Pinned Tweet
Raymond James analyst, Simon Leopold, and @ofershapiro, CEO of @ResolightAI, will discuss the emerging market and applications to employ optical processing to scale AI. Raymond James cites potential implications for compute and networking: $NVDA $AVGO $AMD $ANET $CSCO, etc.
"Resolight.ai believes the real breakthrough comes from coupling co-packaged optics (CPO) with the right network architecture. The data should be converted once at the transmitter and once at the receiver โ and remain light throughout the network. This is the promise of Resolightโs native photonic processor: light-speed optical switching designed for AI scale-up networks. This architecture is built to reduce energy beyond what CPO can achieve on its own, unlock much larger scale-up radix, and enable AI clusters that are larger, faster, and more efficient." - Simon Leopold
๐
Thursday, July 30, 2026 | 11:00 โ noon Eastern
๐Investors can register via Raymond James.
๐ฅA recording will be available through resolight.ai/
๐ Read more now at: resolight.ai/
#photonics #hypercloud #ai #datacenter #photo
@OpticaWorldwide @AdvancePhotonic
๐ช๐ต๐ ๐๐ต๐ผ๐๐น๐ฑ ๐๐ต๐ฒ ๐ฝ๐ต๐๐๐ถ๐ฐ๐ฎ๐น ๐น๐ผ๐ฐ๐ฎ๐๐ถ๐ผ๐ป ๐ผ๐ณ ๐ฎ ๐๐ฃ๐จ ๐ฑ๐ฒ๐๐ฒ๐ฟ๐บ๐ถ๐ป๐ฒ ๐ต๐ผ๐ ๐๐๐ฒ๐ณ๐๐น ๐ถ๐ ๐ถ๐?
Todayโs AI infrastructure is organized around physical boundaries: GPUs inside a rack, racks inside clusters, and clusters connected through increasingly complex network hierarchies.
But AI workloads don't naturally follow those boundaries.
Training and inference have different communication patterns, bandwidth requirements, and latency sensitivities. Yet the network topology often dictates how compute resources can be grouped and utilized.
What if we could reverse that relationship?
Instead of forcing workloads to fit a rigid network architecture, we could design networks that adapt to the communication requirements of the workload.
That is one of the opportunities behind *Nanosecond OCS for the AI Data Plane*: making optical connectivity more dynamic and reducing the constraints imposed by traditional switching hierarchies.
The next step in AI infrastructure isn't just connecting more GPUs.
It's giving operators greater freedom in how they use the GPUs they already have.
resolight.ai/
#AIInfrastructure #NanosecondOCS #AINetworking #Photonics #DataCenters
1
1
2
โ๐ง๐ต๐ฒ ๐ฒ๐ฎ๐๐ถ๐ฒ๐๐ ๐๐ผ ๐๐ผ๐๐ฟ๐ฐ๐ฒ ๐ฎ๐ป๐ฑ ๐๐ต๐ฒ ๐ด๐ฟ๐ฒ๐ฒ๐ป๐ฒ๐๐ ๐๐ฟ๐ฎ๐ป๐๐ฐ๐ฒ๐ถ๐๐ฒ๐ฟ ๐ถ๐ ๐๐ต๐ฒ ๐ผ๐ป๐ฒ ๐๐ต๐ฎ๐ ๐๐ผ๐ ๐ฑ๐ผ๐ป'๐ ๐ป๐ฒ๐ฒ๐ฑ.โ
Described by Excitonic as the โweekโs most quotable line,โ this statement from @ofershapiro, CEO and Co-Founder of Resolight, captures a fundamental challenge in scaling AI infrastructure.
Read the full article by Marcus Chen in Excitonic: excitonic.com/devices-photonโฆ
In its latest coverage of the @AdvancePhotonic Coalitionโs first five manufacturing specifications, Excitonic highlights Resolight and the significance of ๐ก๐ฎ๐ป๐ผ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ข๐๐ฆ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐๐ ๐๐ฎ๐๐ฎ ๐ฃ๐น๐ฎ๐ป๐ฒ.
As demand for optical components continues to grow, increasing manufacturing capacity is only part of the solution. The other opportunity lies in rethinking the network architecture itself to reduce the number of components required in the first place.
With an architecture targeting:
โข Up to 75% fewer optical transceivers
โข Up to 40ร better network energy efficiency
Resolight aims to address both optical supply chain constraints and the economics of AI workloads.
@MariMinetaClapp
#AIInfrastructure #NanosecondOCS #Photonics #CPO #OpticalSwitching #AINetworking
1
4
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.
๐ก๐ฎ๐ป๐ผ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ข๐๐ฆ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐๐ ๐๐ฎ๐๐ฎ ๐ฃ๐น๐ฎ๐ป๐ฒ ๐๐ฎ๐ธ๐ฒ๐ ๐ฎ ๐ฑ๐ถ๐ณ๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ ๐ฎ๐ฝ๐ฝ๐ฟ๐ผ๐ฎ๐ฐ๐ต.
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 ๐ก๐ฎ๐ป๐ผ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ข๐๐ฆ ๐ฎ๐ฟ๐ฐ๐ต๐ถ๐๐ฒ๐ฐ๐๐๐ฟ๐ฒ ๐๐ฎ๐ฟ๐ด๐ฒ๐๐ ๐๐ฝ ๐๐ผ ๐ณ๐ฑ% ๐ณ๐ฒ๐๐ฒ๐ฟ ๐ผ๐ฝ๐๐ถ๐ฐ๐ฎ๐น ๐๐ฟ๐ฎ๐ป๐๐ฐ๐ฒ๐ถ๐๐ฒ๐ฟ๐ ๐ฎ๐ป๐ฑ ๐๐ฝ ๐๐ผ ๐ฐ๐ฌร ๐ฏ๐ฒ๐๐๐ฒ๐ฟ ๐ป๐ฒ๐๐๐ผ๐ฟ๐ธ ๐ฒ๐ป๐ฒ๐ฟ๐ด๐ ๐ฒ๐ณ๐ณ๐ถ๐ฐ๐ถ๐ฒ๐ป๐ฐ๐ ๐ถ๐ป ๐บ๐๐น๐๐ถ-๐ฟ๐ฎ๐ฐ๐ธ ๐๐ ๐๐ฐ๐ฎ๐น๐ฒ-๐๐ฝ ๐ป๐ฒ๐๐๐ผ๐ฟ๐ธ๐.
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-โฆ
#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 ๐ก๐ฎ๐ป๐ผ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ข๐ฝ๐๐ถ๐ฐ๐ฎ๐น ๐๐ถ๐ฟ๐ฐ๐๐ถ๐ ๐ฆ๐๐ถ๐๐ฐ๐ต๐ถ๐ป๐ด (๐ข๐๐ฆ) ๐ถ๐ป ๐๐ต๐ฒ ๐๐ ๐ฑ๐ฎ๐๐ฎ ๐ฝ๐น๐ฎ๐ป๐ฒ.
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-โฆ
#AIInfrastructure #NanosecondOCS #Photonics #OpticalNetworking #CPO #DataCenters
What if the answer to AI networking's growing supply constraints isn't just manufacturing more optical components, but designing networks that need fewer of them?
In a new article for @TheNextPlatform, Joe Fay explores the @AdvancePhotonic's newly released standards and their implications for scaling optical networking.
The article features Resolight and insights from our CEO and Co-Founder, @ofershapiro, on Nanosecond OCS for the AI Data Plane.
As AI infrastructure scales, the challenge extends beyond manufacturing capacity. Traditional network architectures still rely on repeated optical-electrical-optical conversions, adding power consumption, hardware requirements, and complexity.
Resolight's approach targets up to 40ร better network energy efficiency and 75% fewer optical transceivers by keeping data in the optical domain.
As Ofer puts it:
โ๐ง๐ต๐ฒ ๐ฒ๐ฎ๐๐ถ๐ฒ๐๐ ๐๐ผ ๐๐ผ๐๐ฟ๐ฐ๐ฒ ๐ฎ๐ป๐ฑ ๐๐ต๐ฒ ๐ด๐ฟ๐ฒ๐ฒ๐ป๐ฒ๐๐ ๐๐ฟ๐ฎ๐ป๐๐ฐ๐ฒ๐ถ๐๐ฒ๐ฟ ๐ถ๐ ๐๐ต๐ฒ ๐ผ๐ป๐ฒ ๐๐ต๐ฎ๐ ๐๐ผ๐ ๐ฑ๐ผ๐ป'๐ ๐ป๐ฒ๐ฒ๐ฑ.โ
The new APC specifications represent an important step toward making native photon routing practical within an open AI networking ecosystem.
Read the full article by Joe Fay in The Next Platform: nextplatform.com/connect/202โฆ
Thank you to Joe Fay and The Next Platform for covering this important industry development.
#AIInfrastructure #NanosecondOCS #Photonics #OpticalNetworking #CPO #DataCenters
13
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:
โ
Up to 40ร better network energy efficiency
โ โ
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.
๐ Read our latest press release to learn more about Nanosecond OCS and its implications for AI workload economics: resolight.ai/releases/resoliโฆ
#NanosecondOCS #AIInfrastructure #Photonics #OpticalSwitching #CPO #AINetworking
2
69
Weโre at PECC 2026!
Resolight is proud to be a Platinum Sponsor of the @OpticaWorldwide + @AdvancePhotonic PECC Summit, taking place October 20โ21 in Sunnyvale.
As AI clusters continue to scale, traditional electronic switching is becoming a growing constraint on network performance, power, and complexity.
๐๐ ๐ฅ๐ฒ๐๐ผ๐น๐ถ๐ด๐ต๐, ๐๐ฒ'๐ฟ๐ฒ ๐ณ๐ผ๐ฐ๐๐๐ฒ๐ฑ ๐ผ๐ป ๐ก๐ฎ๐ป๐ผ๐๐ฒ๐ฐ๐ผ๐ป๐ฑ ๐ข๐๐ฆ ๐ณ๐ผ๐ฟ ๐๐ต๐ฒ ๐๐ ๐๐ฎ๐๐ฎ ๐ฃ๐น๐ฎ๐ป๐ฒ, ๐ฟ๐ฒ๐๐ต๐ถ๐ป๐ธ๐ถ๐ป๐ด ๐ต๐ผ๐ ๐ฑ๐ฎ๐๐ฎ ๐บ๐ผ๐๐ฒ๐ ๐ฎ๐ฐ๐ฟ๐ผ๐๐ ๐๐ ๐ถ๐ป๐ณ๐ฟ๐ฎ๐๐๐ฟ๐๐ฐ๐๐๐ฟ๐ฒ.
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.
Letโs 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
๐ฐ @BBC's @regwatcher looks at a question that is becoming central to AI infrastructure:
๐ช๐๐ ๐ ๐๐๐๐๐๐ ๐๐ ๐๐๐๐๐ ๐๐๐ ๐๐๐๐๐ ๐๐๐ ๐๐ ๐
๐๐๐ ๐๐๐๐๐๐๐?
Read the full BBC article: bbc.com/news/articles/c2dwg3โฆ
The article captures the broader shift now happening across the industry: copper and electronic networking are reaching physical, thermal, and efficiency limits as AI clusters scale.
Optics can move data more efficiently. But as Ofer Shapiro, CEO of Resolight, explains in the article, the full benefit comes when light is used not only to carry data, but also to process and switch it. That is the architectural challenge ahead.
If data is constantly converted from photons to electrons and back again, much of the efficiency advantage is lost at the switching layer.
Resolightโs approach is built around keeping data in the optical domain and replacing traditional electronic switching with photonic processing. This is where the next phase of AI networking begins: not just faster interconnects, but a fundamentally more efficient network.
#AIInfrastructure #Photonics #OpticalNetworking #DataCenters #PhotonicProcessing #EnergyEfficiency
1
1
37
Resolight will be at the 2026 OCP Global Summit in San Jose, October 12โ15.
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
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
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
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
Itโs 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-tโฆ
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
Resolight is proud to be a Platinum Sponsor of PECC 2026, hosted by @OpticaWorldwide and the @AdvancePhotonic , taking place October 20โ21 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
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















