Enterprise software budgets aren’t growing fast enough to cover every new AI tool. So something else is getting cut. According to new data reported by The Information, Anthropic, OpenAI and a group of younger AI companies are now taking a real share of the money companies spend on software. This isn’t new spending sitting on top of the old stack anymore. It’s coming out of it.
For two years, the story was that AI budgets were “experimental” money, kept apart from core IT spending. The Information’s report points to that line fading. When the model makers show up in the same budget as your CRM, your analytics tools and your help desk software, they’re competing with those vendors directly.
📊 What the Data Signals
The headline finding is simple. Model makers and AI-native startups are winning dollars that used to go to established software vendors. A few things make this shift stand out:
- The labs are selling straight to companies. Anthropic and OpenAI no longer just supply the models inside other vendors’ products. They sell enterprise seats, APIs and coding agents directly to the teams that write the checks.
- Upstarts are replacing tools, not just adding features. AI-native companies are pitching “do the job” products instead of “help you do the job” dashboards.
- Budgets are mostly flat. When CFOs keep total software spend steady, every dollar that goes to an AI lab is a dollar some other vendor loses.
That last point matters most. You don’t get a zero-sum fight until the budget stops growing, and that’s where many IT departments are now.
⚔️ Two Ways to Read It
The disruption view: Per-seat SaaS pricing is exposed. If an AI agent handles work that used to need ten licensed users, the buyer needs fewer seats. Incumbents charging per person are selling the exact thing AI is designed to reduce.
The incumbent view: Big software companies still own the data, the workflows and the procurement relationships. Salesforce, Microsoft, ServiceNow and others are putting agents into products customers already pay for, often bundled or discounted. Their bet is that “good enough AI inside the tool I already have” beats “great AI I have to integrate myself.”
Both views can be true at once. The likely result is a split market. The labs take horizontal work like coding, writing, research and general knowledge tasks. Incumbents defend the areas where proprietary data and compliance lock customers in.
🔍 Why It Matters Now
The timing isn’t random. Coding agents became one of the first AI categories where companies could easily show a return, and engineering budgets are large. Once a company sees measurable productivity gains in one department, it starts asking which other subscriptions it can drop.
There’s also a competitive angle for the labs. Selling directly to enterprises gives Anthropic and OpenAI better margins and a closer relationship with customers than earning API fees through someone else’s app. It also makes them competitors to many of their own biggest API customers. Expect that tension to sharpen as the labs push further into finished applications.
✅ What Businesses Should Do
If you manage a software budget, or you sell software, a few moves make sense:
- Audit seat-based contracts before renewal. Find tools where AI has already reduced active usage. Those are your easiest savings.
- Track AI spend as one line item. Many companies spread it across teams and credit cards. You can’t negotiate what you can’t see.
- Avoid single-vendor lock-in. Model quality and pricing change every few months. Build so you can swap providers.
- If you’re a SaaS vendor, rethink pricing. Outcome-based or usage-based models hold up better than per-seat pricing when your customer’s headcount is shrinking.
- Watch your API supplier. If a lab could ship your product as a feature, plan your moat now: proprietary data, deep workflow integration or vertical expertise.
🔭 What Comes Next
What stands out here is the direction of travel. A year ago, the question was whether companies would pay for AI at all. Now the question is who loses budget when they do. Watch next quarter’s earnings calls for traditional SaaS vendors. Any mention of “seat compression” or slower net revenue retention will tell you how fast this shift is really moving.
The full breakdown of the spending data is available in The Information’s original report.