The market for AI-integrated radio access networks will reach $35 billion by 2030, according to research firm Dell'Oro Group, in one of the first published projections treating AI-RAN as its own revenue category.

The forecast offers a single endpoint and a single number. It does not name the base year, the implied compound annual growth rate, or how Dell'Oro separates hardware from software in its model. For a category that barely existed on vendor balance sheets two years ago, that level of opacity is the story.

What is AI-RAN actually?

AI-RAN refers to the embedding of machine-learning inference — and in some designs, training — inside the radio access network. That is the layer of a mobile network that connects user devices to the core.

Vendors pitch the approach as a way to improve spectral efficiency, cut energy use, and let operators run non-telco AI workloads on the same compute platform that handles radio processing. The acronym has only entered industry vocabulary in the last 24 months. No settled definition exists for which products qualify.

Why does a $35 billion number carry weight?

Dell'Oro is the firm most operators and vendors turn to when sizing telecom equipment markets. Its quarterly RAN reports set the baseline for capex discussions across the industry.

A first-time forecast of this size, if it holds up, would make AI-RAN one of the fastest-growing categories the firm has ever tracked. The headline is therefore the kind of figure that ends up in investor decks, board presentations, and regulatory submissions for the rest of the decade. Spread evenly across the six years to 2030, the $35 billion works out to roughly $5.8 billion a year — a sliver of total telecom capex, even at the endpoint.

What is the forecast not saying?

The projection carries no vendor split, so readers cannot tell whether Dell'Oro expects the revenue to flow to traditional RAN vendors, to silicon suppliers, or to hyperscalers building AI infrastructure for carriers. It offers no regional breakdown between North America, Europe, and Asia-Pacific — the three regions where 5G mid-band rollouts are most advanced.

It does not state whether the figure includes services and integration, or only equipment. And it does not address how Dell'Oro treats open-source AI-RAN efforts that generate no vendor revenue at all.

None of these omissions is unusual for a first-cut forecast. All of them will determine whether the $35 billion number ages well.

What does the operator side look like?

The figure measures what vendors sell, not what operators spend. Carriers that build AI-RAN capabilities in-house — using their own data science teams and open-source toolkits — will incur costs that never appear in any vendor's revenue line.

Conversely, operators who buy AI features as part of a standard baseband upgrade will see the spending land in their RAN capex column without any separate line item. The $35 billion figure does not distinguish between those two outcomes.

What should readers watch next?

The next test for the Dell'Oro projection will be the firm's first AI-RAN-specific update, expected alongside its standard RAN quarterly.

If that report begins to break out base year, regional mix, and vendor share, the $35 billion endpoint will become a usable planning tool. Until then, it is a direction of travel — a marker for where the industry says it is heading, not a measured distance to a known destination.