AvidThink 2026 July Recap

July was all quiet on the event front, with a few exceptions like the RAISE Summit in Paris, which we didn’t attend, but friends who were there said it was both cool and hot at the same time (because of AI and weather). AMD held their own GTC-style event in San Francisco’s Moscone West: AMD Advancing AI 2026, which our partner, Jim Carroll, managed to catch. The slowdown in conferences allowed the team to focus on prep work for the busy fall conference season starting in August, beginning with the Future of Memory Summit (FMS) in Santa Clara, CA, followed by the online IEEE Hot Interconnects, then IEEE Hot Chips (Stanford), AI Infra Summit (Santa Clara, CA), FutureNet Asia (Singapore), OCP Global Summit (San Jose), and more! Our media partner, Jim Carroll of Converge Digest, will be on the ground at most of these events capturing content for our newly revamped joint NextGenInfra.io site!

Speaking of NextGenInfra.io, check out our revamped site! It loads faster than before, is easier to navigate, and includes new features like browse by topic and browse by company. We’ve streamlined and reduced the friction for downloading our reports — give it a try. And while you’re there, take a stroll down memory lane with our NextGenInfra archives. Hard to believe that our collaboration with Jim has resulted in an exclusive collection of over 1100 interviews from over 350 organizations over the years! BTW, since it’s a migration from our previous site, you might spot a few glitches. We’re continuing to make edits and upgrades. If you see an issue, drop us a note at [email protected]. We’d really appreciate it.

Upcoming reports that we’re working hard on include the 3rd edition of our AI for Networking (AIOps and automation) report and our Enterprise Connectivity (SASE, SD-WAN, Campus) report. And by popular demand from our readers and sponsors, we will be refreshing our Data Center Networking showcase and report for the Fall. We’re calling it the Data Center Networking Report 2026 Fall Edition (how’s that for creativity). In the meantime, we continue to engage the market and sponsors on two new reports: (1) Satellite/Non-Terrestrial Networks (NTN) and (2) Next-Gen Edge Computing, with a focus on AI inferencing workloads. If you’re interested in participating in any of these, drop us a note at [email protected]. Or fill in the form at our brand spanking new website: https://nextgeninfra.io/contact/. We’re pretty sure that form is working.

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

Visit NextGenInfra.io and give us your feedback on the new site.

—The AvidThink Team

Ah, the bliss of summer with a nice break from conferences! I did travel once in July for a private client event, but unfortunately I can’t share too much about it. No complaints on that trip though - less than a 2-hour flight, lovely venue, wonderful people.

I’m taking advantage of being home to work through our backlog of projects and other side tasks that have been pending for a while. One of those projects was updating our NextGenInfra site, and we finally pushed it through after many months of delayed updates. A goal for the new site was to remove tracking cookies and avoid having to pop up the irritating cookie opt-in (especially for visitors from certain parts of the world, like the EU and California). We believe that we’ve achieved that. Another goal was to streamline report downloads, so it’s as frictionless as we can make it. Again, I think we’ve achieved that goal — give it a try!

We’re continuing to tweak our NextGenInfra site and working on our reports in anticipation of a busy Fall. I’m looking at our calendar, and it involves many weeks of back-to-back conferences, often overlapping (Network X and OCP, Mplify Americas and FutureNet Asia, FYUZ and FutureNet Americas, and more). One thing about networking and adjacent industries: there’s no lack of events and conferences.

In any case, I’ll keep this month’s note short. If you are interested in any of the reports we’re working on (check out our report roadmap), or if you would just like to talk and catch up, drop us a line at [email protected].

Until next time, hope you’re able to stay cool (it’s been a brutal summer in many parts of the Northern Hemisphere)!

– Roy for the AvidThink Team

Check out our most recent articles on the AvidThink website:

At the risk of sounding like a broken record,(or more precisely, like a broken record player on repeat), grab your copy of the 2026 Spring Edition of our Data Center Networking Report, before we refresh it in the Fall! Seriously though, clicking that link gets you to the new showcases for our reports. Our showcases now include a key takeaways short section, plus improved summaries of the showcase videos. Do let us know if you like it; or not — drop us a line at [email protected].

6G

NTIA Launches Multinational Initiative for Secure 6G Leadership The NTIA and more than 20 participating countries launched the “Call to Action for 6G Leadership and Security” to coordinate standards, harmonize spectrum, secure trusted vendor supply chains, and develop open, interoperable, AI-native networks. Reflecting on 5G’s costly rollout and limited transformative applications, participants–including Nokia–emphasized early international alignment and networks that evolve at software speed.

T-Mobile Targets C-Band and 2.7 GHz Spectrum for Its 6G Strategy
T-Mobile views upcoming upper C-band and 2.7 GHz auctions as opportunities to strengthen its position ahead of 6G, edge AI, and physical AI, while maintaining that it has sufficient fixed wireless access capacity and has not seen significant AI-driven mobile traffic growth. Its SpaceX Starlink direct-to-device agreement extends beyond 2026, although D2D represented only 0.0003% of peak network usage; Q2 2026 postpaid account net additions reached 277,000, exceeding analyst expectations despite unchanged elements of company guidance.

Infrastructure Sharing Could Improve the 6G Business Case
Operators facing constrained budgets and weak returns on 5G are reconsidering network sharing: China Telecom and China Unicom’s approximately 1.5 million shared base stations reportedly produced $56.4 billion in capital savings and $6.5 billion in annual operating savings. PLDT and Dito have agreed to share Philippine towers, indoor coverage, and subsea cables, while TPG Telecom will pay Optus AU$1.6 billion over 11 years for rural tower access; meanwhile, Telstra’s failure to replace a repeatedly flagged $15,000 component illustrates the risks of excessive cost-cutting.

5G

BMW Adopts Verizon Standalone 5G for Connected US Vehicles
BMW became the first U.S. automaker to use Verizon’s nationwide standalone 5G network, adding direct 5G and LTE connectivity to BMW, Mini, and other BMW Group vehicles manufactured for the U.S. market. KDDI’s Global Communications Platform will support ConnectedDrive infotainment, applications, remote services, and telematics while giving BMW control over connectivity and data routing.

Cohere Wins $28 Million Contract for Covert 5G/6G Drone Sensing
The U.S. Department of War’s Future G Office awarded Cohere Technologies a $28 million contract to develop a multi-waveform RAN prototype that turns commercial 5G and 6G infrastructure into a covert integrated sensing and communications system for detecting, tracking, and helping defeat drone swarms. The prototype combines Cohere’s Pulsone-branded Zak-Orthogonal Time Frequency and Space modulation with conventional OFDM software architectures while remaining indistinguishable from ordinary mobile traffic.

AT&T and Ericsson Demonstrate 5G-Based Drone Detection
AT&T and Ericsson used Massive MIMO radios, multi-static sensing, advanced signal processing, and AI to turn existing 5G infrastructure at AT&T Stadium into a distributed radar system, tracking authorized drones flying at 300 to 400 feet without dedicated sensing hardware. FAA rules prohibit unauthorized flights within one nautical mile and up to 1,000 feet above World Cup venues on match days; authorities detected 1,500 drones and seized more than 700 during the tournament, including 53 around AT&T Stadium.

Verizon Adds 193,000 Fixed Wireless Subscribers in Q2
Verizon added 193,000 FWA, 155,000 fiber, 184,000 postpaid phone, and 77,300 prepaid subscribers in Q2, reaching 17.1 million combined FWA and fiber connections; quarterly revenue was $34.3 billion, while mobility and broadband service revenue rose 2.8% year over year to approximately $23.4 billion. Since becoming CEO in October, Dan Schulman has introduced the $70-per-month Verizon One converged offering and announced a $4 billion BT joint venture and a Google dark-fiber agreement worth more than $1 billion, with further multibillion-dollar AI infrastructure deals expected; projected 2026 capital expenditure is $16.0 billion to $16.5 billion.

AT&T Internet Air Surpasses 2 Million FWA Subscribers
AT&T Internet Air exceeded 2 million subscribers, more than half of whom also have an AT&T wireless plan, after adding 279,000 net subscribers in Q2 despite limited promotion beyond offers such as a $99-per-month bundle for customers aged 55 and older. Verizon added 193,000 FWA subscribers during the quarter and had approximately 6.2 million accounts earlier in 2026, while T-Mobile–at approximately 8.5 million accounts–continues extensive marketing but no longer reports quarterly FWA additions.

Cloud

Ericsson Defends Proprietary Cloud Strategy as Operators Pursue Multivendor Platforms
Deutsche Telekom plans to replace the Ericsson Cloud Native Infrastructure Solution supporting its non-standalone 5G services with a Mavenir standalone core by 2028, advancing its multivendor Horizontal TelCo Cloud strategy and reducing vendor silos. Ericsson maintains that its integrated CNIS approach remains viable while expanding public-cloud options, including trials with seven operators for an on-demand core service developed with Google.

Microsoft Azure Surpasses $100 Billion in Annual Revenue
Azure exceeded $100 billion in full-year revenue for the first time as Azure and other cloud services grew 43% year over year, their fastest rate since 2022; quarterly Microsoft revenue rose 18% to $90 billion, net income increased 31% to $35.8 billion, and Intelligent Cloud revenue climbed 32% to $39.3 billion. A $3.2 billion Anthropic investment gain supported earnings, while Microsoft spent $41 billion in quarterly capital expenditures plus $5.6 billion in finance-lease payments and expects next-quarter capex to exceed $50 billion as it expands data-center capacity.

Google Cloud Growth Accelerates as Alphabet’s AI Capital Spending Surges
Alphabet reported quarterly revenue of $119.8 billion, up 24%, and net income of $112.1 billion–including $98 billion in equity-security gains–while operating income increased from $31.3 billion to $40.8 billion and free cash flow fell to negative $5.9 billion. Google Cloud revenue jumped 82% to $24.8 billion, operating income reached $8.8 billion, margins expanded from 20.7% to 35.6%, and backlog rose to $514 billion from $460 billion in the prior quarter and $106 billion in Q2 2025; quarterly capex more than doubled from $22.4 billion to $44.9 billion, up $9.3 billion sequentially.

AI

Ookla Finds US 5G Networks Unprepared for Uplink-Heavy AI Applications
Ookla found that most 5G networks are not ready for emerging AR and multimodal AI workloads, with the U.S. ranking last among 22 countries for the share of throughput allocated to uplink traffic and twentieth for latency. Wearable AI could shift traffic from today’s roughly 90/10 downlink-to-uplink ratio toward 60/40, exposing limitations in U.S. Time Division Duplex deployments, where uplink and downlink share a channel rather than using the dedicated paths available with Frequency Division Duplex.

AT&T Uses Intelligent Model Routing and Open Source to Cut AI Costs by Up to 90%
AT&T processes an average of 45 billion tokens daily and says its cache-aware proprietary AI gateway can reduce operating costs by up to 90% by routing each request–and even individual turns within a conversation–to the best model for cost, speed, and quality. The company has also introduced OTel 2.0 under its Models as a Service strategy, post-training the telecom-focused open-source model on more than 400 billion tokens using AMD GPUs, GSMA data, and Microsoft Foundry.

European Commission Details AI Act Transparency Requirements
The European Commission’s AI Act rules require providers to disclose chatbot interactions, embed machine-readable markers in generated or manipulated content, identify deepfakes and unsupervised AI content concerning matters of public interest, and warn users about emotion-recognition or biometric-categorization systems. Enforcement powers take effect August 2, while AI systems already on the market have until December 2 to comply with synthetic-content marking and detection requirements.

AI Power Constraints Push Data Centers Toward Unconventional Designs
Power, water, land, and permitting constraints are driving experiments with residential AI servers, ocean-cooled underwater pods, and solar-powered orbital data centers; New York has imposed a one-year moratorium on facilities consuming 50 megawatts or more, while community opposition led Blackstone to cancel a Virginia project. Span and Heata are testing home servers that reuse waste heat for domestic heating and hot water, although residential broadband latency, homeowner incentives, equipment noise, and distributed operations remain significant challenges.

CuspAI Raises $450 Million and Launches AI Materials Foundry
Two-year-old British startup CuspAI raised $450 million in a Series B led by Kleiner Perkins and NEA, reaching a $2.6 billion valuation less than a year after securing more than $100 million in Series A funding. Its new AI Materials Foundry has more than 45 founding partners across the U.S., Europe, and Asia-Pacific, including Nvidia, Meta, Samsung, Hyundai, Henkel, Applied Materials, Tokyo Electron, and Lam Research.

Microsoft Selects AMD Helios for Azure AI Training and Inference
Microsoft will deploy AMD’s rack-scale Helios AI system in Azure for large-scale training and frontier-model inference, joining customers including Meta, OpenAI, Oracle, and Tata Consultancy Services when shipments begin later in 2026. Helios integrates Instinct MI455X GPUs, sixth-generation EPYC “Venice” CPUs, Pensando networking, and ROCm software into a unified architecture supporting Microsoft’s internal development, customer workloads, and Azure AI services.

AT&T Releases OTel 2.0 Open-Source Telecom AI Models
AT&T introduced OTel 2.0, an AMD-trained open-source model suite designed to understand network data, telecom standards, and vendor telemetry that general-purpose models frequently mishandle. Building on OTel 1.0 and the GSMA Open Telco AI initiative launched at Mobile World Congress in March, whose post-trained models have exceeded 18 million downloads, AT&T expects the platform to strengthen bargaining power and contain costs as token usage grows by double-digit percentages.

SoftBank Expands OpenAI-Powered Security Service to 3,000 Enterprises
SB OAI Japan, SoftBank and OpenAI’s joint venture, expanded its Patching as a Service offering to 3,000 enterprise customers in its first full-scale commercial rollout. Combining OpenAI cybersecurity models with SoftBank infrastructure, the service assesses source code and simulates external attacks, detecting an average of 280 potential vulnerabilities per 10 million lines of code in early deployments, with approximately 25% classified as high-risk.

Open RAN

Rakuten Symphony Receives $2.14 Million for Open RAN Trials in Ukraine
Japan’s METI awarded Rakuten Symphony a $2.14 million grant within a broader $4.34 million initiative to build a virtualized, cloud-native open RAN environment for Ukraine, with Rakuten Mobile participating and completion targeted for May 31, 2027. Funded through METI’s FY2024 Supplementary Budget, the project will demonstrate lower-cost, flexible, resilient connectivity and autonomous network operations supporting reconstruction during the conflict.

KDDI and Samsung Improve Peak 5G Throughput by Up to 52%
A field trial launched in late 2025 used Samsung’s RAN Speed Optimizer on KDDI’s standalone 5G network across Tokyo and nearby areas, increasing average downlink throughput during peak periods by 31% and delivering gains of up to 52% in dense urban locations. The AI system replaces manual, cluster-level tuning with real-time prediction and automated cell-level parameter adjustments, reducing operating costs while improving connection speed and reliability.

Software-Driven AI-RAN Promises Greater Capacity and Efficiency
Industry leaders see open RAN’s vendor flexibility and cloud-native foundation enabling AI-based optimization of performance and spectral efficiency, followed by shared edge infrastructure for network and AI workloads. Nokia’s AI-native anyRAN software with NVIDIA’s Aerial AI-RAN supports three Open RAN-compliant deployment paths: plug-ins for existing infrastructure, standalone nodes, and cloud-native implementations on accelerated commercial off-the-shelf servers.

Satellite

SpaceX Reportedly Seeks Spectrum for a Full-Service Mobile Network
SpaceX is reportedly exploring acquisitions and federal auctions scheduled for 2027 to obtain urban spectrum and evolve Starlink from rural satellite broadband into a full-service global wireless network competing with AT&T, Verizon, and T-Mobile. Incumbents have spent an estimated $110 billion over the past decade consolidating low-band spectrum, while SpaceX has shown investors a prototype handset and is developing terrestrial infrastructure and hybrid technology to improve indoor and tree-obstructed coverage.

GSMA Intelligence Says Telcos Want Greater Control of Satellite Connectivity
GSMA Intelligence views the May 2026 agreement by AT&T, Verizon, and T-Mobile to pool spectrum and jointly invest in non-terrestrial direct-to-device technology as evidence that operators want greater control over satellite capacity and performance. A neutral-host, shared-ownership model could reduce dependence on Starlink, secure more favorable terms, avoid exclusive arrangements, and benefit wholesale alternatives such as AST SpaceMobile.

Private 5G

Private 5G Market Forecast to Exceed $6.6 Billion by 2030
SNS Telecom & Research expects private 5G spending to grow at a 34% CAGR and surpass $6.6 billion by 2030, while Dell’Oro forecasts 20% annual growth for private wireless RAN systems to more than $1 billion and GPU-RAN is also projected to reach $1 billion by decade’s end. Enterprises are combining private connectivity with compute, AI, and security across campuses, portable backpack-style systems, and wide-area networks to connect robots, sensors, and frontline workers from edge to cloud.

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?