How Microsoft Prepared Itself for AI Failure

Microsoft, to put it mildly, disappointed with its latest earnings report, didn't it? It wasn't the most pleasant sight, and the consequences will be hard to fix.

Microsoft announced a "strong" quarter with impressive growth, but as soon as the public delved into the numbers, the strong quarter turned into a panicked sell-off, leading to "the largest one-day loss in dollar terms in Microsoft’s history."

It quickly became clear that the market was primarily displeased with what was related to Microsoft's relationship with AI, particularly with OpenAI. Dig deeper, and you'll find that the concern extends beyond just one company's risky relationships with another.

This is a major crack in the foundation of AI. Not the first, not the last, but definitely the loudest.

Microsoft is not alone in this. Results are continuing to come in, and everyone is worried about the AI bubble — including OpenAI, Anthropic, Google, and other platforms. In any case, that's what I'm hearing.

So why did Microsoft's positioning in AI lead to a panic sell-off?

Microsoft's "unique" approach to innovation

A quick history lesson. Let me tell you guys the story of Bill "Old Reliable" Gates.

It's no secret that since the days of MS-DOS, Microsoft has had a reputation as a company that supposedly did less inventing and more, let's say, "improving existing technologies."

Apple has claims (Windows). So do WordStar and Visicalc (Office). Even Nintendo and Sony (Xbox).

But Active Server Pages were a hit!

Listen, I'm not accusing Microsoft of plagiarism or a lack of innovation. They are doing just fine. Just like VHS beat Beta or Netflix beat Blockbuster, Microsoft's strength has always been in turning innovative technologies into solutions and then dominating their unique market niche — a powerful intersection of business, consumers, and developers.

As the company grew, the history was as follows: Microsoft waited for someone else to innovate, and then imitated, eventually forming partnerships, and perhaps acquiring. They brought this new solution to their mass technology market, carefully not leaning too much towards consumers.

Of course, there were some well-documented mistakes — Internet Explorer 6, Windows Me, Windows Phone, and Zune, for example.

Now the question hangs in the air. Is CoPilot just a Zune that talks back to you?

Evaluate AI strategies for yourself

Speaking of different companies' approaches to AI and their strategies — it is important to understand the real capabilities and limitations of modern AI systems in practice. The gap between marketing promises and reality is often huge.

Microsoft plays the role of an actor playing the role of another actor

Honestly, I haven't been following Microsoft closely in the AI race because I believed they were just doing what they always do. Sitting on the sidelines, placing bets, waiting for their market/ecosystem solution to become clear so they can pounce on it.

Of course, they have CoPilot, and they are quite happy to integrate it into their consumer offerings (365), developer offerings (GitHub), and even their infrastructure offerings (Azure AI).

I mean, for a comparison of strategies, look no further than how Google Chrome and GSuite are capturing markets, while Microsoft is quite content to promote Edge and 365 through its own ecosystem. There was a time, not so long ago, when you needed a Mac. Now not so much. Microsoft is always there, patiently sitting on the sidelines and placing bets.

Well, one of those AI bets looks... complicated. In its earnings report, Microsoft acknowledged that over half of its unfulfilled performance obligations (RPO), a staggering $281 billion out of a total of $625 billion, are tied to OpenAI, and thus, related to how OpenAI will figure out how to stop burning money and start making it.

“RPO represents contracts that customers have signed, but Microsoft has not yet fulfilled. It is a measure of future revenue that is already booked. Microsoft's report showed that about $281 billion of this unfulfilled volume of obligations comes from one customer, a company that is still burning money and searching for a sustainable business model.”

Wow.

The bet that OpenAI will "figure it out"

Microsoft, like all of us, is waiting for OpenAI — and other platforms and providers — to figure it out.

But the bigger picture, the one that is worth many thousands of words generated by ChatGPT, is that the AI bubble continues to inflate. AI should have been at the equivalent of the “computer in your pocket” phase of the mobile-first revolution by now. It seems that AI may still be racing toward the “fortune favors the bold” phase of cryptocurrencies and NFTs.

If we've learned anything from the outcomes of the “chatbots and porn” AI phase, it's that sitting on the sidelines and innovating by “me too” won't work here.

What Microsoft's miss means for the future of AI

It's not for me to tell a large successful corporation what they should have done. But for the billions that Microsoft poured into OpenAI, they should have been directing OpenAI on what to do.

Microsoft is actually in an excellent position as the next wave of AI begins to become dependent on proprietary unstructured-to-structured data, and they didn't need a crystal ball to tell them that. They have corporate connections, consumer attachment, a layer of developers, and distribution. They should have owned the connective tissue between AI models and corporate workflows, consumer products, and development.

Microsoft should have been preparing itself for this next wave instead of playing around with a 3.3% CoPilot penetration at the MS Office level. Even though Google looks worse in this regard, they are once again pulling ahead, just like GSuite, just like Chrome.

In the 2020s, despite all its mistakes and craziness, Microsoft still owns this triad corporate market, which is part consumer, part business, part developer. If Microsoft goes down in history for anything other than Windows, it will be for creating and developing this market.

So if Microsoft doesn't feel the pulse of the AI client, and they don't feel it, then no one feels it, and no one actually feels it, and we are all flying blind regarding corporate demand for AI.

This is the crack, and that's why it has been so loud.