inferwire
/
Hardware·3 min read

Nvidia Allegedly Funded Startups to Buy Its Own GPUs

A Wall Street Journal report alleges Nvidia used circular financing to mask GPU oversupply from TSMC and maintain high demand perception.

TL;DR

  • A report alleges Nvidia funded AI startups to buy its GPUs and absorb oversupplied chips from TSMC [^1].
  • Internal friction grew when CFO Colette Kress opposed circular financing and a $250 billion OpenAI deal [^1].

Background

Semiconductor manufacturers must commit to huge silicon wafer orders months in advance to secure factory capacity. When demand shifts or major cloud providers build custom chips instead of buying standard GPUs, hardware vendors risk holding billions of dollars in excess inventory. To prevent sudden revenue drops, companies navigate complex financial arrangements with ecosystem partners to maintain sales volume and supply chain commitments.

What happened

Following the departure of former Nvidia CEO Jensen Huang, a report from The Wall Street Journal alleged that Nvidia engineered GPU and memory scarcity by funding AI startups to buy its chips and construct data centers [^1]. The reported scheme aimed to maintain the perception of strong market demand while offloading an oversupply of semiconductors manufactured by TSMC [^1].

During an August meeting with senior finance leaders, CFO and now acting CEO Colette Kress pushed to reduce these circular financing agreements [^1]. In response, Huang reportedly berated Kress, calling her a "deer in the headlights" after she argued to scale back Nvidia's $250 billion deal with OpenAI [^1]. Nvidia's push into small-scale cloud providers came as major AI hyperscalers began moving away from off-the-shelf components to develop custom solutions [^1].

The strategy left Nvidia committed to an "avalanche" of GPUs from TSMC without immediate organic buyers [^1]. Kress noted that many of the small cloud providers receiving Nvidia's investments could not actually afford the GPUs, creating a risk of financial dead ends for the company [^1].

Why it matters

Circular financing risks obscuring true market demand for foundational AI hardware. By using balance sheet capital to fund buyers, a chip maker can create a temporary illusion of persistent supply shortages while absorbing fixed manufacturing obligations [^1]. When major hyperscalers shift toward custom silicon, relying on under-capitalized startups to buy excess inventory exposes the vendor to significant default risks [^1].

The report also highlights severe governance friction during executive shifts. Internal disputes over a $250 billion commitment and inventory exposure demonstrate how strategic bets on startup infrastructure can strain corporate finance controls when market demand cools [^1].

Practical example

Say David manages infrastructure planning at a mid-sized cloud hosting provider.

His team notices that top-tier GPUs remain scarce and expensive on the open market, even as major tech conglomerates quietly transition to in-house custom processors [^1]. Seeking to expand, David evaluates vendor-backed financing programs offered by a primary chip designer looking to fund small-scale cloud buildouts [^1].

Under the arrangement, David's company receives venture backing tied to purchasing commitments for high-end accelerator chips [^1]. However, as customer demand for standard off-the-shelf GPUs fluctuates, his firm struggles to generate enough revenue to cover the hardware costs. Just as internal pushback mounts at the chip vendor over circular financing risk [^1], David realizes his hosting business is overleveraged on inventory that primary hyperscalers no longer need [^1].

Instead of scaling operations organically, David's firm ends up holding expensive data center capacity while the semiconductor manufacturer faces potential financial dead ends from funded buyers who cannot afford their hardware [^1].

Sources

  1. [1]TechPowerUp — NVIDIA Allegedly Engineered Scarcity By Funding AI Startups To Buy Its GPUs
  2. [2]Hacker News — Nvidia allegedly engineered scarcity by funding AI startups to buy its GPUs