AvidThink 2025 October Recap

If you haven't yet downloaded our latest AI in Networking Report, what are you waiting for? Our "AI in Networking 2025 — Harnessing the AI Deluge" report has been well-received, and we hope you find it useful as we reflect on the current state of AI use in network operations. Up next are our Next-Gen Enterprise Connectivity report, which combines coverage of SASE, SD-WAN, ZTNA, with the latest in enterprise campus network-as-a-service (NaaS), and our Data Center Networking for AI reports. If you have an interest in being featured in those or sponsoring those, you know how to reach us ([email protected]).

Here's what's in store for this month's newsletter:

Stay tuned in the next few weeks for our next-gen enterprise connectivity showcase and report on NextGenInfra.io!

October saw Jim Carroll, our media partner from Converge Digest, and me at the OCP Global Summit 2025. With over 11,000 attendees (up from 7,000 last year) packed into the San Jose Convention Center, it was clear that AI fever continues to sweep the data center market. Congratulations to our friends at the OCP for a well-done event and their ongoing growth!

Quick notes from the show: Liquid cooling has gone mainstream, with scale-up networking technologies, including NVLink, NVLink Fusion, UALink, and the latest OCP ESUN initiative with SUE-T, being hot discussion topics in the corridors. Similarly, issues such as the transition from 800Gbps to 1.6Tbps, CPO (co-packaged optics) versus LPO, copper versus fiber, and SONiC (the open-source network operating system) were debated in the hallways. Many vendors continue to chase faster transmission rates, pursue higher radix switches, extend the reach and size of their xPU clusters while dealing with power and heat constraints. Reach out if you'd like to discuss the conference details further.

The week after the OCP Global Summit, I headed to PyTorchCon in San Francisco. PyTorch is the leading open-source machine learning framework used for developing and training deep neural networks. Originally developed by Meta, it is now managed by the PyTorch Foundation, part of the Linux Foundation. It is helpful to explore one level up the stack from the underlying infrastructure to understand better the demands these training, fine-tuning, and inferencing workloads place on the computing, networking, and storage platforms. Plus, it was a treat to hear from AI luminaries, including Jeremy Howard, Percy Liang, Nathan Lambert, Dawn Song, Jim Fan, Ion Stoica, Sharon Zhou, Jerry Liu, Noam Brown, Robert Nishihara, and many more. Watching the leaders of Ray, vLLM, SGLang, Monarch, and other AI projects on stage, in person, was definitely eye-opening and inspiring—my appreciation to the Linux Foundation for the analyst pass.

It's clear that the AI/ML field is changing rapidly, and new techniques and frameworks appear every week. Many of today's models and systems, although impressive, remain fragile. As noted by Yifan Mo of the HELM project at Stanford, which runs ongoing benchmarks against the leading models, slight modifications to the prompt and context (such as punctuation and word choice) can alter the performance ranking of the models. There's a lot we know, but even more that we don't understand about these models — optimization techniques abound, model architectures beyond the transformer beckon, and agentic flows and reinforcement learning are the latest hot topics. However, the frameworks and pipelines to support both are still in their early stages of development.

Late addition: To close out travel for October, I headed to New York City for the 1-day AWS Telco Analyst Summit where I had the chance to catch up with many telco analyst friends, and tohear the latest from AWS’s telco industry business unit on customer wins, the latest in generative and agentic AI offerings, and how AWS continues to help telco partners worldwide cloudify, modernize, and get AI-ready to be AI-native. Definitely helpful to hear directly from telcos on real-world results, and see the lat

Next week, Jim and I will be attending the Mplify Alliance (formerly MEF) Global NaaS Event in Dallas (November 10-12) for all things NaaS, followed by the Network Automation Forum AutoCon4 in Austin, TX (November 19-21). To cap off the year, I'll also be attending AWS re:Invent the first week of December. Hope to see you at one of these shows!

– Roy for the AvidThink Team

Our media partner, Jim Carroll, captured ~20 thought leadership videos at the OCP Global Summit 2025. Visit the showcase and check out what industry experts had to share: the latest in liquid cooling; scale-up, scale-out, and scale across networking; Ultra Ethernet, UALink, CXL, ESUN, AEC, ACCs, and many more acronyms — all covered in our exclusive videos.

Amidst the $100Bs of commitment in the AI circular economy, one investment stood out in October: $1 billion dollars from NVIDIA to Nokia. Nothing like a billion dollars to drive GPUs into 6G and accelerate AI-RAN—see more below. BTW, Verizon has a new CEO (just like T-Mobile, except it was unexpected).

Cohere Launches Pulsone to Push OTFS into 6G
Intel-backed Cohere Technologies launched Pulsone Technology to promote its orthogonal time frequency space (OTFS) waveform as part of the 6G standard. Led by Ray Dolan, ex-Flarion founder, Pulsone aims to use OTFS for Integrated Sensing and Communication (ISAC), enabling mobile networks to function as radar systems for military, automotive, and industrial applications.

AT&T Envisions AI-Native 6G Architecture
AT&T’s 6G vision centers on an “AI-native” architecture, “network as a sensor” capabilities, and a software-driven evolution from 5G. The operator plans to avoid large-scale hardware overhauls, extending current modernization efforts well into the 6G era to maintain continuous innovation and disruption.

Verizon: 5G SA Is the Foundation for 6G
Verizon plans to use its 5G standalone (SA) core as the foundation for its eventual 6G rollout. The operator views the SA core as critical for achieving lower latency, network slicing, and a seamless transition to future architectures, ensuring continuity between 5G and 6G systems.

Fraunhofer, Nokia, and Ericsson Team Up on 6G Video
Fraunhofer Heinrich Hertz Institute partnered with Nokia and Ericsson on a 6G-era video coding proof of concept for immersive media. The technology is under review by ITU-T and ISO/IEC groups, targeting ratification around 2029–2030, with use cases in 3D media, AI-generated content, and gaming.

5G

Deutsche Telekom Taps Hackers to Bolster 5G Security
Deutsche Telekom and T-Mobile hosted a “Bug Bash” inviting ethical hackers to stress-test a private 5G standalone network. Over five days, researchers targeted physical, remote, and antenna-based vulnerabilities, marking the first large-scale application of T-Mobile’s U.S. bug bounty model in Europe.

Reliance Jio and T-Mobile to Launch 5G SA Roaming
Reliance Jio and T-Mobile are testing 5G standalone (SA) roaming, aiming for launch by year-end. The partnership will enable uninterrupted 5G SA services for subscribers traveling between India and the U.S., supporting the growing ecosystem of global SA roaming agreements.

AT&T Completes Nationwide 5G Standalone Rollout
AT&T’s nationwide 5G SA network enables low-latency services like AR and cloud gaming, while paving the way for network slicing. The new 5G core operates independently from LTE and aligns AT&T with T-Mobile’s capabilities, serving millions of customers across the U.S.

Verizon Names Dan Schulman CEO, Expands 5G FWA Base
Verizon appointed former PayPal CEO Dan Schulman as CEO, succeeding Hans Vestberg. Q3 results showed 261,000 new 5G FWA subscribers, bringing the total to 5.38 million, with revenues of $33.8B and net income of $5.1B. Schulman will guide Verizon through its Frontier acquisition in early 2026.

AT&T Replaces 40% of Nokia Radios with Ericsson Gear
AT&T replaced 40% of Nokia radios with Ericsson as part of its $14B modernization effort. The carrier added N77 C-band spectrum at 15,000+ sites and is expanding its 3.45GHz coverage. All Ericsson sites will feature Open RAN and Cloud RAN, with nationwide 5G SA coverage now live.

Ericsson Deploys 5G for Saudi Railway’s Smart Trains
Ericsson and Saudi Railway Company are deploying 5G to enhance onboard connectivity and operations. The pilot route will test use cases like real-time video streaming and IoT connectivity, supported by a new 5G innovation center focused on applications and training.

China’s Operators Surpass 1.14 Billion 5G Users
China Mobile, China Telecom, and China Unicom now serve 1.14 billion 5G users. China Mobile leads with 622 million, followed by Telecom’s 292.2 million and Unicom’s 225.2 million. The numbers highlight China’s dominance in 5G adoption six years after commercial launch.

T-Mobile’s Sievert Departs Following Record Q3 Growth
T-Mobile added 2.3 million subscribers in Q3 2025, its best quarter in over a decade. The company gained 1M postpaid phone customers and 506K FWA subscribers, bringing FWA totals to 7.9M. Fiber acquisitions added 54K customers, raising total broadband users to 8.8M.

T-Mobile Expands Enterprise Offerings with 5G Advanced
T-Mobile launched Edge Control and T-Platform to target enterprise customers with 5G Advanced capabilities. Edge Control enables local breakout for ultra-low latency, while T-Platform offers unified management across all business services from a centralized dashboard.

AI-RAN/Open RAN

Nvidia Invests $1B in Nokia for AI-Native RAN Development
Nvidia invested $1B for a 2.9% stake in Nokia to co-develop AI-native RAN infrastructure optimized for Nvidia GPUs. The collaboration will extend into data center and optical domains, laying groundwork for distributed AI inferencing and U.S. telecom competitiveness.

Samsung Wins Major Vodafone Open RAN Contract in Germany
Vodafone selected Samsung as a lead vendor for one of Europe’s largest Open RAN rollouts in Germany, with the first sites live in Hannover and full deployment in Wismar by early 2026. The project will expand across Europe, while Ericsson remains sole vendor in other regions.

AT&T and Ericsson Complete First Multi-Vendor Open RAN Call
AT&T, Ericsson, and Fujitsu’s 1Finity achieved the first Open RAN call using third-party radios on AT&T’s network. The milestone marks progress toward AT&T’s goal of moving 70% of traffic to open hardware by 2026, showcasing maturing interoperability in open fronthaul integration.

SoftBank and Samsung Partner on AI-RAN Research
SoftBank and Samsung are collaborating on AI-RAN research to enhance efficiency, reliability, and automation. Their joint work spans 6G field trials in 7GHz spectrum, AI optimization in Layer 1, shared AI infrastructure, and large-model-driven network orchestration.

Dell Introduces Single-Server Cloud RAN Solution
Dell unveiled the PowerEdge XR8720t, the first single-server Cloud RAN solution designed for Open RAN deployments. Powered by Intel Xeon 6 with up to 72 cores and 24 SFP28 ports, it supports full DU workloads in a compact, ruggedized system to accelerate Open RAN adoption.

AI

Google and Airtel to Build $15B AI Hub in India
Google and Bharti Airtel will invest $15B to create an AI hub in Visakhapatnam, featuring a gigawatt-scale data center and subsea network. Supported by AdaniConnex and India’s “AI Mission,” the project aims to boost national AI infrastructure and economic output by $15B.

Cisco Expands Spectrum-X Partnership with Nvidia
At Nvidia GTC, Cisco joined Nvidia’s Spectrum-X Ethernet platform via its Silicon One and SuperNICs. Cisco will integrate Nvidia networking silicon with its own operating systems to deliver optimized Ethernet solutions for AI data centers.

Qualcomm Unveils Datacenter AI Accelerator Chips
Qualcomm introduced its AI200 and AI250 datacenter accelerators to compete with Nvidia and AMD. Built on the Hexagon NPU architecture, the chips target large-scale inference workloads in cloud and sovereign AI environments, with releases slated for 2026–2027.

South Korea to Deploy 260K Nvidia GPUs for AI Infrastructure
South Korea’s Ministry of Science and ICT and major conglomerates will deploy 260,000 Nvidia GPUs worth up to $10B to power national AI models. The GPUs will support the National AI Computing Center and major cloud providers like Naver, NHN, and Kakao.

SD-WAN

Versa Integrates SD-WAN with Microsoft Entra for SASE
Versa Networks integrated its Secure SD-WAN with Microsoft Entra Internet Access to deliver an automated, unified SASE solution. The collaboration enhances branch office security and performance through streamlined configuration and Microsoft’s SSE stack.

Private Networks

BMW Deploys 5G at “AI Car Factory” in Hungary
BMW and Magyar Telekom launched a hybrid public-private 5G network across a 400-hectare EV factory in Debrecen, Hungary. The “AI car factory” supports connected sensors, autonomous robots, and AR tools across 15% indoor coverage with dedicated on-prem RAN infrastructure.

OneLayer Raises $28M to Secure Private 5G Networks
Private network security startup OneLayer raised $28M in Series A funding led by Maor Investments and Chevron Technology Ventures, bringing total funding to over $43M. The capital will expand OneLayer’s cybersecurity offerings for enterprise 5G and LTE networks.

Airbus Expands Private 5G Rollout for Digital Manufacturing
Airbus is deploying private 5G networks with Ericsson across global facilities to boost industrial automation and traceability. After launches in Hamburg and Toulouse, the rollout will extend to sites in Spain, the UK, Canada, and the U.S. through 2026.

If you would like to meet with AvidThink at one of the following upcoming events, please click on the event below to request a meeting.

if this newsletter was forwarded to you, why not sign up for your own subscription?