Nvidia Projects Longer Lifespans for Blackwell GPUs

Longer-lived hardware architectures could allow data centers to extend depreciation schedules through 2027 and beyond.

Updated on Oct. 21, 2026 in Semiconductors

Isometric editorial illustration of a modular server rack unit, using clean geometric forms and a muted color palette to represent data center infrastructure. — Nvidia’s revised hardware longevity projections allow data center operators to extend depreciation schedules, improving the financial viability of long-term infrastructure investments. AI Illustration. Upload story photo >

Live Poll

Does an extended lifespan for high-end technology make large business investments a safer bet today?

Nvidia CEO Jensen Huang announced that the Blackwell and Vera Rubin GPU architectures are expected to have a useful lifespan of seven to eight years. This projection significantly exceeds the four-year lifespan previously used in data center investment models.

Why it matters

Extending the expected useful life of hardware allows operators to spread capital costs over more years, improving the financial viability of infrastructure projects. This adjustment shifts the economic model for data center investments as firms look to secure better credit terms.

Nvidia projects seven to eight years of useful life for Blackwell and Vera Rubin, outstripping the standard five to six-year industry depreciation cycle. This shift impacts capital expense modeling for massive data center infrastructure.

The players

Nvidia

A developer of graphics processing units and data center AI infrastructure that leads the market in accelerated computing hardware.

Jensen Huang

The CEO of Nvidia who oversees the company's roadmap for Blackwell and Vera Rubin AI architectures.

CoreWeave

A specialized cloud service provider focused on large-scale GPU infrastructure, currently maintaining A100 deployments through 2029.

The details

Operators calculate annual hardware expenses by amortizing total costs over an expected useful life. By asserting that newer architectures possess structural longevity, Nvidia provides data center owners with a basis to reduce annual depreciation charges. This extended asset life also acts as collateral, potentially allowing firms to secure more favorable credit terms for infrastructure financing. Blackwell GPUs are currently in volume production, with the Vera Rubin architecture expected to ramp in the second half of 2026.

2020: Nvidia A100 GPUs launched.
2025: Blackwell GPUs entered volume production.
H2 2026: Vera Rubin architecture ramp.
October 21, 2026: Jensen Huang provided GPU lifespan projections.
2029: CoreWeave A100 contracts expire.

The Tech Race

Nvidia is attempting to recalibrate the financial baseline for AI data centers by exceeding the typical five to six-year depreciation schedule. This strategy positions their latest architectures as long-term capital assets rather than rapidly cycling consumables.

Data center operators and enterprise investors can expect an shift in capital expenditure planning as depreciation timelines expand. This change will first affect large-scale infrastructure providers like CoreWeave who are currently extending their hardware contracts.

The takeaway

Nvidia aims to normalize an eight-year hardware cycle to sustain the $1 trillion revenue growth projection for its current and future architectures. Investors should track future earnings calls for adjustments to data center depreciation metrics and infrastructure contract terms.