2025 is the Year of Competing AI Business Models (Premium)

Last year, I said I would not pay for AI. And I’m not the only one, so Microsoft and Google have both adjusted their business models accordingly. Both of the respective changes are improvements. But each is different from the other. And one is superior, meaning it’s the better value for customers.

If you didn’t read I Will Not Pay for AI (Premium), the short version is that Microsoft and Google were extorting customers by making AI features–Copilot-branded in Microsoft’s case and Gemini-branded in Google’s–exorbitantly expensive add-ons to existing paid subscriptions. And I was having none of it. I don’t mind paying for Microsoft 365 and Google Workspace–indeed, I pay for the commercial and consumer versions of each–but I have no interest at all in paying for features that should just be part of those subscriptions. That’s like paying for a word processor and then paying extra for spell checking. Nope. Not going to happen.

The issue here is two-fold. First, AI is or will be a part of virtually all the digital products and services we use, just like spelling checking is part of a word processor. Second, the added cost of these extra AI features was beyond unreasonable.

As just one example, a Microsoft 365 Family subscription gives full Office app usage on desktop and mobile across multiple devices, 1 TB of OneDrive storage, and other perks to six users for $120 per year. But a Copilot Pro subscription–the consumer version of Microsoft 365 Copilot–costs an additional $20 per user per month. If just one user on Microsoft 365 Family paid for Copilot Pro, the annual cost would be three times higher, or $360. If all six users went all-in, the total annual cost would jump from $120 to an ungodly $1560. That’s not just incommensurate to the added value, it’s insane.

Google’s pricing was similar. It has offered Gemini add-ons for its commercial Workspace customers and Google One plans that include Gemini Advanced, roughly following the same pricing scheme as Microsoft: An additional $20 per user per month for consumers and an additional $30 per user per month for businesses. Again, insane.

But in the wake of my AI editorial, which I published last August, it was clear that I wasn’t the only Microsoft and Google customer uninterested in helping these companies pay for their expensive AI infrastructures. And in the months since then, both have adjusted the business models for adding AI functionality to their respective productivity offerings. The new models are familiar and, in some ways, similar to each other. That said, they are also quite different from each other too.

Here’s the short version.

Microsoft will continue offering multiple Microsoft 365 subscription tiers for its commercial and consumer customers. It will also continue offering those expensive paid add-ons–Microsoft 365 Copilot in the commercial space and Microsoft Copilot Pro for consumers–at the same expensive per user/per month prices. But it will also raise prices on the base subscription tiers across the board and give those customers limited access to Copilot capabilities each month. Access to those features is determined by “AI credits” that are spent as you use AI features and then renewed at the start of each month.

The idea here is solid enough: If a customer keeps running into the limit and wants to keep using Copilot functionality, they can pay for the expensive add-on. Aside from the additional monthly cost, I like this plan quite a bit. What would be even better is a sort of pay-as-you-go system, by which you could buy additional AI credits when you run into the limit. This would be nice for users that only do so occasionally, a sort of middle ground between the base subscription and the expensive middle ground. (I do this now with Duolingo. My wife and I pay for the Super Duolingo subscription, which gives us whatever perks, but one to three times per year one of us feels the need to make an in-app purchase to get some gems, the in-app currency, for whatever reason.)

But whatever. Microsoft’s new business model is “better” than the original overall, assuming the price hikes on the base Microsoft 365 subscriptions aren’t too high. They’re certainly expensive in the few markets in which this change has been rolled out so far, but we’ll have to see what it looks like when they come to the U.S., Western Europe, and other markets.

Of course, this is Microsoft, so nothing is every easy. Today, Microsoft announced it’s bringing AI-based agents to its (free) Copilot chatbot, but only for paid Microsoft 365 Copilot users. (Good luck parsing that one.) Oddly, Microsoft is using the pay-as-you-go model that I outlined above for this service–odd because I wrote most of this article before I even knew about this addition–but only in certain circumstances. If you don’t pay for Microsoft 365 Copilot, you can (have to) pay as you go for agents grounded in your organization’s data, and for agents that act independently on your behalf. But even if you do pay for Microsoft 365 Copilot, you have to pay for the latter type of agent. Somehow, that $30 per user per month doesn’t cover this service. For the cynical, this will feel like yet another SKU of Microsoft 365.

There’s so much more to the confusion here that I almost hesitate to mention things like Copilot+ PCs and Microsoft’s efforts to bring as much AI as possible down to NPU-powered PCs where they can’t harm its bottom line. That these AI features are mostly lame and/or pointless only adds the problem, as is the fact that Copilot+ PCs do to Windows 11 what Copilot does to Microsoft 365 by adding what is essentially yet another SKU with its own unique features, confusing everyone. But let’s move on. Life is too short.

Today, Google, which previously mirrored Microsoft’s Microsoft 365/Copilot offerings with its own Workspace/Gemini Advanced offerings, announced a dramatic change to its business model. And this is, I think, the superior approach, the one that will deliver more value to customers.

Unlike Microsoft, Google is killing its paid AI add-ons (though, for now, it appears to only be doing so with its business offerings). That is, Google previously sold various expensive Gemini add-ons at additional per user/per month fees on top of its normal Workspace subscriptions, but now it no longer does so. Instead, it is adding all those Gemini Advanced features “for free” to those with Workspace subscriptions. But in making this, it’s also raising prices.

But here, too, Google is veering wildly away from the Microsoft strategy. Where Microsoft’s Microsoft 365 price hikes are fairly dramatic and only come with a limited set of monthly AI credits, Google’s price hikes are comparatively reasonable and they come with no limits. That is, Google is finally doing what I wanted all along, what I communicated last August. It is providing additional value on top of existing subscriptions at commensurate price points.

I noted above that I pay for Microsoft and Google commercial and consumer productivity services. And that adding Copilot Pro to my Microsoft 365 Family subscription was incredibly expensive. But with Google, a Workspace Business Standard account was $12 per user per month (paid annually), while a Gemini Advanced add-on would have cost an additional $30 per user per month. With this change, the price of that subscriptions is going up to $14 per month, a 17 percent increase. That’s reasonable. Commensurate.

What Google isn’t doing, of course, is adding any kind of pay-as-you-go option. But that makes sense because there’s no expensive paid add-on anymore. There’s no need for a middle ground, as there is, arguably, with Microsoft’s new business model.

Again, both changes are a net positive, from what I can tell. It’s just that Google’s is better for customers. Microsoft is still trying to subsidize the tens of billions of dollars it spends each month building out its AI infrastructure, and it’s not clear that customers are on board, especially with Microsoft passing the costs along to customers. This is, I think, a hangover from the churn it’s introduced over the past two years.

On that note, though I’m growing weary of the lingering AI doubt I still see out there in the world, I do understand being change-averse. And AI is a change agent like nothing before it, moving at a blistering speed to upend personal and corporate computing. Looking just at Microsoft, which jumpstarted this new AI era with its initial Bing chat announcement in early 2023, it’s like AI is a cancer that’s inflicted the company’s collective ability to think clearly. Microsoft’s primary customer base is commercial and it values conservative, slow-moving consistency above all else. It is, in other words, as change-averse as is possible. And yet, here’s Microsoft, continually lobbing AI grenades in the metaphorical room, and then not waiting for the dust and smoke to settle before lobbing another one. The churn on AI is astonishing and, frankly, a bit alarming.

I wrote about the impact of this madness and epic lack of strategy on Windows the other day. But Microsoft’s “damn the torpedoes” approach to AI is much broader than the one product I happen to care about the most. The impact on Microsoft 365–what I still think of as Office–is and will continue to be even broader. The impact on Microsoft’s untold number of cloud offerings, even more so. As is the financial impact of building out the infrastructure required to host these services. $80 billion or more in a single fiscal year, the company just admitted.

For Microsoft, this AI era is the first time that its “bet the company” mantra is more real than marketing. AI is either an existential threat to Microsoft or its biggest-ever opportunity for growth. Satya Nadella, his senior leadership team, and the company’s board of directors are behaving accordingly. Every day, it seems, there’s something new. And two years in, most of what’s new is really just a recasting or reimagining of something that was new last month, last year, or even longer ago. Microsoft is investing, partnering, acquiring, and reorganizing on what feels like a rolling basis, trying to figure out a way to make AI not just make sense, but be that growth engine everyone seems to believe it can be. That $80 billion is but the tip of the iceberg. And the rationale is that if Microsoft doesn’t there first, doesn’t “win,” then a competitor will. And that will be the end of Microsoft.

But what does it mean to get there?

This is an over-simplification, but for AI to make sense as a business, it needs to meet the needs of customers, but it also needs to be profitable. Not today, and not next year, not necessarily. But at some point. Microsoft has the financial resources, and ongoing revenues from established products and services, to fund this push for years to come. Many are often confused how seemingly non-profitable startups and online businesses can remain in business, but Microsoft is something else entirely. It is an economic superpower. If any company can make this happen, it’s Microsoft. Only Amazon, Apple, and Google rival it in any meaningful way. And I think most would agree that there’s room in the market for all four of them.

Making sense of AI as a business is one thing. But doing so in real time while it’s on its chaotic rocket trajectory is likely impossible. And yet, Microsoft is trying. In 2023, I (sort of) praised the company for quickly rebranding its AI efforts, a relatively simple act that would have taken years in a different era. In 2024, we figured Microsoft would simply implement on all that it had announced the previous year. But there was more churn, much more, and 2024, if anything, was even more chaotic than the year before. This is a classic “throw spaghetti at the wall and see if it sticks” scenario, but writ large. To the outside world, it feels like the company has no idea what it’s doing. But that’s how it looks from the inside, too. Microsoft literally has no idea what it’s doing. But it’s doing it quickly. The result is chaos.

Now, Microsoft has finally embraced what I feel is a proven business model, by which it offers the same capabilities across the stack in both free and paid versions. It’s not perfect, but it’s better. This is comparable in some ways to what we see in subscription-based online services like Netflix and Spotify. They often (not always) come in free, ad-supported and paid tiers. There’s a lot of variety there, different ways to mix it up, and in some cases (newspapers, Peacock, etc.), there are even paid subscriptions with ads. But let’s just keep this simple. You pay or you don’t pay. If you don’t pay, you get ads.

With Microsoft 365 now, you can pay extra for Copilot or not. If you don’t pay, you get AI credits and limited use each month. If you do pay, you get unlimited usage … except for those new AI agent capabilities. Microsoft is always going to be Microsoft. Nothing is ever simple.

With Google Workspace now, you just get Gemini Advanced, essentially, “for free.” Yes, the prices are going up, but they’re going up even more at Microsoft 365. Google, inexplicably, is removing the complexity. In doing so, it’s making life simpler for customers. And, I am arguing, better.

What a world.