Meta Has Been Busy

**Tyler Crowe** (0:02)
Meta is making even bigger AI bets today on Motley Fool Hidden Gems Investing.
Welcome to Motley Fool Hidden Gems Investing. I'm your host, Tyler Crowe, and today I'm joined by long-time contributors, Matt Frankel and Jon Quast. Today, we're going to dive into a couple mailbag questions. We got questions about GE Vernova, we got questions about REITs, which I think definitely had Matt excited to talk about today. But we want to start today with two relatively large announcements from meta-platforms today, all of which related to AI. One of them was the launch of its Muse Spark 1.1 Artificial Intelligence model, kind of the next iteration of what they've been doing with AI models. And the second was it announced that its plans to put its own AI chip into production is going to start in September with both Broadcom and Taiwan Semi as kind of designers, manufacturers helping them build out their own chip production capacities. Now guys, these sound like really big moves. So help me wrap some context about what we're seeing here with these.

**Jon Quast** (1:05)
Yeah, as far as the model goes from Metaplatforms, this is actually a pretty big upgrade in a couple of areas. You wouldn't normally be inclined to think so. I was tempted to overlook this just going from 1 to 1.1 here with Muse Spark.
But in here that I think is significant, one is the context window. So they're going to provide a 1 million token context window. What this does is it allows an AI agent to essentially work longer on a task without forgetting what it's doing. That's actually a problem with some models out there. You'll send an AI agent to work, it loses context, it forgets what it's doing, it keeps working and keeps spending your money. That's a problem. So a 1 million token context window, this is roughly four times as big as the 1 version of MuSpark. So that's a really significant upgrade. The other big change here that I'm seeing here is this is now being launched to people to use. There's pricing to go with this. And if you look at the pricing, it's more than 50% times cheaper than competitive products from Anthropic and OpenAI.
It's both for the input and the output.
That is really significant when you think about these two businesses because Anthropic and OpenAI, they kind of need the products that they have out there. That's what they do. Meta has a whole other business paying the bills and it does this on the side. So it has the luxury of this aggressive pricing. And so that is something really significant to note with MuSpark 1.1.

**Matt Frankel** (2:49)
So with the chips, I'm not sure if it's as significant as the model. So this is essentially what Google and Amazon do.
These chips, as Tyler mentioned, they're designed with the help of Broadcom and manufactured by Taiwan Semi. This is the essential model that Apple uses to design its own iPhone chips. And the goal here is to really reduce the company's dependence on NVIDIA and AMD processors that are really expensive as the company aims to build out its compute power and double again next year. And really, the idea is that these chips are going to handle the easier side of AI tasks. They're still going to need the more powerful NVIDIA ones for that. Yeah.

**Tyler Crowe** (3:24)
I want to put this in kind of context of everything we've seen from Meta recently, because this isn't what, to me, seems like the biggest major announcement or it's a couple in like what I would say is major announcements from Meta. It's been using a lot of creative financing to build data centers.
To your point, the amount that they're adding this year, doubling that next year, it's made some announcements with smart glasses. It recently announced a prediction market asset. I know I'm missing a few deals and announcement there, but I think it gets to the broader point. There's a lot of things going on here. When I look at Meta and I see all of these things that it's doing, I am less impressed. I see an unfocused company that's throwing a lot of spaghetti at the wall to see what sticks. The company seems to be all in with these new ideas that end up maybe not doing as much, and it all reverts back to the basic advertising model that they've had for so long, with Facebook, Instagram, what have you. Now, I brought this up before, but the company really dodged a bullet, I would say, with all of that investment in data centers for its reality labs, virtual reality efforts, being able to basically pivot quickly to AI and be like, yeah, we don't have to write this down because now it's all AI stuff. So here's my broader question, putting all that in context. Should investors be excited about these new moves and things that they're doing? Because to me, I just see an undisciplined company trying to look like something that it isn't.