AIbaseUpdated

Amazon plans to buy Decart for $7 billion to acquire the AI chip “general adapter”

According to the Wall Street Journal, Amazon is in talks with AI startup Decart for a acquisition, with a potential valuation of around $7 billion (about 46.9 billion yuan). If the negotiations go smoothly and no new…

According to a source familiar with the matter cited by The Wall Street Journal, Amazon is in talks with artificial intelligence startup Decart regarding a potential acquisition deal worth approximately $7 billion (about 46.9 billion yuan). It is said that if the negotiations proceed smoothly and no new bidder emerges along the way, the deal could be finalized soon. In an era of chip shortages and a surge in self-developed technologies, this transaction also reflects the giants’ anxiety over control over computing power.

Impressed by this unique skill of cross-chip adaptation

Decart is headquartered in San Francisco and has a valuable skill: developing cross-chip compatibility software that allows developers to switch seamlessly between different AI chips. As major tech giants fight hard for computing power, the value of this underlying “translation layer” is rapidly increasing. Those who can flexibly move models around won’t be tied to a single chip supplier. For Amazon, which owns AWS, having this scheduling capability also gives its cloud customers an extra advantage.

Once targeted by Anthropic but abandoned

The value of this company did not arise out of thin air. According to PitchBook data, Decart has previously raised over $400 million from well-known venture capital firms such as Sequoia Capital, Benchmark, and Radical Ventures. During the funding round in May this year, its valuation approached $4 billion. More interestingly, Bloomberg reported in August that Anthropic once wanted to acquire it for around $6 billion, but in September, there were reports that the AI lab had given up that idea. Now, with Amazon taking over, there are undoubtedly more uncertainties regarding Decart’s ownership.


Original source

AIbase

Content notes

Original publication and rights belong to the source.

Machine translation · Refer to the original