Jeff Bezos-Backed Tenstorrent Targets Nvidia With Affordable AI Chips

Key Takeaways

• Through his venture capital firm, Jeff Bezos has invested in Tenstorrent.
• The AI chip startup is focused on developing more affordable solutions.
• In a market dominated by Nvidia, Tenstorrent offers an alternative.

Amazon founder Jeff Bezos has invested in the up-and-coming AI hardware company Tenstorrent, contributing to a $700 million funding round that values the startup at around $2.6 billion, Bloomberg reported on Monday, Dec. 2.

In a market dominated by Nvidia, Tenstorrent stands out for its focus on cheaper, more cost-effective AI accelerators.

Tenstorrent’s Latest Funding Round

Since launching its first hardware products in 2021, Tenstorrent has emerged as one of the leading contenders to carve away some of Nvidia’s dominance in the AI chip market.

The company’s latest funding round , led by South Korea’s AFW Partners and Samsung Securities, suggests strong investor confidence in its mission.

But catching up with the high performance of Nvidia’s GPUs may be too much of a challenge, hence Tenstorrent’s focus on mid-range solutions at a more affordable price range.

An Alternative to $70,000 Nvidia Chips

Whether customers require individual processors or scalable data center solutions, Tenstorrent’s primary selling point is its focus on affordability.

According to the website, Individual “Wormhole” chips can be picked up for as little as $999, while larger server racks containing multiple chips are comparably affordable.

In contrast, Nvidia’s high-end Blackwell GPUs cost as much as $70,000 a piece, and full servers can run into the millions.

Tenstorrent claims its systems have “superior performance density for cost,” i.e., they offer users more bang for their buck. But the company isn’t just vying for a slice of the booming AI data center market.

Desktop Processing

While Nvidia has doubled down on large-scale data centers and cloud provisioning, Tenstorrent sees an opportunity in desktop AI.

For customers who don’t want to rely on cloud processing but don’t require an Nvidia setup of their own, Tenstorrent’s AI workstations sit somewhere between ordinary home computers and industrial GPU superclusters.

Crucially, these machines benefit from the chip’s lower power consumption than traditional GPUs, making them appropriate for small-scale researchers and AI developers.