Nvidia’s Next-Gen AI Chip to Cost More Than $30,000

nvidia next-gen ai chip cost

Nvidia announced its next-gen AI chip, a device that’s set to cost as much as a whopping $40,000.

Have you ever wondered how much the GPU needed to power training and deploying AI might cost? Most of these AI chips are Nvidia GPUs, and the company doesn’t typically disclose the list price. However, the company’s CEO did go so far as to say Blackwell, Nvidia’s next-generation graphics processor with an AI chip, will cost more than $30,000.

Powering Artificial Intelligence Is Pricey

Right now, most projects like ChatGPT, Google Gemini, and Microsoft Copilot use the Nvidia H100, or the “Hopper” generation of AI chips. Even though Nvidia won’t list a price for these, analysts have estimated the chips to cost between $25,000 and $40,000.

Roughly every two years, Nvidia releases a new generation of AI chips. These are typically faster and more energy efficient. That means ChatGPT and its cousins can work harder, process more information, and just seem zippier.

This development is expensive. Speaking to CNBC’s Jim Cramer, Nvidia CEO Jensen Huang said research and development costs for the Blackwell AI chips ran to the tune of $10 billion. Naturally, the graphics chipset maker wants to earn that money back, plus some.

While Nvidia does use the buzz around its AI chip announcements to drive its consumer graphics card sales, it does need to recoup R&D costs through sales of the more expensive products.

To that end, Huang showed Cramer the Blackwell chip and said, “This will cost $30 to $40 thousand dollars.”

Who Would Spend $30,000 on an AI Chip?

If you think that sounds like an exorbitant amount to pay for a chip, know this: the top players in AI are already spending an estimated $25,000 to $40,000 for the Hopper generation of AI chips. In fact, Meta announced in early 2024 that it was buying hundreds of thousands of Nvidia H100 GPUs.

Nvidia has so far announced three different versions of the next-gen Blackwell AI chip. It unveiled a B100, B200, and a GB200. The GB200 configuration pairs two Blackwell GPUs with an Arm-based CPU. Each version comes with slightly different memory configurations.

The next-generation of AI chips is expected to ship during the second half of 2024.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *


To Top