Commodity (Premium)

Regardless of your opinions on the efficacy of AI, the speed at which it’s advancing through the industry is both exciting and terrifying. But there’s a dark underbelly to this speed deployment of new functionality, which I see as chaotic, and a major downside to Big Tech firms like Microsoft that are throwing money at AI in what can only be described as unsustainable ways. What happens when AI becomes a commodity, and your business model is built on a model in which customers are asked to shoulder the extravagant costs?

We may find out sooner rather than later: The other side of the AI growth curve will be a trough of disappointment of epic proportions, and if the speed at which AI advances is any indication, the speed at which customers stop paying for it could likewise be historic. This could be a seismic shift that will undermine some of the industry’s biggest players, and right at a time in which they are relying on AI to drive the growth that investors demand. This isn’t tail wagging the dog stuff, like the irrational exuberance of the Azure-fueled cloud era Microsoft just crested, it’s even more ephemeral. There’s no there, there.

I’ve raised this concern several times over the past months or more–as I’m sure you’ve noticed, AI has quickly become an unavoidable topic, one that threatens to subsume everything else I’d rather be writing about–but I was fascinated to discover that Microsoft CEO Satya Nadella recently described the large language models (LLMs) that form the backbone of his company’s overreaching strategies, products, and investments as “a commodity.” As I see it, this was a passive-aggressive shot across the agents of chaos at OpenAI, a company Microsoft over-invested in and is overly reliant on, and a subtle indication to investors that maybe it’s time to cool their jets a bit.

“As AI becomes more capable and agentic, models themselves become more of a commodity, and all value gets created by how you steer, ground, and fine-tune these models with your business data and workflow — and how they compose with the UI layer of human to AI to human interaction,” Mr. Nadella wrote on LinkedIn on the day that Microsoft announced “wave two” of Microsoft 365 Copilot; less than two weeks later, Microsoft announced a similar “next phase” for the consumer Copilot as well.

To understand what’s he’s driving at there, consider where Microsoft is at right now. Having invested billions of dollars in OpenAI and seeing the sudden advances it made with ChatGPT 4 in late 2022, Nadella literally turned the battleship that is Microsoft on a dime and directed the entire company to immediately pivot on AI.

To the public, this new Microsoft arrived in February with the ill-chosen Bing AI chatbot announcement. But internally, the battle cry arrived months earlier, and employees were given a directive: You will embrace AI in whatever product you represent, or you will leave and someone more compliant will take your place. It was a stark example of how different Microsoft is internally under Nadella compared to its public persona. Or it was: Since embracing AI so aggressively, I feel that we’ve come to learn the true Nadella. And that he’s not the robotic, droning public presence we’re privy to. He is, instead, an assassin.

When Nadella took over as CEO of Microsoft, he oversaw a parade of product groups to determine whether they could exist in this new version of the company. In a scene reminiscent of the “Bobs” in the movie Office Space, each was told that they had to conform in two ways. They would need to be financially viable as a business, or provide a plan for getting there. And they would need to align with the cloud focus.

For some businesses, this was simple: Office 365, for example, was already the poster child for how a legacy software business could transform into a cloud services dynamo via Office 365 and now Microsoft 365. Others required a bit of selling: Phil Spencer infamously convinced Nadella that the unprofitable Xbox business could transform from a hardware-based business into one centered on cloud services–Game Pass–that would work anywhere, not just on Xbox consoles. And others were left in a bit of a lurch. Windows, for example, had long benefitted from volume licensing revenues from its commercial customers, so the plan was to transform that business by updating it like an online service. The result was Windows as a Service (WaaS), now called “continuous innovation,” and few customers would agree that the constant churn is a net positive.

AI triggered a second realignment for all of Microsoft, and it had to happen quickly. This is what led to the chaos of 2023, when Microsoft announced AI service after AI services, recalibrating brands and capabilities at a breathtaking pace and in public. But there was a big difference between the cloud push and the AI push: Where Microsoft’s cloud capabilities were all developed internally as a direct result of Ray Ozzie’s Internet Services Disruption strategy in the early 2000s, the AI push was almost 100 percent based on an outside party, OpenAI. And OpenAI cannot be trusted.

The chaos of 2023 ended in the biggest chaos of all: Co-founder and CEO Sam Altman was unexpectedly and suddenly fired by the OpenAI board of directors in November, kicking off an incredible four-day period in which he returned to the company, gave the horrified Microsoft a non-voting board seat–Nadella and company learned of Altman’s ouster just hours before it happened–and began a purge of those who had opposed him. In the span of one year, Microsoft’s vulnerability was exposed in a horrible and very publicly way. And it triggered a great rethinking of this partnership that culminated in Microsoft acqui-hiring Mustafa Suleyman and most of Inflection AI, not as a hedge against OpenAI, but really an organization that could replace OpenAI in time.

The thing is, Microsoft’s partnership with OpenAI, for all the money and Azure resources that Microsoft had poured into the company, was somewhat one-sided. OpenAI was, and is, free to partner with any company it chooses, and it of course chose Apple, infamously, this past year, as one of those other partners. OpenAI is also free to compete with Microsoft’s AI services, and it does, leading to awkward moments in which each is forced to crap on the other’s products in marketing its own, in what is perhaps the weirdest form of coopetition I’ve ever seen.

2024 has brought its own forms of chaos, but one thing is clear: Microsoft is now moving aggressively to sever its reliance on OpenAI. And, in doing so, it is being even more aggressive in competing directly with this company, leveraging its huge customer base, exhaustive product catalogs, and industry-leading infrastructure. OpenAI needs Microsoft as much as Microsoft needs OpenAI, but that dynamic is shifting as Microsoft pushes to alleviate that need. And this is all Nadella’s doing.

And in Nadella’s words, Nadella’s public words, it is the customer services that it creates and not the language models they require that is important. The models are the commodity, and it is the AI services and agents that matter. It is OpenAI that is the commodity. And Microsoft that matters.

As horrifying as the Altman drama was to Microsoft’s leadership in late 2023, one has to assume that this simple sentence from Nadella has triggered a similar horror with OpenAI’s leadership. In fact, it’s difficult not to believe that the recent machinations there, with its CTO and vice president of research unexpectedly leaving in late September and then the company completing a new investment round that values OpenAI at $157 billion, aren’t the direct result of this Nadella laser-guided bomb. He is, again, an assassin.

The problem for Microsoft, no matter what happens with or to OpenAI, is that Nadella’s words apply to it as well. As I wrote back in March, when everything is AI, nothing is AI, meaning that Microsoft can never have exclusive ownership of any AI features, and that these capabilities will–and are–spreading everywhere. Today, Microsoft charges exorbitant fees for AI–Microsoft 365 Copilot is an astonishing $30 per user per month over the base subscription fee–and it’s unsustainable. Microsoft is investing tens of billions of dollars per quarter investing in its AI infrastructure, $19 billion in the most recent quarter alone, and it’s unsustainable. The entire business, the reason Wall Street is still bullish on Microsoft, could accurately be described as unsustainable. This cannot continue forever.

And so Copilot “wave 2” was launched with a hope and expectation that customers will bite. But there’s little evidence that any are all that interested. A source told me this past week that Microsoft 365 Copilot uptake is low single digits percentagewise, and there’s little in the wave 2 announcements that rises above the level of the types of improvements Microsoft used to just provide to existing customers as part of their existing plans. Microsoft may still win this battle by becoming the infrastructure for other AI services. But I have serious doubts about its first-party services. Just as its customers do.

Put another way, it’s not the LLMs that are the commodity, it’s AI itself. And in a world in which Microsoft cannot keep these capabilities to itself, that means that Microsoft, too, is a commodity. So let the best AI win, I guess. But AI won’t drive any big market shifts, I bet. Instead, Microsoft, Google, Amazon, and Apple will simply retain whatever advantages each has over the others. The wild card, of course, is OpenAI. But the most likely outcome there is implosion, given this company’s volatility. And it will happen quickly, I bet. That’s the model it’s established for itself. And I can’t imagine many will mourn its passing.