$BULLA$NVDA The market’s first reaction to “Anthropic generates $6.6 million in revenue per employee, second only to Nvidia” is: the AI application layer has finally set a new efficiency benchmark; there are winners beyond Nvidia, and some are even starting to talk about “compute becoming commoditized, with models diverting the hardware premium.” But I think this actually confirms Nvidia’s moat rather than threatening it. High revenue per employee essentially means Anthropic is using very few people to generate enormous API revenue. Behind every dollar of revenue supporting its inference workloads is a compute cost paid to $NVDA (and cloud providers). Its “high efficiency” comes from an asset-light model that outsources capital expenditure to compute suppliers. Consider the historical precedent: when SaaS companies were generating high revenue per employee in the 2010s, AWS’s capital spending accelerated instead. The boom in asset-light application layers fueled demand for the capital-intensive infrastructure layer. Looking at options data, $NVDA’s near-term implied volatility hasn’t risen on this kind of “diversion” narrative, and skew still favors calls—suggesting that smart money isn’t pricing in “Nvidia being replaced.” On-chain and futures markets also show no sustained inflows into decentralized compute tokens that would signal a shift away from traditional providers. What’s really worth watching is this: Anthropic’s high revenue per employee means it relies more heavily on continued fundraising to buy compute. If fundraising slows, the visibility of $NVDA’s orders could prove more fragile than the market expects—but that’s a demand-side risk, not a loss of market share. Alongside today’s convergence of opportunities in Eaton’s data center business and Amphenol’s AI infrastructure, the infrastructure chain is still expanding, not contracting. What do you think? Different views are welcome in the comments.