Alex Mercer: Meta spent $14.3 billion to hire Alexandr Wang and build a proprietary AI model called Muse Spark—and now Mark Zuckerberg has to figure out how to sell it, a problem Wang was never hired to solve.
Jordan Hale: Wait — okay, $14.3 billion and nobody in that room was like, hey, who's actually going to take this thing to market?
Alex Mercer: I think that's basically the core tension here — Wang is a world-class model builder, that's what he was brought in to do, but what Meta now has on its hands is a commercialization problem, and those are two completely different skill sets.
Jordan Hale: What's fascinating is — and this is the part that got me — Zuckerberg has spent years positioning Meta as this open, developer-first AI company, and now he's got a proprietary model he has to sell, which is like, the exact opposite energy.
Alex Mercer: Right, and if you look at the data, Meta's previous AI wins have come from distribution — embedding models into apps people already use — not from convincing enterprises to pay for a standalone product.
Jordan Hale: So when you zoom out, this isn't really a story about one AI model — it's about whether Meta even knows how to be a company that sells AI versus one that just uses it.
Jordan Hale: Llama 4 dropped in April 2025 with benchmarks that — and this is Meta's own people saying this — were fudged. That's the word that came out internally. Fudged. And Zuckerberg's response was to spend $14.3 billion, acquire 49% of Scale AI, and bring Alexandr Wang in as the first Chief AI Officer in company history. That's a massive swing.
Alex Mercer: And Wang delivered. Roughly one year later — April 8, 2026 — Meta Superintelligence Labs ships Muse Spark. Internally codenamed Avocado. First proprietary foundation model Meta has ever released. That's a real engineering win.
Jordan Hale: In twelve months! Wang brought fifty researchers with him, built out MSL from scratch, and shipped a flagship model. I mean — that's kind of remarkable, right?
Alex Mercer: It is. I'm not dismissing that. But here's the problem — analysts are saying Muse Spark puts Meta 'back on the map,' and in the same breath saying Meta is still far behind OpenAI, Anthropic, and Google in actual market share. Both things are true simultaneously. Which means 'back on the map' is basically describing a position where you're still losing.
Jordan Hale: Okay but — wait, that's the thing though. Wang was hired to fix a credibility problem, not a market share problem. Those are different jobs.
Alex Mercer: Exactly. And that's why Zuckerberg's challenge now is genuinely harder than what Wang was asked to do. Wang's mandate was the engineering rebuild. Zuckerberg's mandate is commercialization. And no hire solves that.
Jordan Hale: Which is a wild position to be in. You spend $14 billion-plus, you get the model, you get the team — and you still have to figure out how to sell it. The selling problem didn't come with the acquisition.
Alex Mercer: And then look at what they're charging. $7.99 a month for Meta One Plus, $19.99 for Meta One Premium. The 2026 capex guidance is $125 to $145 billion. I keep coming back to this arithmetic. At $20 a month you'd need something like 500 million paying subscribers just to approach $120 billion in annual revenue. Meta has three billion users. What percentage do you think actually convert to paid?
Jordan Hale: No way it's close to the number you need.
Alex Mercer: Right. And that's the core tension. Meta's Q1 2026 revenue was $56 billion — up 33% year over year, 82% gross margins. The ad business is genuinely strong. But the stock is sitting around $570 and lagging every other tech megacap. The market's looking at $125 to $145 billion in capex and asking — what exactly is this paying for?
Jordan Hale: See, and this is where I get suspicious, because — the crazy part is — Meta Business Agent just launched, over a million businesses already using early versions on WhatsApp and Messenger. That's real traction. But then you remember Reality Labs is still losing four billion dollars a quarter.
Alex Mercer: Eighty-three billion cumulative since Q4 2020.
Jordan Hale: Eighty-three billion! And now the pitch is — hardware is the Trojan horse for AI subscriptions. The glasses, the devices, they deliver the software. But — wait, that's the exact same logic that hasn't worked for years. Why does it work now?
Alex Mercer: I don't know if it does. One analysis I saw put it pretty bluntly — an $8 subscription cannot plug a $19 billion hardware hole. That's not math that closes.
Jordan Hale: So then what is the $125 billion actually for? Because if it's not subscriptions and it's not hardware—
Alex Mercer: I think it's infrastructure optionality. Compute capacity, inference at scale, building the kind of foundational layer that gives Meta leverage five years from now — not two quarters from now. But that's a very different story than 'we're launching a $7.99 chatbot subscription.' Those aren't the same pitch.
Jordan Hale: And there's something else I keep thinking about. Meta spent a decade as the open-source champion. Llama was their identity — developers built entire ecosystems around those open weights. Muse Spark is proprietary. Closed. You access it through Meta's apps or you pay for it. That's not a business pivot. That's a personality change.
Alex Mercer: Maybe, but — the open-source strategy also never had a clear revenue path. Llama got Meta developer goodwill. Did it get Meta market share against OpenAI? Against Google's Gemini?
Jordan Hale: No, but it gave Meta something OpenAI doesn't have — a developer community that isn't paying them. And now they've walked away from that. So if Muse Spark doesn't actually close the gap with OpenAI and Anthropic, Meta just gave away their one differentiated thing and got nothing for it.
Alex Mercer: That's probably the real question investors are sitting with. Not whether Zuckerberg can sell Muse Spark subscriptions — but whether he can convince them to fund a multi-year infrastructure bet whose returns show up in competitive position, not in quarterly revenue. That is a fundamentally harder sell than a product launch.
Jordan Hale: Man, okay. When you zoom out, that's basically what this whole year has been — Meta asking investors to believe in a vision that keeps getting more expensive and keeps moving the goalposts. Like, Reality Labs was the long-term bet. Then the LLM pivot was the long-term bet. Now Muse Spark is the long-term bet. At some point you have to wonder if there's ever a short-term.
Alex Mercer: Yeah, and what matters here is that the market has actually rewarded Zuckerberg for that framing before — the efficiency year worked, the stock recovered. So the question I keep coming back to is: does he still have enough credibility to pull that move a third time? Or has the well run dry?
Jordan Hale: That's the one I genuinely don't know the answer to. Catch you next week.
