At the Brooklyn 6G Summit (B6GS) this year, Ronnie Vasishta, Senior Vice President at NVIDIA, delivered a keynote that captured the sense of urgency and opportunity surrounding the intersection of AI and next generation networks. His message was clear. The telecom industry has arrived at a rare moment where a transformational technology is not only available but ready to be embedded into the very foundations of future networks. Whether the industry acts quickly enough to seize this opportunity is the real question.
Ronnie pointed out that telecom has never had access to a tool as powerful and adaptable as AI. Revolutions usually force industries to rethink long held assumptions, and this one is no different. The idea that past experience can predict the future no longer holds. Telecom will need to unlearn familiar approaches, especially where they limit progress, and develop a new mindset better aligned with the realities of the AI era.
A major theme was the changing nature of network traffic. The world is already seeing a strong uptick in application driven traffic, and the next decade is expected to multiply this several times over. Uplink demand in particular is becoming more pronounced as applications grow more interactive and more generative. Ronnie shared an example from NVIDIA’s own live network measurements, showing just how bursty and unpredictable traffic becomes when a single user runs a generative video application. Scale this to millions of users and the challenge becomes obvious.
The spectrum situation will only tighten further. While more spectrum is needed, improved spectral efficiency is essential, but alone will not keep pace with the next wave of use cases. AI native optimisation, both in the RAN and across the network fabric, becomes the only practical route to sustaining growth without unsustainable cost.
Ronnie described a future where billions of intelligent devices operate in highly iterative patterns, exchanging data and reasoning in real time. Not all connections will behave the same way, and not all intelligence will sit in the cloud. A spread of capabilities will exist across the device, the edge and AI factories. The compute required to support these interactions cannot rely on traditional single purpose silicon or general purpose compute. Instead, a software defined, accelerated compute architecture is needed, capable of evolving quickly as AI models and interfaces change. This is where NVIDIA sees its role, and where it believes telecom must follow.
He emphasised that the pace of AI innovation is far faster than what telecom is used to. In AI, ideas move from research to prototypes and sometimes even to products within months. In telecom, ideas can take years to move through standards and into networks. Ronnie argued that this gap needs to close and that software defined, accelerated compute platforms can help make this possible. If the core platforms evolve quickly, the network functions running on them can evolve rapidly too.
This led into NVIDIA’s work on the AI native RAN and the wider AI grid concept. NVIDIA has been building an ecosystem of hardware and software platforms that can run radio functions, AI models and digital twin simulations on the same architecture. The goal is to give researchers and industry teams a way to test, train, simulate and deploy on compatible systems. A digital twin environment like Aerial Omniverse, for example, allows realistic RF behaviour and entire city landscapes to be brought into a workstation for rapid experimentation. The same hardware can then run live RAN applications, making the transition from research to production much simpler.
One of the biggest announcements tied to this vision is the partnership between NVIDIA and Nokia. With the introduction of the Aerial RAN Computer Pro (ARC-Pro), a GPU platform designed to sit directly at radio base stations, NVIDIA aims to show that accelerated compute can meet the thermal, power and size requirements of real network environments. According to Ronnie, this removes the perception that GPUs cannot operate efficiently in telecom deployments. It also enables the idea of a 6G network that can be upgraded primarily through software rather than through frequent hardware refresh cycles.
Another interesting part of the keynote was the discussion about monetisation. Ronnie suggested that future value will flow less from raw data transport and more from the intelligence the network helps generate. He used the lens of tokens as an analogy, positioning them as a unit of value that represents AI output. While the terminology may evolve, the core idea is that networks will need new ways of articulating the value they contribute to AI driven applications. This becomes especially important when AI workloads depend heavily on the latency, reliability and availability that the network provides.
Throughout the keynote, there was a consistent call for telecom to be more ambitious. Ronnie argued that evolution is no longer enough. A revolutionary shift is needed if the industry wants to be central to the AI era rather than simply a passive enabler. That shift involves adopting software defined infrastructure, embedding AI deeply into the air interface and RAN scheduling, and ensuring research and production cycles move faster. It also requires a change in mindset, culture and language across the industry.
The keynote concluded with a reminder that networks are becoming part of a wider computational fabric, not just a transport layer. As the number of connected intelligent devices grows into the hundreds of billions, the network’s role expands from delivering connectivity to orchestrating intelligence. With the right tools, architectures and collaborations, 6G could be the first generation where the network is designed from the start to be AI native.
For those at the Summit, NVIDIA’s demo area provided a glimpse of what this might look like. For the rest of us, the message was clear. The 6G countdown has begun and the AI era is already here. The question is whether telecom can move fast enough to define its place within it.
The video of the talk is embedded below:
Comments
Post a Comment