Nvidia: AI infrastructure is a durable asset that can hold its value
During an interview with Yahoo Finance during GTC Berlin, Nvidia CEO Jensen Huang said the company’s AI computing systems are assets that are productive, fungible and durable, with a useful life that can be extended to 8 years or longer, rather than the 4-year cycle that investors and lenders typically assume.
Huang said these systems already have really, really high residual value, and that lenders may in the future be able to use them as collateral. He added that underwriting Nvidia's assets... is the lowest of any asset out there. Nvidia’s official blog also described AI factories as durable, saying that GPUs and systems can continue generating revenue for years after delivery, extending their effective useful life.
The Nvidia AI Infrastructure account recently posted that data show NVIDIA AI infrastructure is a durable, investable asset that can hold its value, with new applications and workloads continuously entering its interchangeable platform.
Original sources:
Title: Productive, Durable, Fungible: How NVIDIA AI Factories Maximize Return on Investment Publisher: NVIDIA Blog (Shruti Koparkar) Publication date: 2026-10-01 URL: https://blogs.nvidia.com/blog/productive-durable-fungible-ai-factories/ Source type: official company blog Verification: full text read in this session, verification date 2026-10-25 Supporting content: defines AI factories as durable and explains that GPUs and systems continue generating revenue for years after delivery
Title: Jensen Huang Says Nvidia GPUs Could Last Over 8 Years, Not 4, Carry ‘Really High’ Residual Value Publisher: Yahoo Finance / Benzinga (Namrata Sen) Publication date: 2026-10-24 URL: https://finance.yahoo.com/technology/ai/articles/jensen-huang-says-nvidia-gpus-113005747.html Source type: media report (interview citation) Verification: full text read in this session Supporting content: Huang directly used the terms productive, fungible and durable, and the view that the useful life is over 8 years and the lending risk is minimal
ABAB AI Insight
Nvidia is repositioning AI computing hardware from traditional capital expenditure into a financeable, collateralizable infrastructure asset. Huang’s remarks at GTC Berlin are consistent with the company’s early-October blog post, emphasizing that systems can still serve different workloads through successive rounds of software optimization, thereby sustaining residual value and leasing demand.
This claim directly addresses market concerns about rapid GPU depreciation. Traditional data center equipment is often depreciated over 4 to 6 years, while Nvidia cites cases such as A100 units still in commercial use and CoreWeave extending 2020 equipment bookings to 2029 to argue that the actual economic life is much longer. If lenders recognize residual value and interchangeability, financing costs could fall, supporting larger amounts of third-party capital entering AI data center construction.
Nvidia has previously worked with institutions including Apollo, BlackRock and Blackstone to channel more than $500 billion in third-party capital into AI infrastructure. Positioning GPUs as the lowest-risk collateral is a key part of this financing structure. If actual residual value falls short of expectations, some of the company’s guarantee arrangements could add pressure to its balance sheet; if the leasing market and software optimization continue to support utilization of older-generation equipment, capital recycling efficiency will be higher than in a traditional semiconductor cycle. Key indicators to watch are the stability of older-generation GPU lease prices and the actual credit allocation ratio lenders apply to residual value.
ABAB News · Cognitive Law
GPU useful life extending from 4 years to more than 8 years depends on software optimization and interchangeability across multiple workloads.
High residual value shifts AI hardware from a depreciating asset to financeable infrastructure that can be pledged as collateral.
The positioning of the asset as the lowest lending risk depends on sustained demand across customers and generations rather than the performance peak of a single generation.