Nadella's Flip-Flop
There are perhaps few things in the current AI revolution that keeps my opinions grounded than watching Microsoft’s evolving position in the post-ChatGPT world. It can be tempting to imagine investors just frenetically changing their labels to Microsoft as “AI winner” or “AI loser” every few months, but the reality is Microsoft’s CEO himself has quite conspicuously evolved the way he wants to position Microsoft in the AI landscape. Ben Thompson recently alluded to Nadella’s flip-flop in a recent piece. It is quite instructive to actually go through Nadella’s own words from October 2023 earnings call:
“…the approach we have taken is a full-stack approach all the way from whether it's ChatGPT or Bing chat or all our Copilots all share the same model. So in some sense, one of the things that we do have is very, very high leverage of the one model that we used, which we trained, and then the one model that we are doing inferencing at scale. And that advantage sort of trickles down all the way to both utilization internally, utilization of third parties.”
As you can tell, the “one model” was the sun, and Microsoft’s products were the planets orbiting it. It may seem strange today, but Microsoft saw the tight coupling of the underlying foundation model with the user-facing application as a massive competitive advantage. They had exclusive commercial access to the best model in the world at the time, so a paradigm where the model was very much entangled with the product suited them perfectly. However, when Sam Altman got temporarily fired by OpenAI’s board, the ensuing drama laid bare all the limitations of this approach of relying on a third-party company’s model. So, Nadella started softening his tone about the benefits of having exclusive access to OpenAI’s models.
From celebrating a “full-stack approach” sharing the “same model” in 2023, you can clearly see how much Nadella’s position has changed in his own words from Morgan Stanley TMT Conference early this month:
“I think at this point, it’s fairly clear industry structure-wise when it comes to the narrow way we talk about frontier models, like the American ones are closed, a lot of the Chinese are open. That, I think, is going to be multi-model, right?
…if you’re building any product, whether it’s for coding agents or for knowledge work or wealth management or whatever, you want to access multiple models, right? That’s going to be the case. And that means there’s lots of very careful design that you want to do, which is you want to have the harness not get coupled with the model layer. So people are getting pretty sophisticated in making sure that the harness layer is decoupled.
The other is the context layer also should be decoupled, right? You don’t want effectively the one model to, in fact, vertically integrate into these 2. And that will be the game that will be played. But I’m pretty clear industry structure-wise, where we will go.”
Why is Nadella suddenly preaching the gospel of modularity and decoupling of different layers? Because Microsoft controls the harness (VS Code, GitHub, Azure AI Foundry) and the context (Microsoft 365, enterprise data). What they do not own outright is a frontier model. If the industry structurally integrates the model, the harness, and the context, Microsoft’s right to lion’s share of the profit pool can be legitimately questioned. Nadella needs the model layer to become a commoditized, interchangeable plug-in so that Azure’s orchestrators and development tools can capture the profit.
To be fair, Nadella’s argument makes sense even from enterprise customers’ point of view. But what makes rational sense to preserve their own interests and how AI products actually work to provide customers the best experience can diverge over time. “Copilot Cowork” (which is powered by only Anthropic’s models) is perhaps an early indication of this potentially persistent tension. Ben Thompson eloquently explained this dichotomy (emphasis mine):
“what made Opus 4.5 compelling was not the model release itself, but changes to the Claude Code harness that made it suddenly dramatically more useful. What this means is that model performance isn’t the only thing that matters: the integration between model and harness is where true agent differentiation is found.
This is a very big deal when it comes to figuring out the future structure of the AI industry and where profits will flow, because profits flow away from modular parts of the value chain — which are commoditized — and flow towards integrated parts of the value chain, which are differentiated. Apple is of course the ultimate example of this: its hardware is not commoditized because it is integrated with their software, which is why Apple can charge sustainably higher prices and capture nearly the entirety of the PC and smartphone sector profits.
It follows, then, that if agents require integration between model and harness, that the companies building that integration — specifically Anthropic and OpenAI (Gemini is a strong model, but Google hasn’t yet shipped a compelling harness) — are actually poised to be significantly more profitable than it might have seemed as recently as late last year. And, by the same token, companies who were betting on model commoditization may struggle to deliver competitive products.
For what it’s worth, Sam Altman in a recent TBPN interview indicated OpenAI will also launch their own version of “Claude Cowork”:
OpenAI will obviously" have a version of Codex that "can do other knowledge-work tasks, and control your computer."
"Of course we should have an ability to kick off new tasks from mobile, and we'll do that."
"Really what you want is your single AI that's working for you on a unified backend. Access to all your data and ideas, and your stuff and your memory, and the ability to work across a lot of surfaces."
So perhaps Microsoft will soon have an opportunity to choose from multiple models to offer “Copilot Cowork”. As you can sense, Microsoft and AI labs such as OpenAI/Anthropic seem to have very different ideas on how to integrate the harness layer to the model. They are both talking their own books of course, but the real verdict will come from the customers. If Microsoft wants to make the harness layer decoupled from the model layer but it ends up affecting the quality of the product itself, customers may demand an integrated experience and Microsoft may need to comply to offer the best experience to the users regardless of their own interests.
It also doesn’t help that Microsoft’s relationship with OpenAI only seems to be deteriorating over time. Financial Times (FT) reported Microsoft is weighing legal actions against OpenAI for the partnership with Amazon. Tom’s Hardware explained the key source of tension about this partnership:
“The PR about that latest agreement states that “API products developed with third parties will be exclusive to Azure. Non-API products may be served on any cloud provider.” Under that logic, OpenAI has the freedom to develop and implement new products, but if they offer them as APIs, they have to go through Azure.
Redmond believes that OpenAI’s offering access to Frontier via Amazon Web Services (AWS)’s Bedrock platform would be in breach of the agreement. Getting even more technical, the dispute may well come down to the definition of a “stateless” versus “stateful” when applied to AI models.
Even though it appears to remember your information, a standard chatbot is actually stateless — adding a new question requires the bot to re-process the entire conversation again. A storage and orchestration layer to facilitate something like Frontier is arguably a “stateful” implementation, more specifically a “Stateful Runtime Environment.”
According to FT’s sources, Microsoft thinks that running Frontier on AWS instead of Azure would breach either the spirit or the letter of the contract. This is illustrated by a report that Amazon is pointedly instructing its staff to never say that SRE “enables access” or “calls on” ChatGPT as a backend, instead preferring vaguer terms like “powered by,” “enabled by,” or “integrates with.”
I am not a lawyer, so I don’t know if the court would consider OpenAI’s partnership with AWS a breach of contract with Microsoft. However, we can indeed infer that if you don’t own the model and must rely on legal interpretations of contractual terms to defend your position, your position is a bit…fragile.
In addition to “Daily Dose” (yes, DAILY) like this, MBI Deep Dives publishes one Deep Dive on a publicly listed company every month. You can find all the 66 Deep Dives here.
Current Portfolio:
Please note that these are NOT my recommendation to buy/sell these securities, but just disclosure from my end so that you can assess potential biases that I may have because of my own personal portfolio holdings. Always consider my write-up my personal investing journal and never forget my objectives, risk tolerance, and constraints may have no resemblance to yours.
My current portfolio is disclosed below: