The most obvious way to do this is using a QCL, which can provide a midIR spectroscopy source.
This has been shown in the lab:
Princeton Engineers Develop a Non-invasive Blood Glucose Test that Uses Imaging Technology | Dark Daily
The problem with using a QCL light source is that QCLs are (at least right now) expensive --- hundreds to thousands of dollars per chip, depending on exactly what parameters you need. This cost can probably be brought down substantially IF someone were willing to commit to purchases of millions, but right now few units are sold every year, so there's not yet been a reason to create the sort of automated single-purpose factory we associate with most chips, instead QCLs with different specs are manufactured in small batches on machines that are then reconfigured for the next design.
This manufacturing, BTW, has pretty much nothing to do with the sort of thing TSMC does, and little to do with mass-market semi-conductor lasers. QCLs are constructed atomic layer by atomic layer, each layer a particular element, with ~hundreds of layers in a particular order to hit a particular wavelength.
So QCL cost is the big issue today. Are there ways around this? No other good/general midIR laser source is know. There are a few alternative midIR lasers (CO2 lasers are well known for example) but these are fixed frequencies and can't be designed to target a particular frequency the way you'd like for generic spectroscopy. But if you don't care about generic, just about one or two particular wavelengths relevant to glucose?
It's possible that Apple (or more likely a partner company) have found an alternative light source that can be forced to detect glucose, and that can be manufactured much more easily. It's also possible that Apple and partner have agreed to (or are working out) some sort of plan whereby Apple finances a dedicated QCL fab that, because it is fixed function, is able to crank out millions of glucose sensor QCLs (as opposed to any other sort of QCL) at a per-chip cost of just a few dollars.
An alternative way to solve the problem relies on chemistry, not light. That's being pursued by some companies, but involves consumables --- basically an ink-jet printer model where the printer is cheap but you have to keep buying expensive ink. It's not clear if Google's contact lens solution is based on this, but that's what I would guess.
The way this used to look like it would play out is that the chemistry solution (cheap up-front, on-going monthly cost) would be sold to consumers, while the optical solution would be sold to medical practitioners (just like offices that are well-managed buy laser printers not ink jet printers, given their volumes). Obviously if (if!) Apple have some sort of mass production scheme planned for their QCL (or other?) optical source, this up-ends the expectations. Especially if their partner manufacturer is allowed to sell the chips to non-Apple.
Finally, of course, remember that Apple tries to do lots of things that are stretch goals, and sometimes these play out and sometimes they don't. Remember sapphire? It's extremely misleading to state that "Apple announced" this on CNBC today. CNBC reported something (based on who knows what), and OFFICIALLY, Apple has, as usual "declined to comment". Maybe someone in the team got drunk and talked too much? Maybe Tim Cook, for whatever reason, was willing to talk to a journalist providing unattributed deep background? Who knows?
The point is --- Apple doesn't officially announce stretch goals until they actually hit those goals, for the very good reason that there's a chance that they might not be able to do what they want. Yes, we all want this, but it may not be achievable this year, it may not be achievable next year, hell it may not be achievable in our lifetimes within the constraints of an Apple Watch at Apple Watch (~$300) prices.
