I spent a lot of time in graduate school researching AR/VR technology (specifically regarding its utility as an accessibility tool) and learning about barriers to adoption.
In my opinion, there are three major hurdles preventing widespread adoption of this modality:
1. *Weight*: To achieve powerful computation like that of the HoloLens, you need powerful processing. The simplest solution to this is to put the processing in the device, which adds weight to it. The HoloLens 2 weighs approximately 566g (or 1.24lb), which is a LOT of weight compared to a pair of traditional glasses, which weigh approximately 20-50g. Speaking as someone who developed with the HL2 for a few years, all-day wear with the device is uncomfortable and untenable. The weight of the device HAS to be comfortable for all-day use, otherwise it hinders adoption.
2. *Battery*: Ironically, making the device smaller to accommodate all-day wear means that you're simultaneously reducing its battery life, which reduces its utility as an all-day wearable: any onboard battery must be smaller, and thus store less energy. This is a problematic trade-off: you don't want the device to weigh too much that people can't wear it, but you also don't want the device to weigh too little that it ceases to have function.
3. *Social Acceptability*: This is where I have some expertise, as it was the subject of my research. Simply put, if a wearer feels as though they stand out by wearing an XR device, they're hesitant to wear it at all when interacting with others. This means that an XR device must not be ostentatious, as the Apple Vision Pro, HoloLens, MagicLeap, and Google Glass all were.
In recent years, there have been a lot of strides in this space, but there's a long way to go.
Firstly, there is increasingly an understanding that the futuristic devices we see in sci-fi cannot be achieved with onboard computation (yet). That said, local, bidirectional, wireless streaming between a lightweight XR device (glasses) and a device with stronger processing power (a la smartphone) provides a potential weigh of offloading computation from the device itself, and simply displaying results onboard.
Secondly, Li+ battery tech continues to improve, and there are now [simple head-worn displays capable of rendering text and bitmaps](https://www.vuzix.com/products/z100-smart-glasses) with a battery life of an entire day. There is also active development work by the folks at [Mentra (YC W25)](https://www.ycombinator.com/companies/mentra) on highlighting these devices' utility, even with their limited processing power.
Lastly, with the first two developments combined, social acceptability is improving dramatically! There are lots of new head-worn displays emerging with varying levels of ability. There was the recent [Android XR keynote](https://www.youtube.com/watch?v=7nv1snJRCEI), which shows some impressive spatial awareness, as well as the [Mentra Live](https://mentra.glass/pages/live) (an open-source Meta Raybans clone). In terms of limited displays with social acceptability, there are the [Vuzix Z100](https://www.vuzix.com/products/z100-smart-glasses), and [Even Realities G1](https://www.evenrealities.com/g1), which can display basic information (that still has a lot of utility!).
As an owner of the Vuzix Z100 and a former developer in the XR space, the progress is slow, but steady. The rapid improvements in machine learning (specifically in STT, TTS, and image understanding) indirectly improve the AR space as well.