Yes I think Facebook and Microsoft may lose money on this, you could call that over investing, but for it to be a bubble that burst, companies need to go down, and people lose money on it.
Otherwise you may be looking for a completely different word which is fad, a period where something is over hyped, but quickly settles down like Fidget Spinners.
.com was a bubble because a lot of people lost a lot of money, so much so that it reverberated through the entirety of financial systems globally. I don't think we will see a financial bubble burst due to AI, that is more than 1% of that at most.
META is throwing out money like crazy on their "META" project that will never succeed, and they are doing the same with AI, but they will not go bankrupt on it, and AI will almost guaranteed be more successful than the "META" project.
What I mean is that generally more money will be made on AI, than is lost experimenting with it.

I'm of two minds about this. On one hand, I do think it will burst. It reminds of the ludicrous claims made about the last two few VC tech bubble trends, like VR and Blockchain. The hype wasn't that it was a useful technology. It was that these were the new paradigm shifts. These will change how society fundamentally functions. Obviously they didn't. Obviously those bubbles burst.

Part of the reason, I think, is that the current round of venture capitalists made their fortunes on the internet itself. It was the paradigm shift, and it toppled the way people had done things for a hundred years in a way that can't really be described to anyone who didn't live through it. It colonized and conquered every space humans went, and became ubiquitous. Retail stores found themselves under siege by amazon. Video stores found themselves obsoleted by streaming platforms, cable TV and movie theaters fought for relevance. It made some men richer than God. A computer in the palm of your hand, allowing you access to the totality of human knowledge and the collective of human communication. It was like the fucking ansible.

Those structures have calcified now, and the internet is at its limit for integration. So tech bros latch on to ever more destructive technologies named after ever more dystopian sci fi, figuring that throwing a billion at any random project is worth it if pans out once again, and it becomes the next paradigm shift. The problem is that all the projects they try to elevate are mostly just ways to disrupt existing industries and reform them under their control without worker protections. Uber operated at a loss for 15 years just to turn taxi driving into indentured servitude. Mark Zuckerberg was obsessed with VR because his primary competitors owned a hardware platform (so Google owns Android, Apple Iphone etc) and he needed FB to have one too. Being a tech bro, the reason he pitched as meeting software was to undermine commercial real estate. NFTs were an attempt to disrupt central banking. AI is an attempt by Silicon Valley to cut highly paid tech workers from the payroll.

Sorry, this post got away from me. This is the part that has a bubble timer on it, I believe. LLMs produce garbage code, and garbage art. It has inflicted immediate, incalculable harm on people real lives. Eventually, I believe (if the current world order survives anyway) lawmakers will clamp down on it.

I don't think the VCs care much about the infinite incalculable loss. But there's this idea that (I think) Robert Evans introduced me to. He noted that Fascists love the infinite lie machine. Fascist governments classically controlled the media, and Russia has demonstrated what a wonderful weapon of war LLMs really are. That alone terrifies me. Its worth something to the worst people, and who knows what might happen if say, Peter Thiel wants to continue underwriting it so that way he can, say, direct fascist uprisings against governments that try to regulate him, or I don't know, portray striking workers as domestic terrorists.

AI is improving at the same rate it always has

This is blatantly false, there's a reason there is talk of the cold winter of AI, and the long walk in the desert.
The desert walk was at least 2 decades of very little progress despite big investments, the cold winter was another decade without much progress because of disillusionment so nobody wanted to invest in it.
AI has progressed more for the past 10-15 years than it did for 40 years from about the 70's to about 2010.

For actual AI, nothing has changed.

I assume you mean strong or general AI, and that's not what we are debating here, because that is not a reality yet.

Something that doesn't exist obviously can't burst.

I agree 100%, and yes there are always losers due to competition and failure to compete. Just like some companies are losing money making cars, but cars are obviously not a bubble.
Maybe the most of the hype is over for now among investors, but AFAIK Nvidia who is the biggest supplier of hardware for it, is not seeing a slowdown in demand.
So maybe it's more a slow down of interest in the media?