Marvell Technologies, USA, announced a definitive agreement to purchase optical interconnects manufacturer Celestial AI in a cash-and-stock deal valued at US$3.25 billion. According to the company, the purchase will accelerate its connectivity strategy for next-generation AI and cloud data centers.

“The acquisition of Celestial AI is a transformative step in Marvell’s evolution and expands our leadership in AI connectivity, as scale-up becomes the next frontier in AI infrastructure,” said Matt Murphy, Chairman and CEO of Marvell. “This builds on our technology leadership, broadens our addressable market in scale-up connectivity, and accelerates our roadmap to deliver the industry’s most complete connectivity platform for AI and cloud customers.”

Photonic Fabric

Celestial AI was founded in 2020 by CEO David Lazovsky and COO Preet Virk. Since then, it has closed a number of successful rounds of venture funding, including US$250 million raised this March, US$175 million in March 2024, US$100 million in July 2023 and US$56 million in February 2022.

The company’s technology is built on its Photonic Fabric platform, an interconnect scheme designed to move optically encoded data to the precise point of computation within the electronic GPU or application-specific integrated circuit (ASIC). It is capable of delivering optically encoded data not just to the edge of the processor, as in a co-packaged optics (CPO) scheme, but to any location on the computer die. Celestial AI claims that disaggregating compute and memory allows the Photonic Fabric to deliver 25 times greater bandwidth at 10 times lower latency than CPO interconnect alternatives coming into the market.

Transition to optical links

In announcing the acquisition, Marvell points to the ever-growing demands that AI is placing on data center architectures. “As bandwidth and reach continue to increase, every connection point in the data center must move from copper to optical,” the company says. “For rack-to-rack scale-out and data-center to data-center (DCI) connections this transition has already taken place. The next inflection point is within the rack, within the system, and even within a package—where electrical connections must now give way to optics.”

Celestial AI’s Photonic Fabric technology platform was “purpose-built for this inflection,” claims the tech giant. “It enables large AI clusters to scale both within and across racks using a high-bandwidth, low-latency, low-power and cost-effective optical fabric.”

Marvell also highlights the thermal stability of Photonic Fabric as an advantage of the technology, as it “enables reliable operation in the extreme thermal environments created by multi-kilowatt XPUs and switches.” This resilience allows the technology to be co-packaged vertically with high-power XPUs and switches in a 3D package, so the photonic connection can be made directly into the XPU, rather than from the edge of the die.

The first application of the platform will be all-optical scale-up interconnects, as high-speed XPU (extreme processing unit) links transition from copper to optics to meet the reach and bandwidth demands of next-generation rack-scale architectures.

A look at the numbers

According to the terms of the deal, which should close by the end of March 2026, Celestial AI will receive US$1 billion in cash and 27.2 million shares of Marvell common stock, which are currently worth US$2.25 billion. Marvell says it expects meaningful revenue contributions from Celestial AI to begin in the second half of fiscal year 2028, reaching a US$500 million annualized run rate in the fourth quarter of fiscal 2028 and doubling to a US$1 billion dollar run rate by the fourth quarter of fiscal 2029.

 “Marvell is the ideal home for our Photonic Fabric, with the scale, customer relationships and connectivity leadership to take this platform into high-volume production,” said Celestial AI’s Lazovsky. “Together with Marvell’s scale-up switching strategy, we’re excited to accelerate the transition to optical scale-up interconnect and expand what next-generation AI infrastructure can achieve.”