Microsoft Copilot Pricing After the Work-First Merge

On September 25, 2026, Microsoft closed a chapter it had been writing for two years. The company folded the consumer version of Copilot into a single work-first assistant, ending its attempt to win the personal AI chatbot race outright. The rebuild was the output of a six-month engineering effort, and the plain reading is blunt: Microsoft is done pretending the free and home tiers are the product.

For heavy AI users, this is not a branding story. It is a copilot pricing story. The subscription you pay for, the seat you provision, and the meter that runs behind your team’s daily work are all being renegotiated under a new product identity. Here is what changed, what the tiers cost now, and where the real spend hides once a chatbot becomes a work tool.

The merge, in one paragraph

Microsoft had been running two Copilots: a consumer assistant living in the free and home tiers, and an enterprise assistant bolted onto Microsoft 365. Those two products converged on September 25 into one assistant oriented around work. The consumer chat surface was absorbed, not killed, and it now arrives wrapped in the identity of the M365 product.

The numbers behind the decision explain the direction. Microsoft reported more than 30 million paid Copilot subscriptions as of the end of June, and roughly 90 million paying Microsoft 365 users. The work side dwarfs the consumer side. When a company has 90 million paying seats and a chat product that is not monetizing like one, the rational move is to point every surface at the paid seat.

What copilot pricing looks like now

Copilot pricing was already a maze before September, and the merge does not simplify it. There are four price surfaces a buyer actually touches, and they are not substitutes for one another.

The consumer and pro tiers. A free tier still exists with limited access. Copilot Pro has been the paid individual tier in the $20 per month range, and it is the tier most affected by the merge, because it was the home of the standalone personal assistant. If you bought Pro for personal chat, you are now paying for a work-oriented product that has absorbed that surface.

Microsoft 365 Copilot. The per-seat enterprise add-on has sat around $30 per user per month on top of an existing M365 license. With roughly 90 million paying M365 users as the addressable base, this is the tier Microsoft actually wants you on.

Copilot Studio and the consumption meters. This is where copilot enterprise spend stops being predictable. Agents built in Studio bill on message and credit consumption, not on seats. A seat is a floor. The meter is the ceiling, and the ceiling is not fixed.

GitHub Copilot. The developer product has its own pricing and its own usage-based billing model for organizations. Developer seats are frequently the largest single line in a company’s AI budget, and they are tracked separately from the M365 add-on.

The practical consequence: “what does Copilot cost” has no single answer, because the four tiers above are sold to different budget owners and reconciled by nobody.

Where the real spend hides

The seat price is the number everyone plans around, and it is the least dangerous number in the stack. Three leaks matter more.

Consumption meters inside Studio. When your team ships an agent into production, message volume scales with usage, not with headcount. A pilot with five users costs almost nothing. That same agent after it becomes the default intake path for a department costs something entirely different, and the bill arrives after the fact. Consumption pricing is honest and it is also unbudgetable if nobody owns the meter.

Truncated context that forces retries. Heavy Copilot users in enterprise settings report that conversation history is aggressively truncated, so the assistant loses recent instructions and the user re-explains. Every re-explanation is a paid round trip. This is a soft version of the token cost problem: you pay twice for the same work because the first attempt did not carry enough state.

Overlapping product names that hide duplicate spend. Microsoft has shipped many products named Copilot, each with its own billing path. The merge reduces the surface area but does not eliminate the overlap, because Studio, GitHub Copilot and the M365 add-on remain separately metered. Duplicate coverage is the easiest waste to miss and the easiest to find once you look at the invoice as a whole rather than per product.

What to do about it

The merge is a good forcing function. Treat the next renewal as an audit.

Build one AI ledger across all four tiers. M365 add-on seats, Copilot Pro seats, Studio consumption and GitHub Copilot billing belong on one sheet. Any budget owner looking at only their own line will miss the overlap.

Put a ceiling on every consumption meter. Set hard spend caps and alerts on Studio and on GitHub Copilot usage-based plans before the pilot graduates, not after the first surprise invoice. This is the single highest-value action available to a FinOps team right now.

Reclassify Pro seats honestly. If a Pro seat was bought for personal chat productivity, test whether it is still earning its place now that the surface is work-first. Some seats will survive the test. Some will not, and that is fine.

Compare against the alternatives before you renew. Claude, ChatGPT, Gemini and Copilot all now sell a work assistant with a consumption layer behind it, and the price per effective task varies more than the sticker prices suggest. Do not renew a Copilot estate by inertia just because the merge made the product feel unified. The unification is on Microsoft’s side of the table, not yours.

Instrument the retry rate. Track how often users re-ask a question because the assistant lost context. That number is a direct proxy for wasted spend, and it is the argument you will need when you ask for a cap.

The cross-provider reality check

Microsoft is not making this move in a vacuum. Anthropic’s Opus 5.5 release arrived with lower prices, OpenAI has been shipping API spend controls, and Google’s Gemini pricing has been under pressure. Every major lab is converging on the same architecture: a cheap seat to get you in, a consumption meter to capture the value you create, and enterprise controls to make the meter acceptable to a CFO.

That convergence is the actual story behind the Copilot merge. A work-first assistant is worth more per seat than a personal chatbot, and it consumes more. Microsoft did not abandon the consumer race because consumer AI is unimportant. It walked away because the work seat is where the meter lives, and 90 million paying M365 users are a far better place to put a meter than a free chat app.

For a heavy user, the lesson is structural. The seat price you negotiate is the least interesting number in the contract. The meter behind it is where your budget will actually be decided, and it is the part nobody puts on the pricing page.

The merge itself changes very little about your day. The renewal conversation it triggers should change a great deal.