• NVIDIA has expanded its focus beyond AI RAN with AI Grid to help telcos build a more intelligent, distributed and sovereign infrastructure for the growing token economy to save costs and monetize. 
• Ecosystem: To enable this, NVIDIA is forming a new partner ecosystem from telcos to network infrastructure vendors. which are very keen on the AI Grid concept such as SoftBank, T-Mobile and Comcast as they see considerable revenue opportunity at the network edge, increasingly with Physical AI services.  
• Intelligence: Agentic AI as a network optimizer is the focus with AI Grid nicely complementing the promising native AI capabilities in upcoming 6G networks. 
• Distributed: The buck or but or now token moves beyond the traditional MEC concept to a more distributed and workload optimized architecture to maximize performance and minimize cost per token 
• Sovereignty: The changing geopolitical climate are influencing the infrastructure spending and deployment patterns all over the world, demanding sovereignty. This is prevalent from Canada to Saudi Arabia to Singapore. 

Right off the back of MWC Barcelona 2026, NVIDIA’s GTC 2026, held in March, featured considerably more focus on the telecommunications industry this year. For the first time there was a dedicated telco pavilion in the exhibition hall where several leading telcos, including SoftBank, TELUS, Singtel and T-Mobile and key vendors such as Amdocs and Keysight, showed off their plans for the future. We have been following NVIDIA’s activities in the space since the AI Summit Japan in 2024 where the company first announced its AI-RAN ambitions with SoftBank and Nokia, and there has been considerable progress since then. 

Here are our key telco-related observations from GTC 2026: 

Introducing NVIDIA’s AI Grid – Technically Impressive but Will it Monetize? 

The conversation, which started two years ago focusing on AI-RAN, has now pivoted somewhat to “AI Grid”. This makes practical sense as the economics of embedding AI over thousands of cell sites would seem to be hard to justify, despite the fact that advanced telcos like SoftBank and SK Telecom are pursuing it. Hence, moving to edge nodes for inferencing makes sense for most telcos. At GTC 2026, several telcos announced their intentions to build AI Grids including T-Mobile, AT&T, Spectrum, Comcast and Indosat Ooredoo Hutchison.  

AI-RAN vs AI GRID : AI-Based Infrastructure Shift for Telcos 

Source: Counterpoint Research

Intelligent: Agentic AI as a Telco Game Changer 

Unsurprisingly, agentic AI featured heavily at the show, which was also a huge theme at MWC. NVIDIA’s Nemotron-based large telco model (LTM) is being used by telcos such as Cassava Technologies, NTT DATA, and Telenor to train agents to optimize mobile networks in areas such as traffic management, governance and security. 

NVIDIA’s model has 30 billion parameters and is open source. This will make it easier for telcos to approach L4 automation but also brings into question the commercial viability of vendors who are charging for similar solutions.   

A New NVIDIA-led Telco Ecosystem with 6G on the Horizon 

Source: NVIDIA 2026 GTC 

GTC 2026 witnessed a large number of partnerships in the telco space notable including Nokia, Ericsson and Samsung who are the main non-Chinese RAN vendors in the market. These alliances, accompanied by the other numerous partners announced in spaces such as in the edge solutions, cloud, data and storage spaces, are turning NVIDIA into the nucleus of a growing ecosystem which could challenge Huawei and ZTE’s position in certain geographies.   

NVIDIA did not mention (in any sessions we attended) that they purchased a small 3% stake in Nokia in October 2025. It was very clear in the sessions that they have a keen interest in participating in the discussions on how 6G networks should be designed and functioned and it will be interesting to see if they increase that amount going forward and/or make parallel investments in the space, which would considerably alter the dynamics of the telco infrastructure vendor landscape.   

Distributed Edge Infrastructure Help Power New Use-Cases:   

Telcos have explored the possibility of monetizing the network edge before (when the concept was called “Mobile Edge Computing” or MEC) which did not substantially materialize. It is true that there is now a much healthier ecosystem of hardware and software to enable better monetization of the edge and 5G network coverage is much better than it was a few years ago. Physical AI is also increasingly touted as a way to accomplish this but the era where robotics, drones and autonomous vehicles are ubiquitous and powered by mobile networks is still a few years away. But for the AI Grid concept to gain mass acceptance in the industry, it will have to monetize.   To that, NVIDIA and partners showcased proposed distributed AI implementations to power number of use cases to drive monetizable experiences: 

- Personal AI: An improved AI agent which enables a more human-like experience by running SLMs closer to users at the edge to reduce latency and near- real-time responsiveness. 

- Linker Vision: A smart city solution which delivers much more accurate and faster results for camera monitoring of safety and city management solutions leveraging the distributed network edge. 

- Decart: A hyper-personalized distributed media solution running at the edge which enables content delivery and robotic reasoning in near real time. 

Comcast in was one of the telcos that had particularly interesting use cases for their AI Grid.  The company demoed small-business AI concierges, AI-enhanced advertising. and premium, low-latency gaming tiers.  While the industry has heard of such use cases before the operator plans to use Small Language Models (SLMs) at the edge which the company claims is significantly more efficient than using LLMs while still being able to learn over time based on user behavior and deliver hyper-personalized services 

So, the key takeaway here is to watch these AI Grid pioneers and see if they are successfully able to offer these solutions to their customers, which will ultimately determine how widespread this new network architecture becomes. 

Telcos Lead the Sovereign AI Charge  

Source: NVIDIA 2026 GTC 

Another heavily discussed topic at GTC 2026 was Sovereign AI. To be fair, there is no standard agreement in the industry as to what counts as “sovereign” and what does not, but from the show it was clear that the desire for more local control over AI infrastructure is having a big impact on how networks are being built and designed.  It was also clear that telcos are playing a prime role in Sovereign AI in many parts of the world. NVIDIA announced that thus far 24 Telco AI factories have been built all over the world and this trend is not limited to advanced markets. Min-Yong Ha from SK Telecom, the most advanced telco in the world in terms of advancing telco-driven Sovereign AI, was there discussing the company’s A.X LLM, which runs on sovereign infrastructure and is already running apps used by a majority of the South Korean population. Vikram Sinha from Indosat Ooredoo Hutchison also shared the stage to discuss the progress of its Sahabat-AI Bahasa Indonesia-based LLM, the country’s largest such initiative. It was clear from the show that Sovereign AI will represent a new rule-of-the-road for AI infrastructure spending going forward. 

Final Thoughts 

Overall, there was a much larger emphasis on the telco space at GTC 2026 – and a lot more enthusiasm for what is in store for the sector. It was also clear that NVIDIA is positioning itself to take a much larger role in the space. Before the 6G era begins, telcos are moving to monetize the network edge with lower latency services and eventually Physical AI solutions. Agentic AI is becoming increasingly useful to telcos for network optimization leading up to native AI 6G networks, and Sovereign AI is changing deployment and spending patterns. All of the above is changing the telecom infrastructure vendor landscape, which we believe will change significantly in the near future.