in real prices

Public list prices, checked September 2026. Vendor pricing, discounts and enterprise contracts change — treat these as illustrative, not quotes.

Salesforce

Salesforce lists Sales Cloud Enterprise at $175 per user per month, with Unlimited at $350 and its AI-forward Agentforce tier higher still. Twenty Enterprise users is 20 × $175 × 12 = $42,000 a year.

The point isn't that Salesforce is overpriced — it's a huge platform and the right answer for many businesses. The point is that one important tool can consume a big share of a small company's software budget by itself.

HubSpot

HubSpot Sales Hub Professional runs about $90–$100 per seat per month, and Marketing Hub Professional starts around $890 per month (plus a one-time onboarding fee) before extra contacts. A modest Sales + Marketing Pro setup lands in the low five figures per year fast. HubSpot may be worth every dollar — but the bill becomes material.

The long tail

Below the big platforms, the rest of the stack is where budgets quietly balloon:

• HR software usually starts around $5–$8 per employee per month — a few thousand dollars a year at 50–100 people, before add-on modules.
• Accounting and ERP tools are a classic case of software that makes more sense to buy than rebuild; their per-employee add-ons (payroll, inventory) are what grow the bill.
• Per-seat productivity tools — team chat, docs and wikis, AI assistants — look small but scale with headcount. Business-tier seats run roughly $15–$30 per user per month, so a 50-person team is $9,000–$18,000 a year for each one. Three of them together clear $30,000 a year — before any CRM, HR, ERP, support or analytics.

Does spend change with company size?

There's no universal per-employee rule. A five-person company still needs accounting, a CRM, email, analytics, support and project management — those fixed costs make per-employee spend look high. As teams grow, they gain economies of scale, and spend per employee generally falls even as the total climbs.

That's why a $9,500-per-employee median is a starting point, not a budget: smaller companies often run higher per head, larger ones lower. Either way, a digitally intensive business reaches a six-figure — then seven-figure — software bill well before enterprise scale.

The real cost is the waste

The invoice is only the visible part. Zylo's 2026 data found that about 46% of applications go unused or underused, and industry research from Gartner, Productiv and Zylo consistently puts wasted SaaS spend at 25–40% of the total — unused seats, duplicate tools and overlapping functionality. At large organizations, that waste runs into the millions per year.

But the deeper cost is operational. Every extra system tends to create another login, another source of truth, another integration, another renewal, another place customer data lives, and another spreadsheet gluing two tools that were never meant to work together. That's when SaaS sprawl stops being a finance problem and becomes an operations problem.

Buy the commodity, build the differentiator

So should companies build everything themselves? No — that's the wrong conclusion. I wouldn't build my own email provider, rebuild a payments processor, write accounting software to dodge $99 a month, or clone a team-chat app for a 30-person team. Commodity software is cheap precisely because thousands of customers share the development cost.

A useful way to split the stack in two:

Commodity software — the problem is essentially the same for every company: email, payments, accounting, video calls, authentication, basic collaboration, cloud infrastructure. Buying usually wins.

Operational software — the problem depends on how your company works: your internal CRM workflow, dispatch, onboarding, quoting, approvals, field ops, stock flows, customer portals, internal dashboards, scheduling, and the glue connecting five existing systems.

Historically, building that internal layer meant hiring engineers or paying an agency for months — which made a $20K–$50K SaaS contract look cheap. AI-assisted development changes the math: a small, well-scoped internal tool can now be built and iterated far faster, especially when it's replacing one workflow rather than trying to become a general-purpose product for thousands of unrelated companies.

What I did in my own company

This is the model I ended up using at Kleta. Instead of treating every operational problem as another subscription or a multi-month dev project, I built the systems around the operation itself — CRM, customer flows, ecommerce, appointments, internal tools and automation. That cut our monthly software bill from around $30,000 to under $250, with no in-house developers.

That experience is now the basis of how I work with other companies — the AI Enablement Partner model. I don't sell a strategy deck telling you to "use more AI." I look at the operation, map the software and the manual work, calculate where the leverage is, and build the parts that make economic sense. Built, not advised.

How to audit your own stack

I don't start a software audit with "what can we cancel?" I start with "what does the business need to happen?" Then, tool by tool, I score:

1. Is it essential, or nice-to-have?
2. What does it actually cost per year, all-in?
3. What percentage of its features do you really use?
4. How much manual work exists around it because it doesn't talk to your other systems?
5. Is this workflow generic — or specific enough to your business that owning it would be an advantage?

Then the comparison that matters isn't "SaaS vs custom" in the abstract. It's:

current software cost + manual work created by the gaps + integration upkeep + cost of errors — versus — cost to build + infrastructure + maintenance.

Sometimes the answer is to keep the SaaS. Sometimes it's to downgrade it. Sometimes three subscriptions become one internal tool. And sometimes the most valuable thing to build isn't a replacement at all, but a custom layer that connects the systems you already have.

The bottom line

The median US company spends roughly $9,500 per employee a year on software, and for software-intensive businesses the bill reaches six and seven figures fast — with a big chunk of it going unused. That's not an argument to build a replacement for every tool. It's that there's now enough money inside the average stack to justify a question that was much harder to make the numbers work on a few years ago:

Which parts should we stop renting and start owning?

That's where I work. I audit the stack, find the workflows where custom software has a clear operational or financial return, build the tools, and hand them to the team.

Want to know what's worth replacing?

I can review your current software stack and operating workflows, then map what to keep, what to consolidate, what to automate, what could be built, and the likely return before we start. Book a call and we'll go through it.

Diego Casabe is an AI Enablement Partner and Forward Deployed Engineer. An ex-COO and founder of Kleta, he builds custom AI tools for founders and SMBs across the US, and cut his own company's software bill from $30K to under $250 a month. Built, not advised. (Zylo 2026 SaaS Management Index and vendor list prices checked September 2026.)