Meta's AI Isn't a Flop — the Numbers Tell a Different Story

Futurism's recent "almost nothing" take on Meta's AI output is a classic case of looking at the vibe instead of the spreadsheet. Meta just guided to $60–65 billion in capital expenditures for 2024, mostly on AI compute — that's a bet the size of a small country's GDP. You don't write checks like that and have nothing to show for it. The real question is whether the numbers that matter are moving.

Start with Llama. The open-source model family has become the default base for a massive chunk of the enterprise LLM world, not just in research but in production. Hugging Face download counts for Llama variants are in the tens of millions, and companies like Amazon, IBM, and Microsoft have all put Llama-based services on their cloud platforms. Even if Meta never directly charges for Llama, the strategic value is that it sets the cost baseline for everyone else — including OpenAI and Google. That's a form of leverage that won't show up on a P&L but absolutely shows up in the competitive landscape.

Then look at the actual business numbers. Meta's advertising revenue has grown double-digits in recent quarters, and much of that lift is credited to AI-powered recommendation engines and ad targeting. The company has talked about generative AI improving conversion rates and driving more value for advertisers, which is precisely how you monetize a hundred-billion-parameter model without selling API credits. Wall Street's reaction — the stock has been on a tear since early 2023 — suggests investors see the ROI, even if Futurism doesn't.

The consumer-facing side is admittedly messier. Meta AI in WhatsApp and Instagram is still not a ChatGPT-killer, and the smart glasses are a niche product. But "not winning the consumer AI arms race" is different from "nothing to show." The Ray-Ban Meta glasses actually outsold initial expectations, and the AI assistant now handles billions of queries per month. That's not a flop; that's a toe in the water while the real bet is placed on infrastructure and open-source ecosystem lock-in.

Where Futurism has a point is that Meta still lacks a single signature AI moment that captures the public imagination like GPT-4 or Gemini did. The company is playing a longer game, building the plumbing rather than chasing headlines. But if you judge AI progress by deployment, downloads, and revenue contribution — the numbers that actually drive a business — Meta has quite a lot to show. The "nothing" narrative only works if you ignore every metric except buzz.