AvidThink 2025 November Recap

Apologies for the delay in this months’s email — we held off till we could include AWS re:Invent takeaways from Roy.
November represented the peak of the Fall conference season with multiple overlapping events, which meant we only attended a subset and had to FOMO (I guess it can be a verb now) the rest. Wish we were at KubeCon NA, FYUZ, and SC, but it was physically impossible. Regardless, you can check out Roy's takeaways from the events he was at below. This week, we'll be soft-launching our Enterprise Connectivity report, which is a combined SASE/SD-WAN/ZTNA and Campus NaaS report — stay tuned. BTW, if you have an interest in being featured in or sponsoring our upcoming report on data center networking (scale-up, scale-out, scale-across), you can reach us ([email protected]).

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

Keep your eye on NextGenInfra.io for upcoming showcases — including Marvell’s Industry Analyst Day and the Enterprise Connectivity site!

My November travel included attendance at multiple events, including the Mplify (formerly MEF) Global NaaS Event (GNE) in Dallas, the MeterUp event in San Francisco, Network Automation Forum's AutoCon 4 in Austin, the Las Vegas Grand Prix, and AWS re:Invent 2025 again in Las Vegas.

Mplify GNE: GNE allowed us (ex-IDC analyst/ex-HPE veteran Nav Chander, who's helping us out, Jim, and me) to catch up with friends and read the pulse on the impact of AI and data center buildout on carrier network-as-a-service offerings. Congrats to the winners of the NaaS Excellence Awards (esp. Ray Mota of ACG), to Nan Chen for "retiring" as Chairman Emeritus, and to Daniele Mancuso (Sparkle) and Franck Morales (Orange) for their new co-CEO positions. I don't think anyone should ever use "retirement" and Nan in the same sentence, and I expect him to continue to remain active. Mplify/MEF has come a long way and we're seeing an unexpected resurgence in Carrier Ethernet certification driven by AI workloads. Providing NaaS via APIs for inter-DC/cloud AI-related transport, with guarantees around performance, sovereignty, and carbon footprint, is popular amongst carriers, enterprises, and some hyperscalers.

MeterUp: If you were looking for a company to make networking cool again in the post-SDN era (circa 2011-2015), Meter is probably it. MeterUp was their corporate event, and they landed headliners like Satya Nadella (beamed in), Kate Johnson (Lumen CEO), Sanjit Biswas (billionaire co-founder/CEO of Meraki and Samsara), and Bob Metcalfe (Ethernet fame). Microsoft and Lumen are both Meter partners, and Sanjit is an investor (along with Sam Altman, the Stripe Collison brothers, and other luminaries). Meter's taking a vertically integrated approach to networking, embracing building its own hardware, coding its unified network OS, and operating customer networks as a subscription service (yes, onsite hardware is also part of the subscription). Meter also successfully ran its first analyst summit, led by Nathan Herrman (formerly of Cradlepoint/Ericsson), who had just joined weeks earlier. Good to see continued momentum at Meter — and you can learn more when we publish our Enterprise Connectivity report shortly.

AutoCon 4: The day after MeterUp saw me in Austin, TX, at Network Automation Forum's AutoCon 4 — with attendance growing to 750 (from 500 in Denver last year). Congratulations to Scott Robohn, Chris Grundemann, and the NAF team for a job well done. Continued conversations around the nature and shape of network automation in an AI-dominated world of MCP (Model Context Protocol) and A2A (agent-to-agent) dominated. Almost every expo vendor had some form of agentic capability leveraging the influx of reasoning models, whether powered by models-as-a-service from OpenAI or Anthropic, or by fine-tuned, stripped-down, and optimized open-weight models. Even though there were similarities, I still enjoyed catching up with Itential, IBM, Gluware, Network to Code, and Selector AI on their updated solutions. And as always, seeing Wim Hendrickx from Nokia, on his third week of back-to-back travels, was a treat! BTW, check out the newly released high-level starter framework from network automation veterans (Christian Adell, Ryan Shaw, Dinesh Dutt, Claudia de Luna, Damien Garros, Wim Henderickx) under the auspices of NAF.

Formula 1 LVGP: Next up, an early 5 am morning flight from Austin to Las Vegas, just in time for the Las Vegas Grand Prix (my first time at a Formula One race), courtesy of T-Mobile. It was a fantastic experience and a nice showcase of T-Mobile's portfolio of 5G and 5G-Advanced services, leveraging its network and spectrum strength. Grandstand experience, walking next to Koval Lane and the LV Strip, and feeling the breathtaking rush of air as cars passed at speeds exceeding 200 mph. I can finally understand the thrill of Grand Prix racing and seeing the leaderboard upset in the finals due to technicalities, which shows what sticklers the organizers are for Formula One rules (of which there are countless). Almost as bad as or worse than 3GPP and ITU specifications 😂.

AWS re:Invent 2025: Barely a week passed before I was back again in Las Vegas on Thanksgiving Sunday. This time rubbing shoulders with 60K+ attendees and AWS fanboys (and fangirls). AWS was definitely eager to demonstrate legitimacy and relevancy in the era of Generative and Agentic AI, rolling out updates across silicon, computing/storage/networking infrastructure, workflow/framework, and the app layer. New instances featuring NVIDIA B300 (P6e) weren't a surprise, nor were Trainium 3 GA or Trainium 4 preview, new Nova family foundation models, Kiro (coding assistant) updates, AgentCore expansion (policy, evaluations), or upgrades/updates to EC2, S3, RDS services. Of relevance to telcos are AWS Interconnect (multi-cloud connectivity to GCP and others), AI Factories (fully-managed dedicated Local Zone-type offerings that include both Trainium and NVIDIA offerings for private and sovereign AI workloads), AWS Transform custom (agentic solutions for migrating off legacy codebases and infrastructure), and Nova Forge (an assistive framework/environment for training custom models using pre-curated datasets, including Nova models at various checkpoints, mixed with proprietary datasets). Re:Invents are always chock-a-block with announcements and 2025 didn't disappoint — we'll be digesting all this over the next few weeks so reach out to us if you want to discuss.

I'm almost traveled out for the year, just two more events left this coming week — Marvell's Industry Analyst Day (hopefully some insights into Celestial AI $3.25B acquisition + CPO + more), and Aira Technologies Activate event at the McKinsey offices in Redwood City (local, thank goodness). Until the next time, hoping you have a good holiday season and keep your eyes out for our final report for the year: Enterprise Connectivity 2025 coming very soon!

– Roy for the AvidThink Team

We are working on our data center networking report for release in January 2026, but in the meantime, if you want your fill of AI data center content, UALink, ESUN/SUE-T, Ethernet, GPU connectivity, and more, check out our NextGenInfra.io showcases, including OCP Summit 2025 and the brand new 2025 Technology Ethernet Forum (Ethernet Alliance) showcase. With content from Arista Networks, Broadcom, Cisco, Keysight, and alliance members, as well as leading analysts from 650 Group and Dell’Oro, check it out for the latest insights on the role and future of Ethernet for AI/ML workloads.

AI

Vodacom Partners with Google Cloud for AI Transformation in Africa
Vodacom Group has entered a multi-year agreement with Google Cloud to migrate its data systems and integrate generative AI models, including Vertex AI and Gemini, into core business functions. The initiative aims to enhance customer service and network performance, following a separate recent deal with SpaceX’s Starlink for satellite broadband.

Nokia Pledges $4 Billion for U.S. R&D and Manufacturing of AI-Ready Networks
Nokia announced a $4 billion multi-year investment plan in the U.S., allocating $3.5 billion to R&D for networking and defense technologies and $500 million to expand manufacturing in Texas, New Jersey, and Pennsylvania. The initiative, coordinated with the Trump Administration, aims to bolster U.S. competitiveness in AI-ready infrastructure.

Alibaba Struggles to Meet AI Demand Despite Planned $54 Billion Investment
Alibaba CEO Eddie Wu reported that the company is struggling to keep pace with AI demand, a trend expected to persist for up to three years. Despite planning a $54 billion investment over the next three years, Alibaba’s recent $16.9 billion capex significantly trails U.S. hyperscalers like Google ($91 billion) and Microsoft ($65 billion).

Majestic Labs Secures $100M to Solve the AI “Memory Wall”
Led by former Google and Meta executives, Majestic Labs has raised $100 million to develop a new memory ecosystem designed to overcome the bandwidth and latency limitations of current GPU-CPU memory pooling. The startup aims to build a high-capacity memory pipeline comprising new silicon designs and software toolchains to satisfy AI’s data appetite.

Nvidia Q3 Revenue Hits $57 Billion as Management Rejects AI Bubble Fears
Nvidia reported Q3 revenues of $57 billion, a 62% year-over-year increase, driven by $51.2 billion in data center sales. CFO Colette Kress cited visibility into over $500 billion in revenue from Blackwell and Rubin chips through 2026, while management dismissed “AI bubble” concerns, projecting the company could reach a valuation of $3–4 trillion by 2030.

6G

China Leads with Over 80 6G Trials and Patent Dominance
China has conducted over 80 6G technology trials focusing on ultra-high frequency (Terahertz) spectrum and ground-satellite integration, supported by major vendors like Huawei and ZTE. The nation views 6G as an economic priority, holding a significant lead in patent applications and targeting commercial deployment around 2030.

SoftBank Validates 7GHz Spectrum Viability for 6G in Tokyo Trial
Using Nokia’s massive MIMO technology, SoftBank successfully demonstrated that 7GHz spectrum delivers stable, wide-area coverage comparable to existing 3.9GHz 5G sites. The Tokyo-based trial reported a signal-to-interference-plus-noise ratio (SINR) of 5.9 dB with minimal degradation, validating the band’s potential for future 6G networks.

NTT Docomo Doubles 6G Throughput in Outdoor AI-Native Wireless Trials
In partnership with Nokia Bell Labs, SK Telecom, and others, NTT Docomo doubled 6G throughput in outdoor environments by utilizing AI-powered wireless methods compared to non-AI baselines. The trial implemented AI baseband processing on a GPU server operating at 4.8 GHz, demonstrating the technology’s viability in real-world scenarios.

ETSI Initiates Two-Year Integrated Sensing and Communication (ISAC) Program
ETSI’s ISAC Industry Specification Group has launched a two-year program to define frameworks for 6G sensing, covering device, core, and radio network requirements. The work aims to shape future 3GPP and IEEE 802 standards, addressing key areas such as security, privacy, and the integration of AI with sensing capabilities.

Ericsson Mobility Report Forecasts 180 Million 6G Subscriptions by 2031
Ericsson predicts global 6G subscriptions will reach 180 million by the end of 2031, with early commercial launches expected in the U.S., Japan, China, and India around 2030. The report also highlights that Standalone 5G availability has expanded to over 90 service providers, an increase of 20 since June.

T-Mobile Leverages 5G-Advanced for Intent-Based 6G Evolution
T-Mobile is utilizing 5G-Advanced features like network slicing, six-carrier aggregation, and RedCap as the foundation for an intent-based 6G network. Collaborating with Nvidia and Nokia, the operator is embedding machine learning into the network fabric to enable autonomous optimization and predictive maintenance.

5G

AT&T Reports 80% Speed Increase Following Early 3.45GHz Spectrum Deployment
Following a short-term lease of EchoStar’s 3.45GHz spectrum on nearly 23,000 sites, AT&T reports mobile download speeds are up to 80% faster and FWA performance improved by 55%. The deployment proceeds ahead of the final regulatory approval for the $23 billion acquisition, expected to close in mid-2026.

Germany Mandates Removal of Huawei and ZTE from Critical 5G Infrastructure
Germany has formalized legislation requiring operators like Deutsche Telekom and Vodafone to remove Huawei and ZTE components from critical 5G core networks by the end of 2026. A second phase mandates the removal of high-risk vendor equipment from radio access and transport networks by the end of 2029.

Mobile Core Network Revenue Surges 14% Outside China in Q3
Dell’Oro Group reports that while global mobile core network (MCN) revenue fell 2% due to a steep 39% drop in China, markets outside China grew by 14% driven by 5G-Advanced deployments. The EMEA and Asia Pacific regions showed strong momentum with gains ranging between 9% and 17%.

SoftBank Selects Nokia AirScale for 5G Standalone Deployment in Western Japan
SoftBank has chosen Nokia to deploy its 5G Standalone network in western Japan, utilizing the vendor’s AirScale RAN platform and massive MIMO radios. The partnership underscores a shared focus on AI-integrated networks, with both companies collaborating as founding members of the AI-RAN Alliance.

Verizon Reduces Workforce by 13,000 to Cut Labor Costs
Verizon is eliminating approximately 13,000 roles, or 13% of its workforce, aiming to reduce non-union labor expenses by about 20%. The company expects to record a severance charge of up to $1.8 billion in Q4 2025, with more than 80% of affected employees scheduled to exit by December 2025.

AI-RAN

Nokia and Nvidia $1 Billion AI-RAN Partnership Faces Operator Skepticism
Nokia and Nvidia have formed a partnership involving a $1 billion investment and Nvidia taking a 3% stake in Nokia to integrate GPUs into RAN software. However, mobile operators have expressed concern regarding the financial and logistical viability of deploying power-hungry GPUs at smaller mobile sites without clear ROI or efficiency gains.

Vodafone Selects Samsung for Major Open RAN Rollout in Germany
Vodafone has chosen Samsung as a key vendor for its Open RAN deployment in Germany, citing the vendor’s success in high-traffic environments like the US Super Bowl. The rollout is expected to cover approximately 6,000 sites, or 23% of Vodafone’s German footprint, replacing legacy equipment from Ericsson and Huawei.

Samsung and SK Telecom Partner on AI-RAN for 6G Optimization
Samsung and SK Telecom are collaborating to develop AI-RAN technologies for 6G, focusing on channel estimation, distributed MIMO, and next-gen schedulers. As part of an approved AI-RAN Alliance work item, Samsung will provide model development while SK Telecom offers operational data and nationwide test environments.

Deutsche Telekom Deploys Nokia Open RAN Solution Across 30,000 Sites
Deutsche Telekom has selected Nokia’s 1Finity brand for an Open RAN deployment covering over 30,000 sites in Germany. While the deployment utilizes an O-RAN Alliance standardized fronthaul interface, DT is purchasing the primary compute and radio components from Nokia as a single-vendor system.

Private Networks

Nokia Considers Divesting Enterprise Campus Edge Business
Nokia CEO Justin Hotard signaled a strategic shift to move the company’s enterprise and campus edge business into a portfolio entity, potentially leading to a sale. While Nokia aims to divest from systems integration-heavy markets, it plans to retain its mission-critical private network business for sectors like utilities and rail.

Cloud

SaskTel Launches 5G Standalone with Samsung Cloud-Native Core
Canadian operator SaskTel is deploying Samsung’s cloud-native 4G and 5G core to enable a 5G Standalone network across Saskatchewan. The upgrade supports advanced use cases like smart agriculture and virtual healthcare for a network that covers nearly 90% of the province’s population.

Cloudflare Configuration Error Causes Widespread Service Outage
A recent outage affecting major services like Google and OpenAI was traced to a Cloudflare configuration file that exceeded size limits and crashed traffic management software. Cisco ThousandEyes identified the incident as part of a pattern of “configuration cascade effects” that have caused multiple hyperscaler outages this year.

Network as a Service

Lumen and Meter Integrate WAN and LAN into Unified NaaS Platform
Lumen Technologies and Meter have partnered to combine their respective WAN and LAN NaaS platforms, allowing enterprises to manage their entire network stack through a single interface. The joint solution, available via the Meter Connect portal, aims to provide unified visibility and AI-ready connectivity for enterprise customers.

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?