ResolightAI (@ResolightAI) header
ResolightAI (@ResolightAI) profile picture
Joined March 2026
The next bottleneck in AI networking isn't necessarily the link. It's what happens when data reaches the switch. The industry is making significant progress with CPO, NPO, and other optical interconnect technologies. But faster optical connections still face an architectural limitation when traffic must be converted back into electrical signals for switching. Each conversion adds energy consumption, latency, and hardware complexity. ๐—ก๐—ฎ๐—ป๐—ผ๐˜€๐—ฒ๐—ฐ๐—ผ๐—ป๐—ฑ ๐—ข๐—–๐—ฆ ๐—ณ๐—ผ๐—ฟ ๐˜๐—ต๐—ฒ ๐—”๐—œ ๐——๐—ฎ๐˜๐—ฎ ๐—ฃ๐—น๐—ฎ๐—ป๐—ฒ ๐˜๐—ฎ๐—ธ๐—ฒ๐˜€ ๐—ฎ ๐—ฑ๐—ถ๐—ณ๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐˜ ๐—ฎ๐—ฝ๐—ฝ๐—ฟ๐—ผ๐—ฎ๐—ฐ๐—ต. By separating control from payload and coordinating data transfers at the endpoints, Resolight's architecture enables optical switching without electronically processing every bit along the data path. The goal isn't simply to make existing networks faster. It's to remove unnecessary processing and conversions from the network itself. Because scaling AI efficiently requires more than faster connections. It requires a different switching architecture. Learn more: resolight.ai/ #AIInfrastructure #NanosecondOCS #Photonics #OpticalNetworking #CPO #DataCenters #scalingAI #deeptech
1
28
What if the most valuable network component is the one you don't need? As AI clusters grow, so does the infrastructure required to connect them: switches, transceivers, optical links, power, cooling, and physical space. But scaling compute shouldn't require network complexity to grow at the same pace. Resolight.'๐˜€ ๐—ก๐—ฎ๐—ป๐—ผ๐˜€๐—ฒ๐—ฐ๐—ผ๐—ป๐—ฑ ๐—ข๐—–๐—ฆ ๐—ฎ๐—ฟ๐—ฐ๐—ต๐—ถ๐˜๐—ฒ๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ ๐˜๐—ฎ๐—ฟ๐—ด๐—ฒ๐˜๐˜€ ๐˜‚๐—ฝ ๐˜๐—ผ ๐Ÿณ๐Ÿฑ% ๐—ณ๐—ฒ๐˜„๐—ฒ๐—ฟ ๐—ผ๐—ฝ๐˜๐—ถ๐—ฐ๐—ฎ๐—น ๐˜๐—ฟ๐—ฎ๐—ป๐˜€๐—ฐ๐—ฒ๐—ถ๐˜ƒ๐—ฒ๐—ฟ๐˜€ ๐—ฎ๐—ป๐—ฑ ๐˜‚๐—ฝ ๐˜๐—ผ ๐Ÿฐ๐Ÿฌร— ๐—ฏ๐—ฒ๐˜๐˜๐—ฒ๐—ฟ ๐—ป๐—ฒ๐˜๐˜„๐—ผ๐—ฟ๐—ธ ๐—ฒ๐—ป๐—ฒ๐—ฟ๐—ด๐˜† ๐—ฒ๐—ณ๐—ณ๐—ถ๐—ฐ๐—ถ๐—ฒ๐—ป๐—ฐ๐˜† ๐—ถ๐—ป ๐—บ๐˜‚๐—น๐˜๐—ถ-๐—ฟ๐—ฎ๐—ฐ๐—ธ ๐—”๐—œ ๐˜€๐—ฐ๐—ฎ๐—น๐—ฒ-๐˜‚๐—ฝ ๐—ป๐—ฒ๐˜๐˜„๐—ผ๐—ฟ๐—ธ๐˜€. Fewer conversions mean less energy spent moving data through the network. Fewer components mean less hardware to source, deploy, and maintain. And in power-constrained AI factories, those savings can create room for more compute. The economics of AI won't be determined by GPU performance alone. They will increasingly depend on how efficiently those GPUs are connected. Learn more about our approach: resolight.ai/blog/resolight-โ€ฆ #AIInfrastructure #NanosecondOCS #Photonics #OpticalSwitching #DataCenterEfficiency
1
8
AI networking has reached an inflection point. According to Sameh Boujelbene, Vice President at @DellOroGroup, AI back-end switch sales surpassed front-end networking sales for the first time in Q2 2026. As AI infrastructure scales, the challenge is no longer just delivering more bandwidth. It is reducing the energy, components, and complexity required to move data between accelerators. The @AdvancePhotonic's new CPO specifications mark an important step forward, introducing support for ๐—ก๐—ฎ๐—ป๐—ผ๐˜€๐—ฒ๐—ฐ๐—ผ๐—ป๐—ฑ ๐—ข๐—ฝ๐˜๐—ถ๐—ฐ๐—ฎ๐—น ๐—–๐—ถ๐—ฟ๐—ฐ๐˜‚๐—ถ๐˜ ๐—ฆ๐˜„๐—ถ๐˜๐—ฐ๐—ต๐—ถ๐—ป๐—ด (๐—ข๐—–๐—ฆ) ๐—ถ๐—ป ๐˜๐—ต๐—ฒ ๐—”๐—œ ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—ฝ๐—น๐—ฎ๐—ป๐—ฒ. By keeping data in the optical domain and reducing repeated conversions, native photon routing offers a path toward more efficient and economically scalable AI networks. Read Resolight's latest press release on what this milestone means for AI workload economics: resolight.ai/blog/resolight-โ€ฆ #AIInfrastructure #NanosecondOCS #Photonics #OpticalNetworking #CPO #DataCenters
2
2
73
AI networking needs more than faster optical connections. It needs a more efficient way to switch data. The @AdvancePhotonic Coalition's newly released CPO specifications mark an important milestone, introducing standards support for Nanosecond Optical Circuit Switching (OCS) in the AI data plane. At @ResolightAI, we're developing native photon routing that keeps data in the optical domain, avoiding repeated optical-to-electrical-to-optical conversions. As @ofershapiro, CEO and Co-Founder of Resolight, highlights, our architecture targets: โœ…Up to 40ร— better network energy efficiency โ โœ…Up to 75% fewer optical transceivers For multi-rack AI scale-up networks, the implications go beyond performance. Reducing network power, hardware requirements, and infrastructure footprint can fundamentally improve the economics of delivering AI workloads. The APC specifications are an important step toward making this architecture practical at scale. ๐Ÿ‘€ Read our latest press release to learn more about Nanosecond OCS and its implications for AI workload economics: resolight.ai/releases/resoliโ€ฆ #NanosecondOCS #AIInfrastructure #Photonics #OpticalSwitching #CPO #AINetworking
2
69
Weโ€™re at PECC 2026! Resolight is proud to be a Platinum Sponsor of the @OpticaWorldwide + @AdvancePhotonic PECC Summit, taking place October 20โ€“21 in Sunnyvale. As AI clusters continue to scale, traditional electronic switching is becoming a growing constraint on network performance, power, and complexity. ๐—”๐˜ ๐—ฅ๐—ฒ๐˜€๐—ผ๐—น๐—ถ๐—ด๐—ต๐˜, ๐˜„๐—ฒ'๐—ฟ๐—ฒ ๐—ณ๐—ผ๐—ฐ๐˜‚๐˜€๐—ฒ๐—ฑ ๐—ผ๐—ป ๐—ก๐—ฎ๐—ป๐—ผ๐˜€๐—ฒ๐—ฐ๐—ผ๐—ป๐—ฑ ๐—ข๐—–๐—ฆ ๐—ณ๐—ผ๐—ฟ ๐˜๐—ต๐—ฒ ๐—”๐—œ ๐——๐—ฎ๐˜๐—ฎ ๐—ฃ๐—น๐—ฎ๐—ป๐—ฒ, ๐—ฟ๐—ฒ๐˜๐—ต๐—ถ๐—ป๐—ธ๐—ถ๐—ป๐—ด ๐—ต๐—ผ๐˜„ ๐—ฑ๐—ฎ๐˜๐—ฎ ๐—บ๐—ผ๐˜ƒ๐—ฒ๐˜€ ๐—ฎ๐—ฐ๐—ฟ๐—ผ๐˜€๐˜€ ๐—”๐—œ ๐—ถ๐—ป๐—ณ๐—ฟ๐—ฎ๐˜€๐˜๐—ฟ๐˜‚๐—ฐ๐˜๐˜‚๐—ฟ๐—ฒ. Attending PECC? Stop by Booth T5 to meet our team and discuss the future of optical switching, or schedule a 1:1 meeting with us: resolight.ai/contact-us/ #PECC2026 #Photonics #AIInfrastructure #OpticalSwitching #OCS #DataCenters
1
33
AI infrastructure is entering a new phase, and the data plane needs to evolve with it. As clusters scale across more racks, the challenge is no longer only bandwidth. It is switching speed, energy efficiency, latency, and the ability to move data without adding more complexity at every layer of the network. At Resolight, we are focused on Nanosecond OCS for the AI Data Plane: optical circuit switching designed for AI-scale networking, with photonic processing at the core. Schedule a meeting or contact us here and we can meet after the conference: resolight.ai/contact-us/ #Photonics #AIInfrastructure #OpticalNetworking #OCS #DataCenters #PhotonicProcessing
21
Looking forward to sharing more on Nanosecond OCS for the AI Data Plane at OCP. Breaking out of the rack requires more than photonic interconnect. It requires a new approach to switching. Letโ€™s meet at OCP and discuss how Nanosecond OCS can help scale AI networks more efficiently. resolight.ai/contact-us/ #OCPGlobalSummit #AIInfrastructure #Photonics #OpticalNetworking #OCS #DataCenters
1
1
76
Resolight will be at the 2026 OCP Global Summit in San Jose, October 12โ€“15. AI infrastructure is hitting a new constraint: the network. More GPUs only matter if data can move between them fast enough, efficiently enough, and within a power budget that can scale. That is where photonic processing and optical switching become critical. We look forward to meeting partners, customers, investors, and industry leaders at OCP to discuss how AI networks can become more scalable, faster, and energy-efficient. To meet with Resolight at OCP, contact us here, on social, or at: resolight.ai/contact-us/ #OCPGlobalSummit #OpenComputeProject #AIInfrastructure #Photonics #OpticalNetworking #DataCenters #PhotonicProcessing
1
1
76
Amazing week at #ECOC2026 and so exciting to see industry momentum and innovation around #photonics #CPO #NPO #XPO. Looking forward to @OpenComputePrj and @OpticaWorldwide @AdvancePhotonic #PECC back in California
3
7
46
This week, Ofer Shapiro, CEO and Co-Founder of Resolight, will speak at the @OpticaWorldwide Global Photonics Economic Forum in Mรกlaga. @ofershapiro will join the AI Infrastructure panel discussion alongside leaders from across the photonics and data infrastructure ecosystem to discuss the optical technologies enabling next-generation AI infrastructure. As AI clusters continue to scale, the network is becoming one of the defining constraints: bandwidth, power, latency, packaging, and the ability to move data efficiently across increasingly complex systems. At Resolight, we believe photonic processing will be central to that next phase. We will be available for meetings during GPEF. To connect with Resolight, message us here or contact us at: resolight.ai/contact-us/ #GPEF2026 #Photonics #AIInfrastructure #OpticalNetworking #DataCenters #CPO #PhotonicProcessing #Optica
3
7
35
This week, Resolight will be at ECOC 2026 in Mรกlaga. As optical communications becomes increasingly central to AI infrastructure, ECOC is an important place to discuss what comes next: CPO, optical I/O, photonic processing, and the network architectures needed to support the next generation of AI scale. For Resolight, the focus is clear. AI infrastructure is no longer only constrained by compute. It is constrained by how efficiently data can move between GPUs, racks, and systems โ€” inside a power budget that can actually scale. We will be available for meetings throughout the event. To connect with Resolight at ECOC, message us here or contact us at: resolight.ai/contact-us/ #ECOC2026 #AIInfrastructure #Photonics #OpticalNetworking #CPO #DataCenters #PhotonicProcessing
1
3
7
80
Itโ€™s hard to be a physicist. Even two air-conditioning units can become a thermodynamic crime scene. That is where the interesting problems begin. Read the full blog: resolight.ai/blog/its-hard-tโ€ฆ To most people, they are just two boxes on a wall. But when one unit partially inhales the hot exhaust of the other, it becomes an efficiency problem. The system works harder, consumes more energy, and produces even more heat. The same logic applies to AI networking. At small scale, picojoules per bit can sound insignificant. But inside large AI fabrics, every electrical conversion and every switching operation consumes energy, generates heat, and adds complexity. At sufficient scale, almost every information problem becomes an energy problem. That is why photonic processing matters. Instead of accepting repeated electrical conversions as unavoidable, we can ask whether the network architecture itself is creating the waste. Find the thermodynamic crime scene. Follow the wasted energy back to the architectural decision that created it. #AIInfrastructure #Photonics #OpticalNetworking #DataCenters #EnergyEfficiency #PhotonicProcessing
3
6
26
Resolight is proud to be a Platinum Sponsor of PECC 2026, hosted by @OpticaWorldwide and the @AdvancePhotonic , taking place October 20โ€“21 at the HPE Aspiration Dome in Sunnyvale. The summit brings together leaders from across AI infrastructure, photonics, cloud, semiconductors, networking, and manufacturing to discuss the technologies and supply chains required for the next phase of AI scale. For Resolight, the focus is clear: optical networking is no longer only about transport. It is becoming central to how AI infrastructure scales. We look forward to meeting customers, partners, investors, and industry leaders at PECC to discuss photonic processing, CPO, optical switching, and how the network can become 10ร— more scalable, faster, and efficient. To meet with Resolight at PECC, contact us here, on socials or at: resolight.ai/contact-us/ @ofershapiro @MariMinetaClapp #PECC #Photonics #AIInfrastructure #OpticalNetworking #CPO #DataCenters #CloudComputing #PhotonicProcessing
3
7
48
Looking forward to joining this panel at the Global Photonics Economic Forum in Mรกlaga. AI infrastructure is pushing data center architectures into a new phase, where optical interconnects and photonic technologies will play a critical role in scaling compute efficiently. Excited to join this discussion with @OpticaWorldwide , moderated by Gregg Bartlett, CTO at @GlobalFoundries and the other panelists. See you in Mรกlaga. @ofershapiro #GPEF2026 #Photonics #AIInfrastructure #OpticalNetworking
3
7
59
What if AI infrastructure does not only need more power brought to compute? What if compute also needs to move closer to available power? As AI data centers become more distributed, the network becomes critical. Resolightโ€™s photonic processing is designed to make that fabric more flexible and energy-efficient. Read more at: resolight.ai/blog/making-betโ€ฆ #AIInfrastructure #Photonics #DataCenters
4
7
32
AI scale is not just a compute problem. It is a data movement problem. Optical links help, but electronic switching still creates power, cost, and complexity bottlenecks. Resolight uses photonic processing to rethink how AI networks move data. Read more at: resolight.ai #AIInfrastructure #Photonics #AI
3
7
28
Whatโ€™s more expensive: coffee or AI? โ˜•๏ธ Time to find out: resolight.ai/blog/whats-moreโ€ฆ At first, a picojoule sounds too small to matter. It is one trillionth of a joule. Boiling a cup of coffee takes roughly 84 quadrillion picojoules. But in AI infrastructure, the important part is not just the energy unit. It is per bit. A single AI rack can move enough bandwidth to host hundreds of millions of Zoom calls. Across tens of thousands of AI racks, saving even 1 pJ/bit can translate into massive power savings. In a conservative U.S. AI rack forecast, eliminating just 1 pJ/bit could free approximately 124 MW of continuous power, or about 1.1 TWh annually. But the bigger opportunity is not only cheaper electricity. In a power-constrained AI factory, freed energy can be converted into more compute, more tokens, and more revenue capacity. That is why the industry is investing billions into technologies like CPO, LPO, NPO, and XPO to recover individual picojoules. At @ResolightAI, we believe the opportunity goes further: redesigning the scale-up network to eliminate up to 40 pJ/bit. At AI scale, picojoules are not small. They are economic leverage. #AIInfrastructure #Photonics #DataCenters #OpticalNetworking #CPO #EnergyEfficiency
๐Ÿค– Made with AI
3
1
8
107