Deep Dive: How Meta Plans to Make Money from AI

**Zaid** (0:00)
Welcome back to The Rundown for another weekend deep dive. Today, we are talking about Meta and their new plan to actually make money from AI. Meta has been under pressure from investors to justify the billions of dollars in AI capex spending. And this week, they showed off what they planned to do. Meta is getting into the cloud business to rent out AI computing power. They're also jumping into the AI coding market to go after Anthropic and OpenAI. And they also ship their first ever image model. So in today's episode, we'll dig into the latest AI pivot coming from Meta, and whether it'll be enough to convince Wall Street that Zuck isn't just lighting hundreds of billions of dollars on fire again. We got a great one for you today.
Let's dive in.
Before we get into Meta's new AI monetization strategy, you have to understand why Meta is spending hundreds of billions of dollars on AI in the first place. See, throughout Meta's history, there have been a few make or break moments for the company. There was the shift from desktop computers to smartphones, then TikTok came along and threatened to steal an entire generation of users. Both time Wall Street panicked, but both times Zuckerberg navigated the company through it and copied whatever feature he needed to copy. And Meta honestly came out stronger on the other side. And AI seems to be the latest make or break moment for the company. So the thing is, Meta still makes 98% of its revenue from ads shown to 3-plus billion people that use Instagram and Facebook every day. And AI is really good at predictions. So having smarter AI models will help Meta recommend better content to users. Smarter models mean that you scroll longer. Longer scrolling sessions means you're gonna see more ads. Better ad targeting means advertisers will pay more for those ads. So all of that is helping Meta make more money. And it's how they justify their huge capex. Meta is projected to spend up to $145 billion on capex this year, and they just keep raising that number every quarter. But here's the thing, that strategy from Meta is working, at least according to their earnings. Last quarter, Meta did $56.3 billion in revenue, which was up 33% from the same quarter last year. It's actually the fastest year over year growth the company has seen since 2021 The company said that ad impressions were up 19%, and that the average price per ad was up 12%.
So what that means is that Meta is stuffing their apps with more ads, but they're able to charge more money for those ads because their ad recommendation is getting better thanks to AI. But despite that, Wall Street still wasn't buying Meta's AI story. The stock was down more than 10% in the first half of the year, while the S&P 500 gained 9%. Now I think a part of the reason for that is that investors still had a bad taste in their mouth from watching Zuck light $87 billion on fire, chasing the Metaverse for the last few years. Fool me once, shame on you, fool me twice, you can't get fooled again. But look, it looks like Zuckerberg is finally reading the room now, because Meta just revealed a whole new set of ways to make money from AI, and the market so far is loving it. On July 1st, Bloomberg reported that Meta was building a cloud business called Meta Compute, and their plan was to sell their excess AI computing power to outside customers. Now, this was a big deal for many reasons, because it was Meta's first entrance into the cloud business. But it also opened up ways for Meta to make money from their AI infrastructure. The first way was by selling access to AI models hosted inside Meta's data centers. So a developer or a business could choose an AI model and send its request through Meta's platform and pay Meta based on how much they use. And through this platform, Meta could offer their own Muse AI models, or potentially down the road, they could also host models from other companies as well. The second and more simpler option though, is that Meta could just rent out the raw computing power from their AI data centers. You know, all the spare servers and GPUs they have sitting all over the country, they could sell that excess capacity. And this would put Meta in direct competition with Neo cloud companies like Coreweave and Nebius, whose entire business model is renting out AI compute. But I imagine at Meta's scale and with their resources, they could offer better pricing than what Nebius or Coreweave could offer. Now, when this news was announced back on July 1st, investors loved that Meta stock jumped 9% on the day, while Coreweave and Nebius stock absolutely tanked. But the thing is, Zuck has been hinting at Meta doing something like this in previous earnings calls. And it makes a lot of sense for them to do this because Meta has spent the last two years building massive AI data centers buying up every NVIDIA GPU that wasn't already nailed down. And with the demand for AI compute being so high right now, Meta can now rent out their extra capacity at a premium and earn a ROI on all that investment. This pivot by Meta shows Wall Street a clear, direct path for them to make actual revenue from their AI infrastructure spending. And look, if this playbook sounds familiar, it's exactly what XAI, AKA SpaceX, just did. XAI built way more data center capacity than GROK could ever use. And instead of eating that loss, Elon Musk started renting it out to companies like Anthropic and Google that will now pay XAI more than $2 billion a month combined. And by the way, Meta going in this direction doesn't necessarily mean that Meta is giving up on their own AI ambitions or admitting that they built too much capacity. It just gives Meta flexibility on ways to monetize their AI infrastructure. They can still train their own frontier models. They can keep improving the recommendation algorithms that power Instagram and Facebook. They can also generate and target more ads. But if there is compute left over, they now can just rent it out for a premium. There was a great piece from the semi-analysis sub stack on Meta's pivot, and the case they make is that this move by Meta actually encourages Meta to build even more infrastructure and not less. And by the way, Meta is not stopping there. They're also working on their custom AI chips as well to put in their data centers. Just this week, an internal memo was reviewed by Reuters which showed that Meta plans to put its own custom AI chips, code name Iris, into production by September. And having their own chip could mean less reliance on NVIDIA, which could mean cheaper computing costs for Meta and fatter profit margins on every GPU they rent out. So Meta is setting themselves up to be in a good spot with their cloud business. But renting out excess compute isn't the only new AI business Meta announced. They're also coming after OpenAI and Anthropic. This week, Meta announced that they were jumping into the AI coding market by launching MuSpark 1.1, which they claim is their strongest AI model yet for coding and agentic tasks. Right now, the AI coding market is dominated by Anthropic and OpenAI. Anthropic is getting a ton of fans with Cloud Code, and OpenAI has Codex. But Meta thinks that they can compete in that arena and take market share, not just because of the capabilities of their new model, but also because of their pricing strategy. Meta is charging a quarter of what Anthropic charges for their top model. So Zuck is betting that by dramatically undercutting Anthropic and OpenAI when it comes to pricing, developers will at least give Meta's new model a shot. And you can tell how much this launch mattered to Zuck because Zuck personally posted the announcement on X, since most of the AI industry is addicted to that platform. This was actually Zuck's first post on X in three years. And of course, Elon Musk had to reply, he replied with the word jinx because XAI had launched their own coding model literally the day before. So Meta is hoping to gain some market share with their new Muse Spark model, but that wasn't the only model they launched this week. To me, the more interesting launch was Meta's release of Muse Image. It's Meta's first in-house model specifically built for generating and editing images. Muse Image will be built directly into Instagram, WhatsApp, and Facebook, and it'll allow users to edit pictures on those platforms using AI. To me, the more interesting business opportunity here is on the advertising side. Muse Image will be integrated into Meta's advertising platform, which will allow businesses to generate images, change styles, and create multiple versions of ads automatically. So imagine if you're a small business advertising on Meta, you could potentially upload a product photo, and Meta's AI could generate multiple variations of ads for you from just that one product photo. So this could be a game changer for small business advertisers, which would mean more businesses spending more money on Meta's ad platform. Meta says they plan to launch a video model called Muse Video pretty soon. So this will only add to Meta's flywheel, because Meta's AI will be making the ads, targeting the ads, and serving the ads. And all of that should lead to more ad revenue. So I guess the question now is that will all these moves by Meta over the last couple of weeks finally get Wall Street off their backs? Well, here's my take.