Gavin Baker: Market Pullbacks Blur Accelerating AI Fundamentals and Compute Economics

Despite a sharp market pull-back that sent many AI-related equities down 40% to 60% and brought Nvidia to its lowest forward price-to-earnings multiple in years, investor Gavin Baker contends that underlying market fundamentals are actually strengthening. Speaking on recent industry trends, Baker notes that key quantitative metrics tracked on the ground—including token generation growth, DRAM spot pricing, and GPU spot rental rates—show accelerating demand rather than deceleration. He points out that public market skepticism stems partly from a lack of visibility into private model developers like OpenAI and Anthropic, alongside rapidly expanding open-source inference clouds.

A central component of Baker's thesis is that the market is misjudging how the next phase of the AI compute buildout will be financed. While widening credit default swaps and rising real yields have stoked fears of a risky debt-fueled capital cycle, operating cash flows for major hyperscalers including Microsoft, Meta, and Amazon accelerated significantly last quarter. Furthermore, as legacy long-term compute contracts roll off and reprice at higher current spot market rates, hyperscaler cash flows are poised to increase further. This dynamics could allow major technology firms to fund upcoming deployments of Nvidia's Blackwell and Rubin chips largely through internal operating cash flow rather than relying heavily on debt.

Baker also reframes the rise of competitive open-source models, arguing that open-source adoption does not diminish overall hardware demand. Instead, open-source model usage shifts profit margins away from the proprietary frontier model layer and into the underlying cloud infrastructure layer. Because generating an open-source token requires the same computational flops, memory, and energy as a proprietary token, lower inference costs drive token elasticity, ultimately increasing total GPU compute hours consumed across the market.