Gavin Baker’s “Federal Reserve of AI” framing puts NVIDIA’s “dark GPU” idea in focus

NVIDIA (NASDAQ: NVDA) is increasingly discussed as the backbone of the artificial intelligence buildout, and on Friday a hedge-fund CIO gave the company a new metaphor. Gavin Baker, chief investment officer at Atreides Management, said NVIDIA could become “the federal reserve of AI,” arguing that Jensen Huang is repositioning the company’s chips into a broader system that supports the AI economy.

Baker’s comments, circulated through an interview summary posted by Yahoo Finance, also introduced the idea of “dark GPUs.” In plain terms, the phrase typically refers to AI processing capacity that is built, owned, or provisioned but not yet fully visible or utilized in the way traditional hardware demand is measured. Baker suggested that these “dark GPUs” could become the first practical test of NVIDIA’s central-utility role.

The reported exchange did not provide additional, verifiable program details such as which customers would participate, how capacity would be contracted, or what measurable milestones would determine success. It also did not outline a specific product name, service offering, or timeline that NVIDIA itself has publicly committed to in the account summarized by Yahoo Finance.

For investors and customers, the “federal reserve” framing matters because it implies a shift from one-off hardware sales toward sustained influence over the compute supply chain. NVIDIA is already widely associated with supplying AI accelerators and the surrounding software ecosystem, but Baker’s positioning suggests he sees NVIDIA moving toward a role in allocating or stabilizing AI compute availability, rather than just shipping chips.

Market participants will likely watch whether “dark GPU” capacity becomes a recurring operational theme that can be linked to observable spending patterns, procurement behavior, or usage metrics across data centers. If it does, the takeaway would be less about short-term demand spikes and more about the durability of AI infrastructure investment and the extent to which NVIDIA’s platform becomes embedded in long-duration capacity planning.

Still, much remains unclear from the available reporting. The Yahoo Finance summary and the surrounding description of Baker’s remarks did not include supporting figures, disclosures from NVIDIA, or references to official company statements tying “dark GPUs” to a named NVIDIA initiative. As a result, readers should treat the “first test” framing as a hypothesis tied to Baker’s interpretation, not as a disclosed corporate plan.