That's an apples to oranges comparison about being a standalone device. Requiring a device for setup vs requiring a device to provide a constant connection to the headset are 2 different things.
The rumor report in the first article starts off with this
"... development will need to be wirelessly tethered to an iPhone or another Apple device to unlock full functionality ..."
It is not very clear so far that that there is going to be a tethered lock to do basic AR functionality. "Full functionality" would probably loop in VR. But VR is likely not the sole function for these goggles.
"...without a neural engine for AI and machine learning capabilities. The chip is designed to optimize for wireless data transmission, compressing and decompressing video, and power efficiency for maximum battery life. ..."
What isn't outlined here is how much Apple added and subtracted on the image processing cores ( the image processor). If the camera modules are made "smarter" on the googles they may not have to transmit back as much. But chopping off the "Neural" cores means that any heavy lifting , more general inference would need to be shipped out and back. (Alexa does that all the time for extended speech recognition and it supposedly it is "standalone". )
I consider my alexa devices to be standalone devices, and they require a phone to setup, just like an apple TV, philips hue lights, etc. Those are different than my apple watch, that doesn't have full functionality without my phone.
Turn off the wifi router in your house and ask alexa and significant question and see if you get an answer.
Look this is somewhat a matter of how you draw the lines around a completed system and how much is required to be done locally inside a single box .
These VR googles are particularly bad ( to at best awkward) for substantive data entry. So pragmatically not standalone in that mode.
Also, I'm not sure where you are going regarding battery life. I'm no expert, but having everything on board a device like the Oculus 2 seems like it would be more battery efficient than constantly having to communicate to a separate device.
Apple's "cherry picked" battery testing task for benchmarking battery life on Mac or iPhone is wirelessly browsing video.
https://www.apple.com/mac/compare/ ( scroll down to battery life and Power )
https://www.apple.com/iphone/compare/ ( again scroll down to battery life and Power )
That puts most of the work onto the wireless system and the fixed function video decoders ( and minimized CPU and GPU core work. ) . They are not picking the most CPU and/or GPU challenging thing here to measure battery life. They are picking one of the easiest. ( really not just Apple, all the laptop and smartphone vendor have drifted toward this as a common benchmark practice. ) .
Some folks are stuck thinking that video watching means Flash plugin and its software video decoder. That is completely decoupled from reality these days. The vast majority of that task has been sucked into non-generic hardware at point. Also similar to folks claiming the M1 cores are super powerful when process some H.265 4:2:2 compressed video faster than a Mac Pro (when that is really more so presence or absence of a hardware video decoders; on or off the CPU cores. )
So an even more customized chip to possible throw even more fixed function logic at these narrow set of tasks at the probable die space (and transistors ) reduction of a much higher number of CPU cores and GPU cores (and AI/ML cores ) is extremely probably going to get even higher battery life. One of the principle design objective was lower battery consumption is quoted directly in the rumor description. If Apple wasn't getting what they explicitly set out to do that would mean they were failing at that. If they have taped out a solution, a substantial number of simulations and prototype work probably indicates that it is working.
If Apple stuffed the CPU and GPU into an Apple wireless keyboard the power consumed and required battery capacity would go up for the keyboard ; not down.
Throw on top then don't necessarily need to put a powerful transmitter on the googles if the other Apple device is within 10-12 ft of the googles. Again wireless Keyboard and Mouse from Apple as major battery consume and they are completely wireless.
The new tech that Apple is leveraging here that the most previous goggle implementations didn't have available earlier is faster wireless bandwidth. Apple is probably using a short range, direct point-to-point wireless connection over WiFi 6 , if not 6E, coupled to heavy hardware compression to make it work. Generic WiFi 5 ( or lower) this would not work well.
Quest 2's WiFi on a generic multiple device shared household network pulling data from a router possible 100 ft away through multiple walls. Yeah that probably would use more power. However, that is exactly what Apple is not going to do for this computational offload. The Apple WiFi watch doesn't need a phone to connect for WiFI services though;
Choose which Wi-Fi network your Apple Watch connects to when your iPhone isn't nearby.
support.apple.com
If these goggles behave like an WiFI Apple watch ( "lIt will be similar to the WiFi-only version of the Apple Watch, which requires an iPhone connection to work "). Requires an iPhone to work isn't completely accurate either over all functions.
Regardless, we are also assuming that these will be competitive devices, which they very well may not be.
Although it has no massive general purpose , "grunt" compute capability in it (or relatively large SSD storage capacity) , the Apple goggle will probably be 2-4x more expensive. The screens , camera modules , RAM , and this custom chip is where most of the money will go.