I guess Apple has to refresh all their lineup asap, given that they need to convinced investors they are all in on AI, which they will claim need M4 to run well/fully.
In the 'old fashion' days investors actually looked at a company's financials to decide whether it was a good investment or not. Now have drifted more into the which company has a better hype train to ride as opposed to does this P/E ratio make any rational sense. Sadly there is now a notion of 'investors' which primarily folks playing the 'bigger fool' game. Find someone else who looks even less at financials and will buy it at a higher price.
If Apple dumps the M3 generation dies too fast they will probably have trouble paying for the R&D. That would be a net loss. AI is a software-hardware system. M3 runs Siri fast but Siri's software is the highly dubious part. Folks act alike Apple hasn't been putting multiple AI/ML accelerators in the Mac for years. It is the software that is moving slow as molasses. Quickly tossing the M3 into the garbage can isn't going to make a difference. Likewise the VAST majority of folks with M1-M3 systems are not going to buy anything that quickly.
Apple probably won't be selling M3 systems in 2 years. ( e.g, the iPad Air pretty good chance goes from M2 -> M4 and M3 is never seen in iPad line up. ). The M3 Max will probably get a 'short cycle' but those systems sell at a higher premium. If Apple skips the M3 Ultra then totally saves on that R&D cost also. So far the rumors indicate that Apple will ride the MBA M3 systems for about a full year. Those are Apple's best selling systems and best chance to use volume to recoup the shared R&D overhead investment that crosses all of the M3 generation.
Windows has a very different issues. Windows 10 is suppose to be desupported in 2025. Yet 60% of folks are still on it. The overall average age of WinPC systems is older than Macs. That market has a vastly bigger "motivation" problem of getting off older hardware onto newer stuff. ( mainly folks who have pre-pandemic systems and didn't upgrade during the that 'buying boom' cycle. )
In contrast, Apple has 60-70+ % folks on the lastest version of the OS. The macOS ecosystem doesn't really have a major "squat as long as possible" faction. The average upgrade cycle is incrementally better than Windows. There is a substantive "keeping my Mac Pro , box with discrete GPUs cards forever" crowd but relatively not that large to aggregate ecosystem. MBP and Mini folks on the Intel who haven't moved the M4 even without "AI magic sauce" would uncork a decent number of folks. Their system is likely falling into Vintage/Obsolete if more than 5 years old. The performance gap is even more large. etc. The AI stuff is just a 'cherry on top'.
The 'new' AI/ML coming excluding M1-M3 to be useful would be telling about 36 Million users "no soup for you". (Apple is averaging about 12M units per year). That is dumb. It has to be useful for those folks. That is VAST majority of the possible Mac user base. The Mac user base is around 80-100+ M. They only every 'turn over'
only about 12-15% of that. If just start selling the M4 that would be less than 6-7% of the user base getting very substantive utility.
The M4 selling point would more so the AI/ML 2-3 years down the road runs better. Not that M2-M3 can't do a decent job on the current stuff. And Apple's AI/ML is extremely likely going to run on some range of A1x SoCs also. That is an even larger user base where software that skips everyone has limited uplift on moving the whole ecosystem forward. Siri has to "suck less" on those systems too.