Did Nvidia Engineer the GPU Shortage? $250B, 0 Proof - shattered.io
Reports suggest Nvidia funded AI startups and cloud providers to purchase GPUs, potentially absorbing surplus chips from TSMC, but no evidence proves the company caused any shortage.
The core allegation is narrow. Reports claim Nvidia financed AI startups and smaller cloud providers to purchase GPUs and build data centers, potentially absorbing surplus data-center chips tied to TSMC supply commitments. The reported motive addresses a glut, not a drought, creating a mismatch with the headline question about an engineered shortage.
The reported shortage also extends beyond GPUs to memory components from SK hynix, Samsung, and Micron. No reviewed report explains how Nvidia could control these separate memory makers, making it difficult to attribute the entire shortage to one GPU designer.
The August Meeting
The reported account centers on an August meeting between Jensen Huang and Colette Kress, Nvidia's CFO. Kress reportedly argued for reducing circular-financing arrangements, where a supplier's money helps fund a customer's purchases from that same supplier.
deer in the headlights
The meeting is linked to a stated $250 billion deal with OpenAI, though this figure comes from the reported account rather than public filings. Nvidia reportedly faced an 'avalanche' of GPUs from TSMC without immediate buyers, according to the sources.
Financing Mechanics
The mechanics, as reported, involve Nvidia investing in young AI companies or smaller cloud operators, which then use that money to buy GPUs for data centers that rent compute to AI developers. While each step is legal in principle, the open questions concern scale, disclosure, and intent.
If funded buyers took delivery of surplus data-center GPUs, they would ease a glut but wouldn't necessarily drain the retail graphics card market, as data-center parts and consumer cards are separate products. No reviewed reporting shows that the financing diverted retail graphics cards or memory from consumers.
Memory Shortage
Memory tightness has documented drivers separate from Nvidia. For example, SK hynix locked up more than half of its 2027-28 HBM4 capacity through long-term supply deals with memory makers, which set their own production plans. While Nvidia influences memory demand through its orders, it doesn't control memory supply.
Related
