Is Microsoft Winning AI — Or Just Funding Its Rivals?

Microsoft continues to expand its AI footprint, but user engagement with Copilot is falling fast. Despite 15 million paid Copilot seats, recent data shows a steep decline in users selecting it as their go-to assistant.

The market sees a strong infrastructure story backed by cloud revenue momentum. But adoption metrics tell a different tale—raising new questions about the durability of Microsoft’s AI positioning.

User Adoption Vs Installed Base

Copilot’s reach spans over 450 million Microsoft 365 subscribers. Yet only a small fraction actively use it, and that share is shrinking. From July to January, primary usage preference dropped from 18.8% to 11.5%.

The gap between licenses sold and behavior observed is widening. Employees with access often default to rival tools, signaling that distribution alone is not driving habit formation. Over time, that disconnect changes how CIOs evaluate renewals, bundling, and long-term platform trust.

If usage keeps sliding even inside Microsoft’s own ecosystem, the bigger question becomes whether Copilot can justify its seat count when renewal decisions shift from experimentation to scrutiny.

It’s a critical signal that even best-in-class distribution channels can’t overcome a weak product-market fit. Microsoft’s enterprise footprint gave Copilot a wide runway, but user experience and daily utility are what ultimately determine product stickiness. With usage intensity declining and alternative tools gaining favor, Copilot risks becoming shelfware inside the very companies it was built to serve.

Model Dependence & Compute Constraints

For all its cloud dominance, Microsoft is not entirely in control of its AI stack. That’s a big deal. Copilot’s core experiences—from enterprise to consumer—are largely powered by OpenAI and Anthropic. Microsoft’s own models exist, but they’re still catching up. This dependence has real-world consequences, especially as Microsoft battles for inference priority on its own Azure infrastructure.

The issue isn’t theoretical. Microsoft confirmed that a shortage of computing capacity has already delayed development of proprietary models. While it’s investing in homegrown silicon like the Maia 200 and Cobalt 200 chips, much of the capacity is still being allocated to OpenAI and third-party customers, not Microsoft’s own Copilot teams. CEO Satya Nadella described the company’s metric of choice as “tokens per watt per dollar”—but that optimization still runs up against physics, silicon supply chains, and geopolitical tensions over chip manufacturing.

This dynamic exposes an awkward reality. While Microsoft helps OpenAI scale, it’s simultaneously building a product that competes with ChatGPT. The result? Microsoft is both supplier and rival in a zero-sum latency game. Until it owns more of the model layer, and scales its own chips more broadly, Copilot will remain vulnerable—not just to external threats, but to the very ecosystem it helped create.

Fragmented Copilot Experience

Part of what makes ChatGPT feel magical is its simplicity. One model. One chat interface. One seamless UX. In contrast, Microsoft Copilot has become a branding labyrinth. There are now Copilots for 365, for GitHub, for Windows, for Edge, for security, for healthcare—and no consistent user experience tying them together.

Even Microsoft employees have raised concerns. CEO Satya Nadella reportedly emailed leadership frustrated that Copilot couldn’t answer a basic prompt in Edge. That bug was fixed, but the fragmentation persists. Internally, Microsoft splits its Copilot teams between consumer (led by Mustafa Suleyman) and enterprise, with limited coordination. That structural siloing is showing up in the product itself—different interfaces, features, and capabilities across versions, with little interoperability.

Users, meanwhile, are left confused. Microsoft’s own surveys show customers don’t understand why some versions of Copilot behave differently than others. The company’s response—that this separation is “by design” to protect work/personal boundaries—rings hollow when competitors like Claude Cowork are earning praise for doing what Copilot can’t: working across applications without friction.

Competitive Pressure From ChatGPT, Gemini & Claude

In 2023, it looked like Microsoft had won the AI lottery with OpenAI. By early 2026, that bet feels a bit more complicated. ChatGPT now boasts over 900 million weekly active users, dwarfing anything Microsoft has built. Gemini and Claude are growing rapidly, with each launching collaborative tools and multi-modal interfaces that rival (and in some cases surpass) Copilot’s capabilities.

Most concerning for Microsoft is the shift in preference. When workers have access to multiple tools—Gemini, ChatGPT, Copilot—they’re picking the non-Microsoft options more often. Users cite better quality, fewer restrictions, and more intuitive design. Anthropic’s Claude Cowork, in particular, has raised eyebrows for integrating smoothly into Microsoft 365—something Copilot still struggles to do.

At the same time, OpenAI is moving faster and more independently. With GPT-5.2 already live and outperforming rivals in benchmark tests, Microsoft is at risk of being leapfrogged by the very startup it helped fund. That’s the paradox: Microsoft fuels OpenAI’s rise, only to watch its own competing product fall behind.

The competitive pressure isn’t just product-based—it’s narrative-based. Right now, ChatGPT is the face of AI. Copilot is… the assistant you didn’t ask for in Word. Until that shifts, Microsoft risks being seen not as a leader, but as a fast follower in the very revolution it helped catalyze.

Final Thoughts: Power, Platforms & Perception

Microsoft’s position in the global AI race is both dominant and defensive. It owns the cloud, sells the seats, and runs the infrastructure. But when it comes to end-user mindshare, Copilot still trails behind faster-moving rivals. The company has the resources to catch up—its LTM valuation still commands a 25.9x P/E and 21.7x EV/EBIT, down from highs earlier last year but still rich by historical standards.

The question is whether those multiples reflect growth in AI, or optimism around the broader business. Microsoft’s Copilot story is a case study in how platform power doesn’t always guarantee product love. It’s also a sign that the AI value chain is shifting—away from surface integrations and toward model control, compute advantage, and user trust.

Whether Microsoft becomes the AI leader or just its infrastructure backbone will depend on how fast it can resolve Copilot’s identity crisis and close the UX gap with ChatGPT and others. Until then, “Microsoft Copilot AI struggles” may remain more than just a headline—it could be the defining narrative of this phase in the AI era.

Disclaimer: We do not hold any positions in the above stock(s). Read our full disclaimer here.