Meta's AI Spending Cuts Free Cash Flow by 91%

That 91% quarterly free cash flow drop is brutal, even by Big Tech standards. Meta's funneling billions into AI infrastructure—GPU clusters, data centers, cooling—and the short-term hit is undeniable. For context, their free cash flow went from robust to nearly wiped out in a single quarter. It's not just training costs; inference at scale burns through cash fast, especially with models like Llama being open-sourced.

What stands out is how this compares to Microsoft or Google. Both invest heavily in AI, but their diverse revenue streams (cloud, enterprise, ads beyond one platform) cushion the blow. Meta leans almost entirely on advertising, making this a concentrated bet. If AI-enhanced recommendation systems or metaverse features don't boost revenue soon, this becomes a risky squeeze from a financial standpoint.

From a practical engineering angle, Meta's likely running thousands of H100s at peak utilization. The energy and cooling costs alone are staggering—think data centers in places with high electricity prices. They're probably optimizing for throughput over latency, given the model sizes. For developers using Meta's open-source releases, this is a double-edged sword: more capable models now, but potential service cuts if cash flow tightens further.

I'm watching to see if this spending pace holds. If free cash flow stays depressed for another quarter, expect belt-tightening in other divisions. For the AI community, it's a reminder that even the giants face trade-offs between innovation and financial discipline.