Arguably they are already powerful enough to make the leap in at least some cases, that said Apple does not push Macs in the same quantities as iOS devices as to justify wholly new silicon for the line. I have been in an argument with a friend for the past year or so about precisely when (or if) Apple will rip Intel out and use their own silicon. I think they will, but I'm going to try to present his argument to the best of my abilities, with my own observations and guesses as context.
Presently the T series is generally gimped variations of the A series. My understanding is that the T2 is essentially an A10 with the high performance cores disabled, and the low performance cores repurposed to manage the myriad of functions that were previously handled by dedicated controllers. They're also necessary for certain features like TouchID, and for Apple to use its own HEVC encoder/decoder.
Presumably the T3 will be keyed off of a later part like the A12 or a later generation, which will make it the first T series chip to also include the neural engine. Assuming they don't disable that part, it will potentially make the NE available to Mac application software.
So what's the real problem with ripping out Intel right now? Given a higher TDP, Apple could hypothetically design a much more performant chip that would replace both the CPU and GPU in their current Mac lineup, but they simply don't push enough Macs to justify the expenditure on custom silicon, right now at least. Probably doesn't help that they have been making existing customers less happy, but it also doesn't help that Intel has simply been re-releasing Skylake in different variations for a good few years now.
If you have a Macintosh able to run Mojave, an SSD, and at least 16GB of RAM, you are probably set. The important thing about Mojave isn't so much any one feature of Mojave, it is the continued support including bug fixes and security updates. Given all of that, most people simply won't need a new system any time soon. There's essentially no reason a laptop from 2012 can't take you all the way to 2022 all other things being equal, so we're starting to see sales stagnate across the entire PC market including the Macintosh, so while Apple is selling phones about as fast as they can stamp them out, the same is not true of Macbooks or iMacs.
Silicon is much more of a volume business. Apple can re-purpose older versions of their A series chips because those are an R&D and capital investment that has already paid off, it isn't a new design, and they just need someone to continue fabbing them. New generations of the A series chips will generally pay off, because they will be selling hundreds of millions of the same iPhone, and usually for greater than one calendar year. This past year, for the first time since the A7 was released, they skipped releasing an -X variant of their A series chips which are generally released to support new iPads (there was no A11X for those who may have forgotten). It is possible that iPads have not been selling in the quantities necessary for even a variant design, and rather than an annual refresh cycle, we could reasonably expect them to move to a biennial refresh cycle, hence why they have skipped the A11X and jumped straight to the A12X.
Intel and Qualcomm have customers, but for Apple's own silicon, Apple's only customer is itself, by design. That also imposes a limitation on themselves that it just might not make financial sense to fabricate a series of chips exclusive to their Mac lineup, appropriate to the TDP of a Mac.
By all means, tear this apart if you disagree.