Lana

Thanks, Carl.

Next up...

Alphabet unveiled a new way to slim down the amount of memory that AI models use this week, and its got chip investors shook.

The company's new TurboQuant algorithm can run large language models using about six times less memory than before.

That's huge.

Giant data center operators have been gobbling up memory chips like nobody's business, and now they might be able to run on a lot fewer.

That could ease the whole memory chip supply crunch thing and, of course, downsize prices.

Unsurprisingly, investors sent shares in three big memory makers — Samsung, SK Hynix, and Micron — on a two-day slide after the news.

There's a divide in the memory chip market.

Those three core players — and their peers — are on one side, making the specialized high-bandwidth memory that sits at the heart of AI systems.

And despite this week's sell-off, their business model is somewhat safer.

AI still runs off their stuff after all.

On the other side are flash memory and storage makers like Kioxia and SanDisk, which aren't quite so fundamental and will likely face a much bleaker reality.

At first glance, TurboQuant sounds like bad news for memory makers of all kinds.

But that might not be the case.

It's known as the Jevons Paradox.

Basically, when something gets cheaper and easier to use, people tend to use more of it, to the extent that it could offset the negative effects of a lower price.

So if AI becomes more efficient, companies might deploy it more widely, build more tools, and serve more users.

And that could end up supporting, or even increasing, demand for chips over time.

That's it for today.