Dell’Oro Group Finds AI RAN Moving Toward Broader Adoption Across the RAN Stack

Dell’Oro Group argues that Artificial Intelligence (AI) Radio Access Network (RAN) is moving into wider deployment planning, with adoption expected across the RAN in the latter part of the 5G cycle and at the start of 6G. The firm cites shifting vendor roadmaps, growing AI-for-RAN distribution, and operator constraints around compute placement.

Market overview

The article says operators have kept openness, intelligence, automation, and virtualization as RAN roadmap pillars, while visibility and adoption vary by technology. It contrasts early 5G attention on Open RAN and Virtual RAN (vRAN) with current emphasis on AI RAN.

What changed

Dell’Oro Group links recent momentum to events including MWC2026 Barcelona and Nvidia GTC, stating that “AI RAN is already happening.” It also says AI RAN adoption is expected to broaden later in the 5G cycle and from the outset of 6G.

The firm frames the current question as “how, what, where, and when,” rather than whether AI RAN and AI-RAN will be deployed. It also notes differences expected across deployment models, compute architectures, hardware choices, functional splits, and underlying technologies.

Current deployment scope

The article states that most AI RAN activity is centered on distributed AI-for-RAN solutions that target performance and efficiency, often using existing 5G infrastructure. It also reports that Huawei and ZTE have collectively shipped “more than 0.6 M AI-enabled boards/plug-ins.”

Dell’Oro Group says MWC Barcelona results showed nearly all RAN roadmaps now include AI RAN capabilities across the full stack, with a focus on AI-for-RAN. It adds that vendors are extending intelligence beyond baseband, including radios.

Example supplier activity

The article cites Ericsson as launching “ten AI-ready radios featuring in-house silicon with Neural Network (NN) accelerators.” It also characterizes the supplier move as adding intelligence into additional RAN layers beyond baseband.

Forecast and market share view

Dell’Oro Group says its next-generation RAN outlook remains broadly intact after MWC 2026 and Nvidia GTC. It assigns likelihood levels for AI RAN, Cloud RAN, and multi-vendor RAN roles in “the second half of 5G and the early 6G Edge Resource Allocator (ERA)” as high, moderate, and low, respectively.

The firm states that AI RAN is expected to surpass “$10 B” and account for “roughly one-third of the total RAN market by 2029,” describing this as “not new revenue.” It characterizes these figures as part of its “latest forecast update.”

GPU-RAN and operator constraints

The article says prospects for GPU-RAN (and AI-and-RAN) are improving but remain “still small, but no longer negligible.” It attributes the shift to low starting expectations and “a gradual change in sentiment,” moving from “outright skepticism to cautious curiosity.”

Dell’Oro Group cites Jensen Huang’s statement: “That base station…is going to become an AI infrastructure platform.” It also points to early operator progress from T-Mobile, SoftBank, and Indosat and says Samsung and Fujitsu are exploring whether GPUs could diversify computing platforms.

The firm says operators remain cautious about Graphics Processing Unit (GPU) RAN and broad-base AI inference distribution due to constraints including “power budgets,” “strict cost controls,” and “massive scale requirements.” It adds that it is challenging to justify “power-intensive compute at every cell site,” and it cites concerns about “the performance-per-watt gap between GPUs and custom silicon.”

The article includes a SoftBank/Ericsson example in which a robot assistance demo operated with latency requirements “of around 100 ms,” using centralized AI inference with compute in a data center using the User Plane Function. It concludes by stating that AI RAN is moving “from hype toward reality.”

Analyst view on adoption paths

Dell’Oro Group says adoption paths will be shaped by trade-offs across “AI inference distribution needs, flexibility, performance, energy efficiency, Total Cost of Ownership (TCO), and Test Traceability Matrix (TTM).” It also asserts that “the overall direction is clear: AI will become an integral part of every layer of the RAN.”

In base-case projections, the article says non-GPU RAN will dominate AI RAN over the forecast period, citing the ability to upgrade existing infrastructure and constraints at the cell site. It also states that this poses a “meaningful challenge” for NVIDIA if it aims to position itself beyond inference.

The article links operator questions to timing and placement of GPUs, saying the inquiry has shifted from why GPUs matter to where and when they make sense. It also says that if NVIDIA expands the RAN from connectivity to a distributed AI platform, the long-term opportunity could exceed the base-case view.

This Analyst Signals brief reflects a neutral, fact-based summary of the original research note.