Nvidia’s Huang Faced GPU Glut Crisis Before Resignation

Nvidia's former chief executive Jensen Huang privately acknowledged that the company was drowning in excess GPU inventory and memory commitments, newly revealed details show.

According to reporting, Huang confessed in August that major cloud computing firms were abandoning Nvidia in favor of developing their own custom chips, leaving the company saddled with prepaid silicon orders from Taiwan Semiconductor Manufacturing Company that were accumulating like an "avalanche." The disclosure came during a tense meeting with Nvidia's chief financial officer Colette Kress, who has since assumed the role of interim chief executive.

Kress had grown alarmed about Nvidia's mounting financial obligations and sought to curtail the company's practice of financing chip purchases for cloud providers and artificial intelligence companies including OpenAI and Anthropic. While initial discussions centered on backing roughly 8 gigawatts of computing capacity valued at approximately $250 billion, Kress managed to reduce Nvidia's guaranteed commitment to approximately 4.25 gigawatts, or up to $125 billion.

During the August confrontation, Huang dismissed Kress's concerns about financial exposure, using derisive language to characterize her position. However, he then detailed the predicament facing Nvidia: major cloud operators were reducing their dependence on the company and pursuing alternative chip suppliers. Simultaneously, the massive volume of wafer capacity and memory components Nvidia had committed to purchasing were arriving in overwhelming quantities.

The revelation validates long-standing concerns from industry observers and hardware enthusiasts. Rather than allowing excess chip production to flow into consumer graphics cards at reasonable prices or broader memory markets, Huang allegedly mobilized Nvidia's balance sheet to finance data center construction projects. This strategy enabled artificial intelligence startups and major cloud companies to absorb Nvidia's excess inventory.

While reports of unsold hardware stockpiles have circulated for months, mainstream attention to Nvidia's situation has only recently intensified. Industry analysts previously documented how Nvidia and peer companies exposed themselves to substantial financial risk through aggressive ordering practices that capitalized on widespread fear of missing out among major technology firms.

Estimates suggest approximately $200 billion worth of Nvidia artificial intelligence chips remain idle in warehouses, purchased by cloud operators lacking sufficient electrical infrastructure or immediate customer demand to deploy them. This accumulation underscores a fundamental mismatch: actual artificial intelligence demand falls far short of the silicon volume that major chip buyers and manufacturers committed to purchasing.