Why Nvidia and Nokia are backing AI RAN specialist ODC

Global operators are joining industry giants like Nvidia and Nokia to inject $45 million into AI RAN specialist ODC.

The Series A funding round for the US-based startup represents an unusual convergence of industry heavyweights. It is rare to see a silicon behemoth like Nvidia, a traditional networking incumbent like Cisco, a legacy radio equipment vendor like Nokia, and global telecoms operators – including AT&T, MTN, and Telecom Italia – participating in the exact same early-stage investment.

This broad consensus demonstrates that the current trajectory of the radio access network (the most expensive and operationally demanding segment of any cellular system) is reaching its physical and economic limits.

ODC is tasked with building technology that integrates AI directly into the radio baseband, replacing rigid human-coded algorithms with machine learning models that can adapt to changing network conditions in real time.

For businesses reliant on advanced connectivity for industrial automation, large-scale IoT deployments, or private enterprise networks, this investment serves as a clear indicator of where corporate connectivity is heading.

The telecoms industry is moving away from single-purpose hardware and leaning heavily into software-defined, intelligent platforms capable of learning and optimising themselves without continuous human intervention.

Ronnie Vasishta, SVP of Telecom at NVIDIA, said: “The industry is moving toward software‑defined, AI‑native telecom networks, which will be essential for the physical AI era.

“ODC’s AI‑RAN stack is a key enabler of this shift, turning today’s 5G networks into a distributed AI computing fabric at the wireless edge. By leveraging the NVIDIA Aerial platform to unify high‑performance 5G with sensing, ODC is helping to raise the innovation bar for AI-RAN and creating a strong on‑ramp to 6G.”

Economic imperative for intelligent radio

To understand why a $45 million investment is turning heads across the industry, one must examine the baseline economics of modern telecoms.

The radio access network accounts for the vast majority of an operator’s capital expenditure and ongoing energy consumption. Cell towers, antennas, and the baseband units that process radio frequencies are incredibly expensive to deploy and maintain.

Furthermore, as global data consumption continues to climb, average revenue per user remains stubbornly flat. Operators are under immense pressure to reduce the cost per bit transmitted, forcing them to find new ways to squeeze more capacity out of their existing spectrum licenses.

For years, the industry has championed Open RAN and virtualised RAN concepts to commoditise hardware and reduce reliance on a handful of proprietary equipment vendors. However, replacing bespoke silicon with general-purpose processors often results in a performance penalty and increased power draw.

Specialists like ODC are positioning AI RAN as the solution to this specific economic puzzle. By embedding AI into the physical layer of the signal processing, operators can optimise spectral efficiency, anticipate hyper-local traffic spikes, and dynamically manage power consumption right at the antenna.

Traditional radio networks broadcast signals at relatively constant power levels based on pre-programmed parameters. An AI-native network, conversely, can learn the specific topographical and behavioural characteristics of a single cell sector.

If an enterprise campus empties out at six in the evening, the AI model can autonomously power down specific frequency bands, saving massive amounts of electricity, and spin them back up milliseconds before a scheduled automated factory run begins. This level of dynamic resource allocation provides operators with a viable path to profitability in the 5G era.

Transitioning to an AI-powered radio network demands an entirely new architectural paradigm, presenting immense implementation challenges. Telecom providers cannot simply deploy a software update to legacy baseband units and expect them to start running complex inference models. They must completely overhaul their edge infrastructure to support high-performance computing workloads right at the base of the cell tower.

This reality brings the IT and telecom worlds into a complex collision. Provisioning GPUs alongside traditional network processors in environmentally harsh, space-constrained locations is a logistical nightmare.

Edge computing sites lack the climate control and redundant power grids of hyperscale data centres. Running AI inference models requires specialised silicon (hence Nvidia’s keen interest in ODC) which traditionally generates considerable heat. Managing the thermal and power dynamics of thousands of distributed AI nodes is a hurdle that operators must solve before AI RAN can achieve widespread commercial deployment.

Beyond the hardware, the data required to train and refine these models is staggering. Network operators must develop the capability to ingest, normalise, and analyse petabytes of radio telemetry daily. Most telecoms providers currently lack the data maturity to handle this volume without overwhelming their own backhaul connections.

Telemetry, therefore, must be processed and acted upon locally at the edge to ensure latency requirements are met—requiring a highly-sophisticated distributed software architecture that many legacy operators are entirely unfamiliar with.

Governing the probabilistic network

Perhaps the most imposing hurdle is the conceptual evolution required to transition from deterministic engineering to probabilistic operations.

Traditional telecoms infrastructure is built on the premise of absolute predictability. A specific input guarantees a specific output, dictated by strict standards established by global regulatory bodies. AI models, by their very nature, infer and predict rather than follow absolute rules.

Placing a probabilistic engine in charge of radio resource management introduces entirely new categories of operational risk. If a machine learning model misinterprets an anomaly in user behaviour and decides to power down a macro cell during a local emergency, the consequences are severe. Network outages caused by AI hallucinations would invite immense regulatory scrutiny and damage enterprise trust irrecoverably.

Consequently, governance frameworks must be established before these systems manage live traffic. Network engineers will need to implement deterministic boundaries around the AI models. These safety nets ensure that regardless of what the neural network recommends, the physical equipment cannot exceed defined operational and safety parameters.

Building such guardrails requires a deep understanding of both radio frequency physics and machine learning validation techniques—two disciplines that rarely overlap in today’s talent pool.

Pallavi Mahajan, Chief Technology and AI Officer at Nokia, commented: “AI is a fundamentally new workload that is reshaping network architecture—driving the need for software-driven platforms, intelligence at the edge, and continuous innovation.

“That shift is putting real pressure on infrastructure and requires architectural change across the network. ODC’s approach to AI-RAN reflects where the industry is heading, moving the RAN toward a more software-driven, AI-ready platform. Nokia’s investment reflects that direction and our focus on enabling AI-native networks across 5G and 6G.”

The human element of this technological adoption also requires careful management. The workforce operating global communication networks has spent decades mastering physical hardware installation and deterministic protocol stacks. Integrating AI requires blending this traditional expertise with data science, machine learning operations, and modern software engineering practices.

Telecom companies will need to upskill their workforce aggressively to survive this transition. The operational silos that traditionally separate internal IT departments from external network engineering teams must be dismantled. A radio engineer and a data scientist will need a shared vocabulary to troubleshoot an AI model that is allocating bandwidth inefficiently during peak enterprise hours.

Furthermore, the nature of machine learning requires continuous model training. The radio environment changes with the seasons as foliage grows and blocks signals, or as new urban developments alter transmission paths. Models must be retrained and deployed continuously, requiring a mature software deployment pipeline that many operators are still struggling to build.

Why the heavyweights are backing AI RAN specialist ODC

The involvement of silicon designers, software networking firms, and legacy hardware vendors in the ODC funding round highlights a broader battle for the future of the telecom edge.

Nvidia clearly views the global network of cell towers as the next massive frontier for its enterprise AI hardware. By processing AI workloads on the same silicon that manages the radio network, operators can theoretically monetise their edge computing assets in new ways, perhaps renting out spare processing cycles to local enterprises when network demand is low.

Organisations investigating private 5G deployments or upgrading their campus networks should demand clarity from their vendors regarding their AI RAN roadmaps. When signing multi-year infrastructure contracts today, executives must ensure the underlying hardware is capable of supporting future machine learning workloads. Investing heavily in rigid, legacy radio equipment right now risks acquiring stranded assets within half a decade.

Enterprises must also begin treating their own network telemetry as a highly-valuable corporate asset. As networks become more intelligent, they will rely heavily on the data fed into them by connected devices. The organisations that successfully harness intelligent infrastructure to drive business efficiency will be those that have spent the preceding years structuring, cleaning, and securing their network data. The era of the static, predictable network is closing, making way for an infrastructure that thinks, adapts, and learns in real-time.

See also: How neuromorphic AI will limit power demands in 6G networks

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the IoT Tech Expo and Cyber Security & Cloud Expo. Click here for more information.

Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.