Put Up or Shut Up (Premium)

We’ve long known that 2024 was the year in which Microsoft would need to deliver on its AI promises. But it’s only mid-January, for crying out loud, and no one expected the software giant to suddenly release a comprehensive set of paid AI services for both consumers and commercial customers just two weeks into the New Year. This is both exciting and scary. And not just for Microsoft’s customers.

There are so many things to discuss with Copilot Pro and Copilot for Microsoft 365, but the core issue, I think, is tied to that old classic, cost and value.

So let’s step back a bit. We all know that AI is both resource-intensive and prohibitively expensive, and that this reality is why only a handful of the world’s most powerful Big Tech companies have the capacity and desire to even do so. And we’ll all remember 2023 as the year when Microsoft got its groove back, moving decisively and aggressively to establish itself as both a leader and a first-mover in this suddenly emergent new market.

But as 2023, wound down, I was impressed by other less obvious signs that today’s Microsoft was different from its predecessors in ways that I think are important. The Microsoft of the past would have allowed its initial branding mistakes, in particular around its use of Bing for its first go-to-market AI offerings in 2023, to stand, not just for months, but for years. But this Microsoft can be a splendid thoroughbred when it needs to be, and it rebranded and recalibrated its offering several times since then. And among the best changes is that Bing today has been largely exorcised from the discussion. Yes, you can access what’s now called Microsoft Copilot from Bing. But Bing is not where you go to access these capabilities, and certainly not exclusively. The normal order of things has been restored, and we can all continue to ignore Bing just as we’ve done for the past many years.

Branding is important, and let’s never forget that Microsoft’s past is littered with many defeats thanks to brands both good (Skype) and bad (Zune). But with its AI branding issues behind it—Copilot is curiously perfect from a branding perspective—Microsoft still ended 2023 in a precarious state: This AI explosion has cost the company roughly $10 billion to $15 billion per quarter so far, but those costs are only going to go up as it continues building out the required infrastructure. And while it has successfully kept the lights on using the profits it makes from other parts of the company, AI has to pay for itself before it can drive its own explosive growth. And it’s not clear at all whether Microsoft is delivering enough value to justify the expense to customers.

Part of the problem, of course, is that Microsoft’s first productivity-focused paid AI service, Copilot for Microsoft 365, limped into its November 1 launch date with the double whammy of expensive pricing ($30 extra per month per user) and restrictive licensing terms—at least 300 users, only for the most expensive Microsoft 365 commercial tiers—that limited its appeal. (Microsoft’s first true paid AI service was Copilot for GitHub, which targets a niche audience.) This led to fears that perhaps there wasn’t any there there, if you will. That Microsoft was purposefully limiting the reach of this offering because its infrastructure costs were too expensive and/or because the offering itself wasn’t all that compelling. If Microsoft releases Copilot for Microsoft 365 but no customers pay for it, did Microsoft really release Copilot for Microsoft 365?

And so we celebrated the New Year wondering how Microsoft would tackle these problems. We know that it is working to lower costs in its cloud datacenters through a variety of means that include both Small Language Models (SLMs), hybrid AI with NPUs on clients, and custom, in-house silicon. And we knew that it would expand the availability of Copilot for Microsoft 365, and provide a consumer version, at some point. But the timing was unclear. And I suspect most people reading this didn’t expect to see any big announcements in the first quarter of 2024, let alone January. There just seemed to be too many obstacles.

Long story short, the January 15 announcement came as a surprise. And there’s no version of this story where we can belatedly rationalize this. Microsoft’s AI infrastructure didn’t suddenly get much bigger and less expensive to run. Its AI capabilities in Copilot in elsewhere didn’t suddenly make a massive leap forward in scale or scope. No, this was where Microsoft knew it would be months ago. And it is simply continuing the torrid pace it established in 2023.

And this might be a mistake.

The problems with Copilot Pro, which is correctly seen as the consumer version of Copilot for Microsoft 365, are many. It’s not clear whether the capabilities customers receive now—and the additional capabilities they will receive when the Copilot GPT Builder goes live—justify the cost. It’s also not clear whether there is an appreciable audience of people—Microsoft defines this audience as “power users like creators, researchers, programmers, and others”—willing to pay an additional $20 per month to access these capabilities. In fact, the audience of people who don’t even understand what these capabilities are, and would likely never pay regardless, is obviously much, much bigger. And there’s a good argument to be made that the free version of Copilot is too compelling to warrant ever paying for Copilot Pro.

Maybe that needs to change. But Microsoft is also treating AI like another tier of Microsoft 365. And on the consumer side, this doesn’t make sense because the costs are so out of whack, and because Microsoft has to date simply delivered these types of updates to Word, Excel, and the other Office apps for free as part of the existing subscriptions. You can go back in time a few years and look at any given month’s updates for Word, Excel, and so on. That’s what this feels like, but with a Neiman Marcus price tag attached.

That is, if you’re a consumer paying $69 per year or a family paying $100 per year for Microsoft 365, a more premium tier with new AI features might reasonably add, what, 20 percent to the annual cost? That would raise those subscription costs to $83 and $120 per year, respectively. But $20 per month is $240 more in one year, and that’s just for one person. The cost isn’t just incommensurate with its value, it’s incommensurate with the base price of the underlying (and required) subscription. Very few customers would ever go from $89 per year to $309 per year. Not in a world in which we’re bitching about the monthly cost of Netflix and Spotify.

You do get some interesting features with Copilot Pro, and across many apps and experiences, but most people aren’t using all of Office as it is. And so the list of advantages gets whittled down when you consider any individual’s needs. For example, I used Word every day for decades, but I rarely ever opened Excel, PowerPoint, or other Office apps that now have new AI features thanks to Copilot Pro. And so my personal value assessment would involve examining the features that would impact me.

It’s a short list. A very short list.

Copilot in Word adds a Draft feature that can help you write a first draft or improve an existing work. You can (more easily) transform text into tables, a feature I would never use. There’s a chat interface for learning more about certain words or terms, a feature people use Google Search for today. And like all good AI of this era, it can summarize documents, a feature that’s more useful for documents you did not write. That’s it. That’s all it does right now. For me, and, I suspect, for most, that doesn’t warrant spending $20, let alone $20 every single month.

Looking at this logically, few people will pay for Copilot Pro now because they can’t justify the value, don’t understand the value, and certainly don’t understand all the crazy new terminology (LLMs, GPTs, prompts, etc.) that Copilot and AI introduce. This puts Microsoft in an awkward position in which it must improve this offering to justify the expense, something that will, in turn, be expensive to it. And there is the fear that a poorly received launch could poison this effort and this brand, that if it has launched too soon, too aggressively, it may never recover. What if Copilot enters a death spiral?

The next few months are going to be very interesting. Microsoft can’t afford to screw this up, but it’s too late to roll back now: Copilot Pro is live, and there are people standing by at Microsoft waiting to take your call. I just wonder whether anyone is going to do so. Or whether this money grab will be the final straw for a customer base that is already tired of paying what they pay today and is not at all interested in what feels like an exponential increase in that cost.

Hang on tight, everyone. We’re about to find out.