How Nvidia’s AI Chips Are Proving Doubters Wrong

NVIDIA’s CEO Jensen Huang displays products onstage during the annual Nvidia GTC Artificial Intelligence Conference at SAP Center in San Jose, Calif., in March. (AFP via Getty Images)

Key Points

• Nvidia stock is approaching its all-time closing high of $235.74.
• Skeptics argued that Big Tech companies underestimated expenses by using an unrealistic six-year depreciation schedule for Nvidia chips.
• Silicon Data reported that Nvidia’s A100 chips stopped depreciating in late 2025 as rising rental income offset the time decay of value.

Nvidia is on a roll. The chip maker’s stock looks set for an assault on record highs and it has a lot to do with the surprising durability of its hardware.

Nvidia shares were up 0.3% at $231.12 in premarket trading. It’s not far off its all-time closing high of $235.74, reached on May 14.

The semiconductor giant has been knocking down various arguments made by skeptics in recent months, from its growth prospects to worries about circular financing. Now it is looking to disprove concerns about how long its artificial-intelligence chips hold their value.

One major argument made by AI skeptics—including Big Short investor Michael Burry—was that Big Tech companies overstate the long-term value of chips bought from Nvidia by using a six-year depreciation schedule. That’s the time frame used to spread the cost of an asset over its estimated useful life and would mean Nvidia customers were systematically underestimating their annual expenses.

Now we are more than six years on from the release of Nvidia’s A100 chip in 2020, which became the workhorse of the AI boom when it was used to train and run OpenAI’s ChatGPT. And the evidence so far is that the critics were wrong. Not only are A100 chips still being used, their rental price has actually increased in recent months.

“Based on our residual fair value estimates, A100 stopped depreciating since late 2025 as rising rental income offset time decay of value,” said Silicon Data, a start-up that researches the market for computing power, in a recent analysis.

Loading tweet…

The key has been the explosion in demand for chips to power inference—producing output from AI models—in addition to the need for cutting-edge models for training, prolonging the use of older processors. Nvidia-backed cloud-computing provider CoreWeave revealed earlier this year that it signed a deal to rent out its A100s through 2029.

There’s no guarantee the current trends will last forever. A new generation of AI data centers or custom chips could render Nvidia’s older processors obsolete more rapidly. But so far, Nvidia has been proven right and its doubters wrong.

Write to Adam Clark at adam.clark@barrons.com

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8