Microsoft's AI Tax: Copilot Price Hikes and the Forced-AI Business Model
Microsoft raised M365 prices up to 43% to cover AI investments. One day later, Zuckerberg admitted agents are behind schedule. The forced-AI business model is here — and open-source is the exit.
The email landed in inboxes across the world in late June. Microsoft, in its characteristically gentle way, wanted to let you know that your bill was going up. Not because you asked for more features. Not because you complained that Word lacked something. But because Microsoft spent billions on AI, and now you get to help pay for it.
On July 1, 2026, Microsoft's new "packaging and pricing update" — corporate speak for "your renewal is getting more expensive" — went live across Business, Enterprise, Frontline, and Government suites. Some SKUs jumped as much as 43%. And the headline feature being bundled in to justify it? Copilot. The AI assistant most of their customers didn't ask for and many actively tried to avoid.
This is the story of how AI stopped being a feature and became a tax.
The Numbers (They're Bad)
Let's look at what actually changed, because the percentages tell a story that press releases won't.
Microsoft 365 Business Basic went from $6 to $7 per user per month — a 16% bump. Business Standard climbed from $12.50 to $14. Office 365 E3 rose from $23 to $26. Microsoft 365 E3 went from $36 to $39. E5 increased from $57 to $60.
Those are annoying but survivable. Then you get to the frontline worker plans.
M365 F1 jumped from $2.25 to $3 per user — a 33% increase. On the lowest-paid workers in the organization. F3 went from $8 to $10, up 25%. And if you remove Teams from F1? The price climbs 43%. That's not a price adjustment. That's a shakedown.
Standalone add-ons aren't spared either. Windows Enterprise per-device licensing jumped 31%, from $5.85 to $7.63. Microsoft 365 Apps per device rose 17%, from $36 to $42. Entra Plan 1 and EMS E3 both climbed by double digits.
For a company with 500 employees on Business Standard, this "update" adds $9,000 a year to the bill. For 5,000 employees on E5, it's $180,000. And what did they get for it? Let's look.
What You're Actually Paying For
Microsoft's pitch is that these aren't naked price hikes — they're bundling new stuff in. And to be fair, some of it is legitimately useful if you're in the Microsoft ecosystem. Office 365 E3 and M365 E3 are getting Microsoft Defender for Office 365 Plan 1 folded in. Business Basic and Standard get URL time-of-click protection (links scanned when clicked, not just when the email arrives). E5 tenants get Intune Advanced Analytics, Enterprise Application Management, and Microsoft Security Copilot.
That last one is the giveaway. Security Copilot — which Microsoft had been selling separately — is now included in E5 at 400 Security Compute Units per month for every 1,000 licenses, capped at 10,000 SCUs. And every affected suite gets "Copilot Chat enhancements" including inbox and calendar awareness and access to Word, Excel, and PowerPoint agents.
So the deal is: you're paying more, and in exchange, you get AI features baked into tools you were already using. Whether you wanted AI in your spreadsheet is not a question Microsoft is asking. It's already there. Your wallet is just catching up to the decision that was made for you.
The Timing Is Suspicious
Here's what makes this taste particularly bitter. On July 2 — one day after the pricing went live — Reuters published an interview with Mark Zuckerberg where he admitted that AI agent development is "going slower than expected." The Meta CEO, who has bet roughly his entire company on AI agents, told the world that the technology isn't where he thought it would be. Not a year ago. Not six months ago. Right now.
This came from the person with arguably the most resources on the planet devoted to this exact problem. If Zuckerberg is saying agents are behind schedule, you can be fairly confident that the AI being bundled into your $14/month Office subscription isn't ready to justify a 12% price increase either.
Then there's the Forbes headline from the same week: "AI Costs More Than the People It Replaced." Companies that replaced human workers with AI systems are discovering that inference costs, API fees, and the infrastructure to keep models running can exceed what they were paying in salaries. The AI savings story that dominated 2024 and 2025 is colliding with the reality of compute economics.
So the picture coming into focus is: AI is expensive to run, it's not delivering the autonomous results that were promised, and the companies that invested heavily in it need revenue to cover the gap. Microsoft found 430 million ways to do that. They're called M365 subscribers.
The Forced-AI Problem
There's something structurally weird about what's happening here. When Salesforce adds a feature, you can usually ignore it. When Slack ships something you don't care about, it sits in a menu somewhere. But AI is different because it actively inserts itself into your workflow.
Copilot shows up in your Word document with suggestions. It summarizes your emails whether you asked it to or not. It generates meeting notes and puts them somewhere you'll find them. Teams now has AI features that generated enough backlash that Microsoft — two days after the price hike — announced they'd let you turn some of them off.
Think about that sequence. July 1: prices go up, partly because of AI investment. July 3: Microsoft caves on Teams AI backlash and agrees to give users an off switch. They raised your bill for a feature people hated enough that they had to make it optional.
This is the forced-AI business model in its purest form. Build AI. Bundle it into everything. Raise prices to cover the cost. Make opting out technically possible but practically friction-filled. Count on inertia and enterprise procurement cycles to keep the revenue flowing.
It works. It'll probably keep working for a while. Microsoft is a $2.9 trillion company with lock-in that runs decades deep. Most enterprises aren't going to rip out Active Directory because their per-user cost went up $3. But the resentment is compounding, and resentment has a way of showing up in renewal negotiations 18 months later.
Meanwhile, Outside the Walled Garden
While Microsoft taxes its captive audience and Zuckerberg admits the agent revolution is behind schedule, something quieter is happening in the background. Open-source models are getting better faster than anyone predicted.
Meituan's LongCat-2.0 dropped in late June — 1.6 trillion parameters, trained and deployed entirely on Chinese domestic hardware with zero Nvidia GPUs. It ranked second globally in Claude Code usage behind Claude Opus 4.8 before it was even officially released. GLM-5.2, DeepSeek V4, Qwen — the open-source frontier keeps pushing forward without the recurring SaaS overhead.
The economics are fundamentally different. A closed model from a big lab comes with per-token pricing, rate limits, usage tracking, and the ever-present threat of a price increase when the provider decides their margins need help. An open-source model you run yourself costs electricity and hardware depreciation. That's it. No seat licenses. No SKU changes. No "packaging updates."
This is why the desktop AI assistant space is getting interesting. Tools that run open-source models locally — browsing your files, executing commands, automating workflows — offer something the cloud providers structurally can't: ownership. You're not renting intelligence by the token. You're running it on your machine, on your terms, with no surprise invoice.
Microsoft's price hike is a reminder of the trade-off. The convenience of an all-in-one cloud suite comes with a recurring cost that only goes in one direction. The open-source path requires more setup but gives you a bill that's predictable and a model that can't be repriced out from under you.
So What Do You Actually Do?
If you're an IT admin reading this while staring at a renewal quote that's 13% higher than last year, here's the honest landscape.
First, if you renewed before July 1, you keep your old price until your next renewal. You also get all the new features regardless. So there's a window.
Second, Microsoft 365 Personal and Education pricing isn't changing. If you're an individual consumer or a school, you're safe for now. The increases target businesses, enterprises, frontline workers, and government — the segments with the least ability to switch quickly.
Third, start auditing what you actually use. A lot of companies are paying for E5 when E3 covers 90% of their needs. The features being added — Defender, Intune, Security Copilot — are genuinely valuable if you're using them. They're expensive paperweights if you're not.
And finally, pay attention to the open-source AI ecosystem. Not because you need to rip out Microsoft tomorrow. But because the gap between "AI as a forced subscription" and "AI as a tool you control" is narrowing every month. The companies that understand both models will make better procurement decisions than the ones that just auto-renew.
Microsoft's AI tax is here. It's not the last one you'll see — Google, Salesforce, and every other enterprise platform are watching this rollout closely. The question isn't whether AI will be bundled into your software bill. It already is. The question is whether you have alternatives ready when the next increase lands.
If you're tired of renting AI by the seat, CopperRiver runs open-source models locally on your Mac — browsing, coding, automating, no per-token surprises. Plans start at $9/mo. The price won't jump 43%.