Celestial AI, a Silicon Valley chip startup aimed at overcoming speed limits in artificial intelligence, announced on Tuesday it had secured an additional $250 million in its Series C1 funding round. This brings total funding to $515 million.
The round was led by Fidelity Management & Research, and included participation from former CEO of Maverick Capital, Tiger Global Management and Cadence Design Systems, Lip-Bu Tan. Existing supporters such as AMD Ventures, Koch Technologies (KDT), Temasek of Singapore, Xora Innovation, a subsidiary of Temasek, and Porsche Automobil Holding, also participated in the round.
Founded in 2020 by Preet Virk, David Lazovsky (CEO) and Michelle Tomasko, Celestial AI uses light instead of electrical signals to create faster connections between AI computing and memory chips.
“The rise of complex inference models and the rise of agent AI has led to a rapid increase in demand for AI infrastructure,” said CEO David Lazovsky.
He explained: “We’ve grown cluster sizes of thousands from several AI processors in our servers to dozens of racks, and multiple racks. These setups require high-bandwidth low-reduced network connectivity to manage large-scale data transfers between processors. Celestial AI’s photonic fabric is the only platform to meet these requirements, setting new standards for bandwidth, latency, energy efficiency and total cost of ownership. We are excited to get the support of global investors who share our long-term mission.”
Challenge the current state of AI hardware
The speed of data transfer, known as memory bandwidth, is an important factor in the advancement of AI systems. It is so important that it is considered in the US government’s export controls aimed at limiting AI development in a particular country. Nvidia is currently leading memory bandwidth using its proprietary NVLink and NVSwitch technologies. This has caused a wave of startup activities, with companies competing to develop alternative solutions. Rivals Lightmatter and Ayar Labs raised $850 million and $370 million, respectively.
Supported by AMD Ventures, Celestial AI uses its own photonics to develop bridge technology. This approach sets it apart from its competitors, aiming to increase energy efficiency and reduce delays.
Its flagship product, Photonic Fabric, is designed for AI computing systems and serves as the backbone of data center computing, networking and memory. This technology provides significant improvements in system performance and energy efficiency, aimed at supporting sustainable and profitable AI advances.
Speaking to Reuters, Razovsky emphasized the importance of balancing speed, space and force. “There’s no good answer other than Nvidia,” he said in an interview at Santa Clara, California headquarters. “The ones developed with photonic fabrics achieve similar results, but with better energy efficiency and lower latency.”
Celestial AI will continue to scale its technology to meet the demands of growing industries, focusing on energy efficiency and cost reduction.
Source link
