Microsoft offers mass buyouts for roughly a quarter of its workforce

Microsoft offered buyout packages to roughly 60,000 employees on Monday. That's a quarter of the company. Not a quiet "restructuring" or some vague "strategic realignment." Just a mass offer asking a huge chunk of its workforce to collect a severance check and leave.

Here's the part that should make you pause: Redmond doesn't actually expect most of them to say yes. The company is betting that only a fraction of those 60,000 people will actually take the money and walk out the door. This isn't a layoff in a nicer outfit. It's a different kind of move entirely.

So what happens when you offer 60,000 people an exit ramp but quietly hope they keep driving? And why would a company with Microsoft's resources need to trim headcount this way? That's what I want to figure out.

Technical Overview

The Microsoft-Inflection deal isn't an acquisition in the traditional sense. They paid $650 million for a licensing agreement and hired most of the staff, leaving Inflection's corporate entity intact but hollowed out. Technically, this means Microsoft doesn't own the model weights outright; they own the right to use them, deploy them, and presumably fine-tune them. It's a strange arrangement that lets Microsoft absorb the talent and infrastructure without absorbing the liability or the regulatory scrutiny that comes with owning the IP.

When you acquihire an AI lab, you're buying more than GitHub repos. You're buying the training clusters, the data pipelines, the evaluation harnesses, and the institutional knowledge of how to keep a large model stable during training. Inflection's team brought specific expertise in fine-tuning and personalization that slots into Microsoft's existing Copilot infrastructure. The quote about employees asking whether to take the buyout suggests this was as much a retention play as a technology grab. Microsoft needed the people to maintain and improve the models they were licensing.

The speculation that Microsoft achieved AGI internally changes how you read this deal. If they already have systems that match or exceed what Inflection built, then the $650M makes sense as a talent acquisition and a hedge against antitrust action. But verifying AGI claims requires rigorous testing. Companies use automated evaluation suites that probe reasoning, planning, and tool use across hundreds of tasks. Here's a simplified version of how that works:

import openai

def evaluate_model(client, model, tasks):
    results = []
    for task in tasks:
        response = client.chat.completions.create(
            model=model,
            messages=[{"role": "user", "content": task}],
            temperature=0
        )
        # Check if response meets the success criteria
        passed = "solution" in response.choices[0].message.content.lower()
        results.append(passed)
    return sum(results), len(results)

client = openai.OpenAI()
tasks = ["Solve this logic puzzle...", "Write a Python parser...", "Plan a trip..."]
passed, total = evaluate_model(client, "gpt-4", tasks)
print(f"Capability score: {passed}/{total}")

Real evaluation suites are far more sophisticated, with human raters, sandboxed execution environments for code, and multi-turn conversations that test long-term planning. But the principle is the same: you need automated, repeatable tests that stress the model's ability to generalize across domains it wasn't explicitly trained on. If Microsoft's internal systems are passing these tests at rates that dwarf public benchmarks, the Inflection deal looks less like a technology purchase and more like poaching talent while keeping their own cards hidden.

Conclusion

Microsoft sent buyout offers to 60,000 people—roughly a quarter of the workforce—and expects only a fraction to actually leave. That gap between the offer and the expectation is the whole story. This isn't a layoff by another name; it's an expensive bluff, or maybe just a signal to Wall Street that costs matter, dressed up as generosity. Either way, the employees who stay are now sitting in meetings wondering why their job was on the list, and whether next quarter's package will come with a deadline.

I'm still not sure what Microsoft thinks it's buying here. A leaner company? A more compliant one? The math only works if they were always planning to miss their target. Watch the acceptance rate over the next thirty days. If it comes in under 10%, this was pure theater. If it hits 30%, they're cutting deep and just didn't want to say the word "layoff."