Pat Gelsinger, the former Intel chief who is now a General Partner at Playground Global, has invested in Delos Data, a company founded by former Intel veterans to address a costly problem inside artificial intelligence data centres.
The company raised $100 million to develop chips and software that improve the movement of data between processors. Its funding round included Matrix Partners, Playground, Socratic Partners, Capricorn, Matter Venture Partners, IAG and DYNAMIQ.
The investment is best read as a bet on a constraint within the AI build-out: larger clusters of processors need to communicate efficiently to train and operate AI models, and poor communication raises energy use and costs.
Dan Daly, Delos Data’s co-founder and chief technology officer, is leading the development of the company’s technology. He said its aim is to deliver faster, more flexible movement of data across AI infrastructure.
Delos is developing networking technology designed to work across different combinations of computing hardware, rather than being tied to one chip architecture. Nvidia remains the leading supplier of AI accelerators and networking technology, making flexibility in the wider infrastructure a significant commercial objective for new entrants.
Gelsinger joined Intel at 18 and later became its chief executive before moving into venture capital investment in deep tech. He has said that inefficient communication between chips can increase both energy consumption and costs.
The immediate funding for Delos sits alongside a more cautious view among investors, who have become increasingly concerned about the possibility that AI-related capital spending could slow.
Naveen Rao, the CEO and co-founder of Unconventional.AI, has founded three AI startups. He sold one to Intel and another for over a billion dollars, while his current company has a valuation of $4.5 billion.
Rao’s experience illustrates how quickly technical work can move while other constraints persist. His company got a chip designed in six months without a dedicated team using AI, yet its growth is being hindered by its inability to hire people fast enough.
For AI companies and their backers, the question is therefore wider than whether processors and software can be built. The pace of expansion also depends on whether capital remains available for the infrastructure required to deploy them.
Comments (0)