Relativity Networks raises $22M to stretch AI data centers beyond power bottlenecks

Relativity Networks, led by Luminar Technologies co-founder Jason Eichenholz, has raised $22 million in SAFE financing to commercialize hollow-core fiber technology for AI data centers, with a $40 million follow-on order from an unnamed hyperscaler. The funding round, announced Tuesday and reported by TechCrunch, included investors Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC, and builds on a $4.6 million pre-seed and a $6.1 million seed round. The company aims to reduce latency between AI data centers, addressing power and land constraints without requiring hyperscalers to develop the technology in-house.

Relativity Networks raises $22M to stretch AI data centers beyond power bottlenecks Luminar co-founder Jason Eichenholz is commercializing UCF photonics research, backed by a $40M follow-on order from an unnamed hyperscaler. By RuntimeWire Staff /author/runtimewire-staff · Published Primary source: TechCrunch https://techcrunch.com/2026/08/19/relativity-networks-raises-22-million-to-bring-a-faster-kind-of-fiber-to-data-centers/ Why it matters AI infrastructure is running into power and land constraints. Relativity Networks is betting that lower-latency fiber can widen the usable map without forcing hyperscalers to build the technology themselves. Relativity Networks https://relativitynetworks.ai/?ref=runtimewire CEO Jason Eichenholz https://relativitynetworks.ai/about-us/?ref=runtimewire , the laser physicist who co-founded lidar maker Luminar Technologies, has raised $22 million in SAFE financing for his effort to make the physical distance between AI data centers less punishing. The financing, announced Tuesday and reported by TechCrunch https://techcrunch.com/2026/08/19/relativity-networks-raises-22-million-to-bring-a-faster-kind-of-fiber-to-data-centers/?ref=runtimewire , came from Rhapsody Venture Partners, Bell Ventures Inc., Faster Than Glass LLC and other investors. Relativity Networks also disclosed a $40 million follow-on order from an unnamed hyperscaler, its clearest evidence yet that hollow-core fiber is moving from a photonics project into data center procurement. Relativity Networks did not disclose a valuation or identify a lead investor. The SAFE structure also postpones a priced valuation until a later equity round, leaving the conversion cap, discount and ownership cost outside the announcement. A photonics founder comes back for another build Eichenholz has spent more than three decades turning optics research into products. He earned a physics degree from Rensselaer Polytechnic Institute and graduate degrees in optical science and engineering from the University of Central Florida's CREOL optics college. Before Luminar, he founded Open Photonics and held technical leadership roles at Ocean Optics, Halma and Newport/Spectra-Physics, according to UCF's profile of his career https://www.cecs.ucf.edu/jason-eichenholz/?ref=runtimewire . Relativity Networks is a return to the academic institution where Eichenholz trained. Co-founder and CTO Rodrigo Amezcua Correa https://creol.ucf.edu/person/rodrigo-amezcua-correa/?ref=runtimewire is a UCF professor whose research group spent years developing anti-resonant hollow-core fiber designs. Amezcua Correa earned his doctorate at the University of Southampton and later worked at the University of Bath and Powerlase Photonics before joining UCF. Eichenholz had stepped away from company building when he encountered Amezcua Correa's work. In a February 2025 UCF interview https://www.ucf.edu/news/ucf-alum-professor-team-up-to-power-next-generation-ai-data-centers/?ref=runtimewire , Eichenholz said the technology convinced him that his commercial career was not finished. The pairing gives Relativity Networks a familiar deep-tech division of labor: Amezcua Correa developed the underlying fiber architecture, while Eichenholz is building the manufacturing partnerships and customer relationships required to sell it. Relativity Networks was incorporated in Delaware in 2023, according to a Securities and Exchange Commission filing https://www.sec.gov/Archives/edgar/data/2021908/000202190826000001/xslFormDX01/primary doc.xml?ref=runtimewire . The same filing shows that Relativity Networks began a SAFE offering on January 20, 2026, and had sold $1 million toward an initial $10 million offering target by February 6. The $22 million announced this week is substantially larger than that early filing. The financing follows a $4.6 million pre-seed round and a $6.1 million seed round backed by Prysmian, GOVO Venture Partners and other investors. Using the round amounts in Relativity Networks' July 2025 funding announcement https://www.prnewswire.com/news-releases/relativity-networks-secures-6m-seed-funding-to-power-the-next-generation-fiber-optic-cable-for-data-centers-raising-total-to-10-6m-302511362.html?ref=runtimewire , Relativity Networks has disclosed roughly $32.7 million in financing. That release's headline rounded the earlier figures to $6 million and $10.6 million, while its body reported $6.1 million and $10.7 million. Selling data centers a larger map Conventional optical fiber carries light through silica glass. Hollow-core fiber uses a glass structure to guide light through an air-filled center, reducing the time a signal spends traveling through glass. Eichenholz told TechCrunch that conventional fiber takes roughly five microseconds to carry a signal one kilometer, compared with about 3.5 microseconds for Relativity Networks' fiber. That difference becomes material when thousands of accelerators must remain synchronized across separate buildings or campuses. Data center operators increasingly need to place compute near available electricity, land and cooling capacity, even when those resources sit farther from an existing cluster. Lower propagation delay gives network architects additional distance before latency breaches the limits of a distributed workload. Eichenholz describes that shift as AI infrastructure's next optimization problem. Chipmakers supplied more compute, networking vendors improved connections within data centers, and Relativity Networks is betting that operators will next redesign where those data centers can sit. Faster fiber could let separate campuses behave more like one machine. The $40 million follow-on order gives that thesis commercial weight, although the disclosed figure is an order rather than recognized revenue. Relativity Networks has not identified the hyperscaler, the amount of fiber involved or the delivery schedule. Those details will determine whether the contract represents a broad deployment or a concentrated infrastructure program with one customer. The speed claim needs a common denominator Relativity Networks' performance claims vary across its public materials. TechCrunch reported a 30% speed improvement. Relativity Networks' homepage https://relativitynetworks.ai/?ref=runtimewire and Prysmian describe data moving nearly 50% faster through an air core, while Relativity Networks' technology page https://relativitynetworks.ai/our-technology/?ref=runtimewire displays a separate figure of roughly 17% faster transmission alongside the nearly 50% language. Those percentages may measure different layers of the system, but they cannot be treated as interchangeable benchmarks without test conditions. Investors and customers will need the fiber length, cable configuration, optical equipment and comparison baseline behind each figure. The practical metric is the latency reduction delivered after the fiber has been cabled, spliced and installed in a working route. Relativity Networks has made progress on that less glamorous part of the problem. Prysmian, Relativity Networks' manufacturing partner and an earlier investor, said on July 23 that the companies had tested a 24-fiber cable with a diameter of 10 millimeters in Dura-Line microducts. The cable was installed at jetting speeds of up to 350 feet per minute at a Tennessee test facility, according to Prysmian's announcement https://na.prysmian.com/resources/press-releases/prysmian-relativity-networks-deploy-highest-density-hollow-core-fiber?ref=runtimewire . Prysmian is producing the fiber in Eindhoven and manufactured the cable at its Claremont, North Carolina plant. That partnership is central to Eichenholz's strategy. Relativity Networks does not need to finance an entire global cable-manufacturing footprint before filling orders. Prysmian provides production and cabling capacity, while Relativity Networks supplies the fiber design and commercial platform, branded ChronoCore. Microsoft already decided the fiber matters Relativity Networks is entering a category with a formidable validator and competitor. Microsoft acquired hollow-core fiber developer Lumenisity https://blogs.microsoft.com/blog/2022/12/09/microsoft-acquires-lumenisity-an-innovator-in-hollow-core-fiber-hcf-cable/?ref=runtimewire in 2022 to improve its cloud infrastructure. Microsoft later expanded manufacturing through outside partners, including Corning, and has described production deployments of its technology. Microsoft can build hollow-core fiber around Azure's own infrastructure requirements. Relativity Networks is pursuing a merchant model, offering hyperscalers and data center developers access to the technology without requiring them to develop or acquire a fiber maker. Prysmian's manufacturing reach is what makes that approach credible. The new capital must now carry Relativity Networks through the difficult stretch between a successful cable test and repeatable deployment. That work includes manufacturing yield, field splicing, installer training, attenuation performance and cost. Conventional fiber owns the installed base and supply chain; hollow-core fiber has to earn replacement economics route by route. Eichenholz has built his career around that transition from optics research to manufactured hardware. Relativity Networks gives him a larger target than a component upgrade. He is betting that data center geography has become part of computer architecture, and that a faster strand of fiber can give operators more places to put the machines.