Back
DensityAI Nears $10B Valuation With Conditional AWS Chip Deal
SiTech AI Team2 წთ. საკითხავი

DensityAI Nears $10B Valuation With Conditional AWS Chip Deal

AI chip startup DensityAI is in late-stage talks to raise hundreds of millions of dollars at a valuation near $10 billion. Management told investors AWS will buy its chips if they hit performance targets.

DensityAI, an AI chip startup founded about a year ago by former leaders of Tesla's Dojo supercomputer program, is in the final stage of funding talks, The Information reported. The company plans to raise several hundred million dollars in a round that would value it at $10 billion. Two people familiar with the matter said that is a high valuation for a business whose chips are still in early development and may be years away from mass production.

A conditional agreement with AWS

Management told potential investors it already holds an agreement under which Amazon Web Services will buy the chips if they reach specified performance benchmarks. One source said Andreessen Horowitz (a16z) is in talks to lead the round. Amazon and a16z representatives declined to comment, and DensityAI did not respond to questions.

A team built at Tesla's Dojo

The startup was founded by Ganesh Venkataramanan, the former head of Tesla's Dojo team, and former Tesla employees Bill Chang and Ben Floering. Roughly 20 former members of the Dojo project now work at DensityAI, including senior-level specialists.

Dojo was Tesla's in-house program for the chips and servers meant to train the neural networks behind autonomous driving and humanoid robots. The company disbanded the team last summer.

3D DRAM stacking and its risks

DensityAI tells potential investors that its chip uses an unusual memory layout to raise computing speed and cut energy use. Memory layers would be stacked directly on top of compute cores instead of beside them, shortening the distance data travels during inference.

The approach carries a clear drawback: stacking temperature-sensitive DRAM above hot compute units is extremely difficult to engineer, and the fragile layers deform or break easily.

Market context

d-Matrix already uses 3D DRAM stacking in its second-generation chips. It raised $275 million at a valuation of about $2 billion and is negotiating a new round, while NVIDIA is also exploring new memory technologies, including work with SK Hynix.

Investors keep funding compute startups despite immature products. Etched, founded four years ago, raised $700 million in August at a $21 billion valuation in a round led by Jane Street. Amazon builds its own AI chips, including Trainium, which targets better price-performance than NVIDIA GPUs, yet it still signs deals with chip startups.

SSiTech

SiTech — AI-powered web development

We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.