NVIDIA Stock Falls Despite Record Revenue

NVIDIA's record-breaking quarter saw data center revenue, its main growth engine, climb 75% year-over-year to $62.3 billion, accounting for over 91% of total sales. The company's forecast for the next quarter is approximately $78 billion, significantly higher than the $72.6 billion analysts had anticipated. The stock's dip reflects Wall Street's shift in focus from the AI training boom to the long-term economics of AI inference. Training a model is a massive, one-time computational expense, but inference—using the model to generate responses—is a continuous, high-volume cost. Analysts project the AI inference market could eventually be ten times the size of the training market. This market transition alters the hardware landscape, favoring efficiency, low latency, and cost-per-query over the raw power of training-focused GPUs. This opens the door for competitors, including AMD's MI300 series accelerators and custom silicon from major tech companies, to capture a share of the growing inference market. For Apple, this trend aligns with its long-term strategy of vertical integration and reducing reliance on single suppliers. The company is already developing its own AI accelerators, internally codenamed Baltra, to move away from NVIDIA's hardware. This follows Apple's successful transition from Intel to its own M-series silicon for Macs. Instead of using NVIDIA for its AI infrastructure, Apple has been leveraging Google's custom Tensor Processing Units (TPUs). This indicates a strategic decision to bypass NVIDIA's ecosystem, a move that could influence the development of AI-driven features within iOS and macOS by optimizing for custom hardware rather than NVIDIA's more generalized platforms.