MICROSOFT: BEWARE THE BIG BRUSSELS UNBUNDLING

TL;DR

Microsoft doesn't have its own AI model. And yet it starts in the dominant position - and may maintain its lead - during the next round of the AI Wars.

Redmond's bet: AI models will become commodities. But context won't. And Windows sits beneath everything - seeing every file opened, every workflow crossing applications, every piece of institutional memory. So you can switch AI models with a dropdown. But you cannot switch away from Windows without rebuilding your entire computing environment.

This creates a system that captures context but that nobody leaves. Not because leaving is forbidden, but because leaving is far too complex and expensive.

We have found three things that could break this dominant position. First, Microsoft refuses to implement context portability - the ability to take all that context to another provider - until Brussels forces it. This may allow competitors to position themselves as trustworthy alternatives whilst Microsoft's moat drains.

Second, personal preferences may capture executive attention, pulling the most important AI use out of the productivity suite.

And third, infrastructure costs may outrun per-seat revenue, forcing visible throttles that weaken Microsoft's value proposition.

Any one scenario weakens Microsoft's position. Two together could reshape the contest entirely. All three at once may turn the system nobody leaves into the system people actively migrate from.

//

The longer read👇🏼

The System No-one Leaves. But for How Long?

The frontier AI providers have spent years racing to build smarter models. But, as already discussed, intelligence is now almost table stakes.

Microsoft CEO Satya Nadella put it this way: "As AI becomes more capable and agentic, models themselves become more of a commodity."

The decisive factor in the current Battle for AI Supremacy is, therefore, likely to be which Big Tech behemoth controls the context AI models need to be useful. Which includes being able to see your emails, calendar, documents and conversations. As well as the institutional memory scattered across your organisation's systems.

And as soon as you start to look closely, the depth of the ironic implications of Redmond’s position reveals itself. Microsoft is not a frontier developer. And yet despite this key deficiency in a battle for dominance - lacking the very product the market is fighting over - it was always going to emerge in pole position. Because Microsoft sits in, under or around most of our work output.

Which explains why Microsoft has opened the intelligence layer whilst tightening its grip on context. Users will be able to choose between OpenAI's reasoning models and Anthropic's Claude in Copilot. Developers can build agents powered by either in Copilot Studio.

But you cannot switch from Windows context to an alternative without migrating your entire computing environment. The operating system sees everything you do - every file opened, every application launched, every cross-application workflow. That visibility doesn't require your consent to share. It's architectural. Windows is the layer beneath everything else.

The only question we need to ponder, therefore, is whether its position on the starting grid translates into a long-term durable advantage. Or whether regulators and competitors chip away at the foundations faster than Microsoft can consolidate them.

Why Microsoft Emerges All Guns Blazing

A quick glance across your working day confirms why Microsoft starts in a strong position on the map. Let’s imagine that you need to make a decision on a proposal before lunch. Once all Microsoft’s apps and components are tightly integrated you’ll be able to prompt Copilot in Word to assess what it recommends, given everything it knows.

Copilot, of course, immediately understands what you're doing and why you’re asking it for help. Because it already has the context. It knows where to pull everything together from. It reaches back through yesterday's Teams discussion and pulls figures from a spreadsheet - which someone unbeknownst to you updated this morning, which turns out to make a material difference. On that basis it realises it needs to check the current policy document to ensure the proposal remains in policy. Given what it finds, it drafts a final recommendation and drops it into an email for you to review.

Let's be clear here. You didn't tell it where to look. Or how to connect these pieces.

This works because everything runs through the same system. Windows sees what you're doing. Microsoft 365 is where the documents are edited. And conversations about the deal are happening in Teams, while the schedules are coordinated in Outlook. Behind this, a unified data layer connects everything you've written, received or stored. When you make a request, the system retrieves what's relevant, sends it to an AI model running on Microsoft's infrastructure and returns an answer.

The same login grants access. The same security rules govern what can be seen. The same audit trail records what happened.

This becomes weight-bearing context. Not because the individual components are particularly sophisticated, but because they're joined without friction. The operating system where work happens, the productivity suite where it's recorded, the OneDrive memory layer that spans both. The Edge browser and other integrations extend the reach, but these are the foundations.

The system buys consistency, speed and compliance by construction. It also buys lock-in. Leaving the rails Microsoft has built means rebuilding that a complete corporate memory elsewhere. Which is not forbidden. But it is very complex and expensive. Uneconomically so.

This opening position on the battlefield, therefore, creates what might seem like unassailable context gravity. Microsoft presents as the obvious stock to back in the Context Wars.

But Theres a Flaw

This all sounds very compelling. But there is a potentially huge ‘whoops!’ in this plan.

The integration that makes the system so valuable also pins a large, fluorescent target on Microsoft’s back. Total context ownership by one large player will effectively lock everyone in to one platform. And that, to say the least, is an anti-competitive position.

The EU has previously disallowed instances which feel like minor infractions compared to the complete ownership of context. For instance, Microsoft bundled Teams with Office 365, arguing this was simply good product design. Of course your chat app should connect seamlessly to your documents and calendar. Brussels, however, disagreed. The European Commission forced unbundling, required data export on request and mandated interoperability standards.

Teams still works for us as it ever did. But it no longer comes automatically with the suite - and organisations can now extract their Teams data and move it elsewhere without losing context or history.

This wasn't a one-off remedy. It was the application of an existing template. And the same template will almost certainly be applied, in spades, to Microsoft's play for all context. Just because Microsoft can, Brussels will reason, doesn't mean it should.

When Copilot becomes essential infrastructure threaded through Windows, Office, Teams and Edge, regulators will examine how anticompetitive the behaviour is. Can organisations switch to alternative AI assistants without rebuilding their entire workflow? Can the enterprise use a different productivity suite without stabbing itself in the face?

Is context reasonably portable? Or does Microsoft's memory layer create lock-in by design?

The technically correct answer - that it’s the entire platform and its seamless integration that makes this context possible - won't satisfy competition authorities. Microsoft's response will need to be expressed in product releases, not as a technical argument.

We might therefore expect that Windows will need to make context as portable as a Word document. That context will include where every input was made and where every output came from. The memory export feature will need to be extensive - so extensive that any organisation can extract their context and take it elsewhere, wherever they deem optimal. And the context delivered will need to be of such high quality that competitive AI assistants run like native apps. Anthropic in Word, Google in Teams, without the experience degrading.

Just as Word documents now look great in Google Docs or on Apple Macs.

We can conclude, therefore, that regulators are likely to limit the extent of Microsoft's dominance. And Redmond likely knows this. Which helps explain the partnerships already in place with Google and Anthropic. The Brussels unbundling is likely to be coming - the only question is whether Microsoft builds for it proactively. Or fights it reactively.

But regulation isn't the only challenge. The economics become more complex - and perhaps even marginal - as usage scales.

Context density improves answers, but retrieval, storage and inference all cost money. Right now, Microsoft charges per seat - a predictable monthly fee that hides the actual computational cost underneath. As people use Copilot more intensively, as the system pulls from richer context across more apps, as queries become more sophisticated, the gap between per-seat revenue and per-query cost narrows.

Azure can absorb this for a while. Microsoft has the infrastructure, the capital and the strategic patience to subsidise adoption. But not indefinitely. If Azure's growth rate slows whilst AI-related capital expenditure and energy costs continue climbing, something has to adjust.

Either pricing changes - adding usage caps, tiered access or additional fees for deep context retrieval. Or scope narrows - grounding windows get tighter, cross-app context becomes selective, the seamless integration that defines the platform starts feeling constrained.

Then there's the OpenAI relationship. Microsoft still routes enormous demand to Azure through its partnership, and that volume creates leverage with both OpenAI and enterprise customers. But the terms have shifted. The revised agreement between the two has narrowed Microsoft's exclusivity, whilst OpenAI pursues multi-cloud distribution and direct enterprise relationships.

If OpenAI's agentic workflows start capturing executive attention - if decision-makers begin their day in ChatGPT rather than Outlook - then context starts flowing to wherever the thinking happens. Microsoft becomes the execution layer. Still important, but no longer strategic.

All of this, of course, only breaks the strategy if Microsoft assumes it needs to control the model. If the advantage Microsoft seeks is actually continued ownership of the rails - rather than the entire train set - then model independence strengthens rather than weakens its position.

This would allow customers choose their AI engine. And Copilot becomes the neutral orchestrator that works equally well with OpenAI, Anthropic and Google. Or whoever comes next.

But that requires Microsoft to make neutrality feel natural rather than bolted-on. If switching between AI providers within Microsoft's ecosystem feels clumsy, enterprises will look for alternatives. And regulators will demand it.

Microsoft's best strategy, therefore, may be to make context portable from the off. And if that feels seamless to all, the moat will adapt. And likely keep Microsoft safely locked up inside for at least five years.

Can Competitors Disrupt Microsofts Strategy?

Most certainly. And especially if they execute brilliantly over the next 18 months.

That would mean Google finally making its whole greater than the sum of its parts. Chrome, Gemini and Workspace will need to stop feeling like separate products and start acting like a unified platform. Taking back the technical lead from OpenAI in browsers, Chrome could then continue to become the place where you do almost all things. Which means we'd also need to assume that Gemini becomes indispensable across all this browser-based context. Workspace would need to start winning more often against Microsoft 365 as well. Perhaps because regulated industries and government recoil at Microsoft's smash-and-grab for context. And then Brussels starts waving people away from the platform too.

If Google makes a credible play for less invasive context - promising similar capability without operating system visibility - it could own a significant slice of the territory. Why would you allow Microsoft to see everything happening on your devices when Google can deliver comparable intelligence just from your browser and cloud apps?

At which point, Microsoft's advantage contracts sharply. Its strategy remains decisive only in organisations where the memory layer is already rich, deep and of continuing relevance - where years of Windows, Office and Teams use have created context that's genuinely expensive to rebuild. And even there, only until Brussels imposes context portability.

The OpenAI challenge is also browser-based. Given ChatGPT's domination of the consumer space it is, by default, where many executives start their day. The recent addition of the daily Pulse report - fuelled via extra context from using the new Atlas browser - is fuel to this fire. And if decision-makers begin their morning in ChatGPT rather than Outlook - if that's where they think through problems, draft plans and work out positions - then context starts accumulating outside of the productivity suite. And Copilot becomes the thing that executes them downstream in Word and Outlook. To what extent could OpenAI then push further? And how long before Microsoft's integrated experience starts looking like execution infrastructure - not strategic platform.

Anthropic's threat, of course, is of a very different nature. It can be expected to position neutrality not as a compromise but as premium positioning. The AI for organisations that cannot afford vendor lock-in. And in regulated sectors - finance, healthcare, public administration - that describes most serious buyers. And we can suppose that these organisations might well start testing different models to see which works best.

Whilst none of these threats can threaten Microsoft's stack entirely, each attack chips away at Microsoft's hope to own it all. Browser gravity may feel like a better compromise than operating system visibility. And model independence might become mandated. But even before then, a decision to make context portable might temper Microsoft's ambition. And send a signal of trust to regulators.

Thread that needle successfully and the position holds. Fail, and the weight-bearing context becomes contested ground.

What Microsoft Will Likely Do Next

So how does Microsoft respond to all this? We might consider a potential shift in how the company presents itself - and how it builds.

The platform that nobody currently leaves may be far better positioned as the platform that people can leave - but choose not to. And that means making the infrastructure layer far more visible. Right now, most organisations use Microsoft's unified memory layer. But how many really need to understand how it works or where their data lives? When context is wrapped in, the answer might become 'mostly everyone'. And the opacity, which has so far been something of a benefit to Microsoft, suddenly becomes a vulnerability. If regulators or competitors raise questions about lock-in, the default response becomes suspicion.

The logical move? Making the identity controls, memory layer, the inference pipeline and the policy centre into front-and-centre products. Allowing administrators to see these components in ways they can test and audit. And allowing context to become portable.

Would Brussels leave Microsoft be? Proactive, pre-emptive compliance is, in fact, leadership. And it might just improve rather than hurt Microsoft's claim to suck up all context.

Of course, the feature set would need to be genuinely comprehensive - good enough that any organisation could extract their complete context and reconstitute it elsewhere if they chose to. The goal isn't to encourage switching. It's to remove the fear of being trapped, which is what prevents adoption in sectors where governance matters most. Finance teams, biopharma, healthcare providers, public sector buyers - they all need to know they can leave before they'll commit to staying.

This would mean model choice can then become native rather than grudging. Policy teams can select which AI engine runs which workflow without the experience degrading noticeably.

In parallel, Microsoft is likely to focus on tuning the economics. Retrieval is likely to become more disciplined by default. Smaller grounding windows, smarter caches, more selective cross-app context. Copilot will have access to everything but learn to pull in only what's likely truly relevant, rather than everything just in case. The pressures here are both to manage costs and to reduce the sense of invasive surveillance.

We might expect all this to happen faster than Microsoft would prefer. Because the alternative - waiting for regulators to force each capability as a separate remedy whilst competitors position themselves as the trustworthy alternative - seems, from the outside at least, increasingly expensive.

Better to build for portability as a product strategy than fight it as a compliance burden?

But there's tension here that won't resolve neatly. The more open Microsoft makes the system, the lower the switching costs become. The more they enable model choice, the less they benefit from OpenAI's partnership as traffic gets driven to competitors. The more comprehensive the memory export, the easier it is to rebuild context elsewhere. Threading this needle - making the system feel open where policy demands it whilst keeping the integration tight enough to remain valuable - will determine whether Microsoft's position holds or erodes over the next few years.

The System Nobody Leaves - Until the Big Brussels Unbundling?

It is certainly true that Microsoft doesn't have its own AI model. It’s reliance on OpenAI for intelligence, with some hedging through partnerships with other providers, does make strategic sense in the battle to win the AI Wars.

Because Microsoft isn't competing on intelligence. It's competing on context. And if the battle is really about context - about who controls the data, documents, conversations and institutional memory that make AI useful rather than simply clever - then Microsoft starts with something formidable. The system nobody leaves. And this could make Microsoft the dominant player in enterprise AI for years.

Unless one of three things breaks the strategy. Or, more dangerously, two in combination.

First, we can imagine that Microsoft refuses to implement portability until Brussels demands it. This will allow competitors to position themselves as the trustworthy alternative. At which point, the moat may drain faster than Microsoft can adapt.

Second, a model-first personal environment may captures executive attention and pull context away from the productivity suite. If OpenAI's agentic workflows or Anthropic's memory spaces become where strategic thinking happens - where senior people start their day- then context starts accumulating outside Microsoft's platform.

Third, costs could outrun monetisation and force throttles users can feel. Which weakens the product and helps the competition.

Any one of these scenarios weakens Microsoft's position. Two together might reshape the contest entirely. All three at once may turn the system nobody leaves into the system people actively migrate from.

But how likely is it that the other players on the map will exert the pressure required to derail Microsoft? Next time we’ll take a closer look, starting with Google.