Dell'Oro Group finds optics central to AI networking at OCP 2026
13 companies named across 14 categories, one of 272 articles referencing Dell'Oro Group. Previous coverage: Mplify summit shows gap between federation and execution in NaaS (Oct 2026).
Companies mentioned
Best suited for
- Seniority
- Architect
- Job function
- Network / Network Architect
- Persona
- Network Architect
- Buyer role
- Architect / Technical Evaluator
- Decision stage
- Where to buy
- Adoption curve
- Early Adopters
- Technology maturity
- Accelerated Attention
- Industry
- Information Technology / Networking / Enterprise & Data Center Networking / Data Center Switching & Fabrics
Our classification, not the publisher's statement. Best suited for, not only for.
Optical innovation was one of the biggest themes at OCP 2026, and the analyst observed that optics sessions nearly doubled from last year. As AI networks demand more bandwidth, optics and switching silicon are becoming more tightly intertwined.
Co-Packaged and Near-Packaged Optics
NVIDIA moved its Spectrum-X Ethernet Photonics switches into production, with early customer adoption and deployments beginning. Celestica introduced the DS6005, a liquid-cooled 102.4 Tbps Ethernet switch based on Broadcom's Tomahawk 6-Davisson CPO platform. It offers 64 × 1.6 Tbps ports in a 1.5OU form factor, using co-packaged optical engines and field-replaceable external lasers.
In its rack-scale architecture announcement, Arista identified open CPO/NPO alongside XPO as next-generation optical approaches for increasing network-rack density. Arista cited a path from 1.6 Pb/s with current OSFP-based designs to 6.5 Pb/s per rack. The analyst noted this is an architectural density outlook, not evidence that NPO is already shipping in volume.
Pluggable Optics Increase Bandwidth Density
NVIDIA and Cisco introduced plans for Ultra-Dense Small Form Factor Pluggable (USFP), a proposed 400 Gb/s-per-lane design with a path to as much as 12.8 Tbps of aggregate module capacity. The companies are targeting four times the bandwidth density and approximately 1.3 times the power efficiency of 200 Gb/s-per-lane pluggable designs. That could connect more endpoints without adding network tiers while retaining deployment flexibility and field serviceability.
The analysis states that the story is not simply CPO versus pluggables. Adoption will reflect density, power, reach, reliability, operating model and economics. Customer qualification, serviceability, production yields and shipment volumes remain the metrics to watch.
100T Switching and 1.6T Connectivity
Broadcom's Tomahawk 6 underpins Arista's 7060XE7 portfolio, which supports dense 1.6 Tbps connectivity with air- and liquid-cooled options. Celestica's Tomahawk 6-based DS6000 and DS6001 are available to order, while the DS6005 extends the family into CPO. Cisco reported initial customer shipments of G300-powered 102.4T systems.
Cisco also introduced the 51.2T Silicon One G302, with 200G SerDes and 1.6T connectivity for top-of-rack applications, and planned G302 shipments before year-end 2026. NVIDIA is advancing Spectrum-6, including Meta's announced adoption for next-generation AI infrastructure. Market readiness differs across announcements, orders, customer sampling and shipments.
Scale-Up Extends Beyond the Rack
Microsoft's Maia 300 scale-up domain connects 12,288 accelerators across 128 racks. It combines electrical links within racks with optical inter-rack connectivity and custom transport for accelerator-to-accelerator communication. Scale-up can be defined by workload communication requirements rather than physical rack boundaries.
Arista's Etherlink SU-144 proposes open, Ethernet-based scale-up for as many as 144 accelerators in a single-hop domain and up to 1,024 across racks. Celestica and AMD's Helios collaboration focuses on scale-up switches using Ultra Accelerator Link over Ethernet (UALoE) for AMD's rack-scale AI platform. Meta's Tahan scale-up switch is a 51.2T, Broadcom Tomahawk 5-based Ethernet switch in the MTIA 300 system, and Meta says it is shipping.
OpenAI and Broadcom's Jalapeño is an inference-focused processor, not an Ethernet scale-up switch. Its broader system platform incorporates Broadcom networking technologies. Together, these developments challenge a hard rack boundary between scale-up and scale-out, while raising questions around latency, transport semantics, congestion control and operational complexity.
Competition in Scale-Up Switching Silicon
Upscale AI introduced its 115.2 Tbps SkyFabriX switching silicon as part of its standards-oriented Token Fabric portfolio. Astera Labs unveiled its Orion rack-scale PCIe switching architecture, aimed at connecting multiple accelerators within a pod. These announcements show suppliers beyond incumbent networking-chip vendors targeting the market.
Ethernet, PCIe, NVLink, UALink and emerging open standards will be evaluated against bandwidth, latency, memory semantics, flexibility and cost. No single interconnect is likely to fit every workload. Scale-up is becoming a meaningful new battleground for silicon share.
Hyperscaler Customization and Open Networking
Meta's IcePack liquid-cooled rack houses 102.4T IceCube scale-out switches offered in two silicon variants: Broadcom Tomahawk6C and NVIDIA Spectrum-6. Both use Meta's FBOSS software and OCP Switch Abstraction Interface (SAI). NVIDIA separately detailed its Spectrum-6-based N-IceCube collaboration.
Cisco highlighted Silicon One deployments across hyperscale networks, including P200 systems in Microsoft and Alibaba environments. NVIDIA and Google also announced Falcon Transport support for future ConnectX-10 SuperNICs and BlueField DPUs. These examples show hyperscalers combining merchant silicon with their own software, protocols and system designs.
Analyst Outlook
The analysis concludes that AI networking is no longer simply about faster switches, as optics, switching silicon and network architecture co-evolve. CPO is entering early commercialization, while NPO and other optical integration paths are gaining visibility. Ethernet scale-up is extending from the rack into larger accelerator domains.
Optics, switching silicon and network architecture are increasingly interdependent in the developments described from OCP 2026. For Decision Insights readers, these shifts bear on optical integration, 1.6T connectivity and Ethernet scale-up decisions. This Analyst Signals brief reflects a neutral, fact-based summary of the original research note.
Blog post, originally published by Sameh Boujelbene at delloro.com.