I think you’re missing the point. That’s what happened. But why? Why did management pull the plug before the ratchet of iterative development could really start working? Why did they go into it with a ‘Version 1 must sell or we’re out’ mentality?
It’s not for lack of cash to support a more patient approach.
I suspect it's because Microsoft wasn't ready for the kind of lift needed to really make it succeed and it wasn't until version 1 was coming together that they realized just how big that lift would be.
If we take Vision Pro as where HoloLens needed to reach technologically, they needed better displays, better sensors, and much more compute power. Those aren't core areas where Microsoft invests, and all three of those are very expensive, require a lot of people, and only the compute has a pretty reliable ability to predict when you'll hit a given goal (assuming you know how much compute you'll need - if you don't, that too is indeterminate).
And because Microsoft is an enterprise focused company, and HoloLens an enterprise focused product, Microsoft had to do a lot of the lift to sell the use case. In Apple's case, they can turn Vision Pro over to a huge army of developers to find those use cases. Consumer products are more supply-driven because the developers are often solving their own personal wants, and they'll sell your product for you saying 'hey, did you know the iPhone will help you get a ride home from the bar'. In enterprise the customer needs to see the use case and then they hire developers to build that.
Apple started working on Vision Pro in 2015. They were more than an order of magnitude off on compute than where they needed to be (probably wasn't sure exactly where they needed to be), the technology for the display hadn't been invented yet, and LIDAR for the gesture detection still looked like this.
I saw a talk by Raffaello D'Andrea back in about 2012 (can't seem to find it online) where he talked about how quadcopters came to be, and why so suddenly. Turns out small quadcopters are a threshold technology. Until tech crosses a specific threshold, they're impossible, and once they cross that threshold, they rapidly move from feasible to astounding. Quadcopters needed 4 things: powerful, compact motors requiring rare earth magnets and a few other details, very small sensors which were only arriving due to smartphones (well, the military was driving that a bit faster, but civilians weren't beneficiaries), batteries with high energy to weight ratios which were also being enabled due to phones, and enough compute power in a small enough package to close the control loop (sensor, analysis, compute, motor control) fast enough - about 60hz. Any slower than that and it's unstable and falls out of the air. And all of those things came together in the aughts.
And with some incremental gains in the first three, and Moores law driving the last one, going from something that can hover in a lab to something that can to aerobatics and catch things in the air takes a couple of years and basically every problem from there out is a software constraint - can programmers figure out how to get it to recognize an object and catch it, because the platform could easily do that almost immediately. We don't need to make meaningful gains in compute, battery, motor, or sensors because every quadcopter problem is really a problem of 'are programmers clever enough to maximize the use of what is already being built' with the answer being 'no, not really' - so human programmers are what more advanced quads are gated behind.
Vision Pro is very much the same thing. HoloLens has limited application because some of those techs just hadn't gotten over the line yet, and Microsoft was in no position to do the kind of fundamental research needed to get them there. Existing VR headsets have limited markets for the same reasons. Vision Pro gets everything just over the line enough to turn most problems into software problems - either on their end in terms of OS and APIs and what they will expose to developers (they're holding a lot back for user privacy reasons, and some I suspect because they don't have the compute available to do so) or on the end of developers themselves to come up with ideas for this. My guess is that from here out the software will reliably fail to maximize the utility of the hardware, and that gap will grow rather than shrink - Apple will be able to make incremental hardware gains faster than developers can write code. I don't mean that there are no further breakthroughs needed - things like object recognition are lacking in the product and may take ages to arrive - rather than in terms of what the product can in theory do, the gap between that and what users can actually do will expand rather than contract, which is a pretty good metric for any general compute platform vs a specialized compute platform that tends to get any new capability utilized immediately.
Apple didn't invent the displays (these are Sony displays), but they're close - they have their own microLED effort going - though with a different goal - 400ppi displays for Watch/iPhone rather than 4000ppi - we'll likely never know if Apple shot for this and failed, saw Sony was there and shifted their effort, etc. They didn't invent the lidar, but they were the first to bring it to the consumer market and design around it - there's a lot of software around the gesture detection. And they did invent the compute - nobody else can put that much compute in that power budget right now, nobody can composite 23 megapixels of pass through video at 12ms off of a battery.
Apple had a lot of other utility for all of this tech than just the Vision Pro. It drove a bunch of other device features, so it wasn't R&D that would be lost if they never got there. But for Microsoft, HoloLens was carrying that entire load, and I think it's a pretty easy call to see that the product just was unlikely to be big enough to pay for it all - especially if you have this consumer giant chasing a similar goal that will have massively larger economies of scale and democratize that so much that enterprise will just build off of the consumer platform. Remember when Blackberries were the enterprise phone? There is no enterprise phone now. Everyone just buys iPhones. Microsoft maintaining their pricing premium once Apple showed up was a long shot.