Nokia plans to start supplying GPUs to telecom operators, staking a position in the emerging market for AI-powered radio access networks. The move puts the Finnish vendor in direct competition with a new class of infrastructure players, as mobile operators weigh whether intelligence at the network edge justifies a costly refresh of radio hardware.

The announcement, reported by Light Reading, signals Nokia's willingness to gamble on AI-RAN, an architecture that folds GPU-based computing into the radio access network itself. Instead of treating AI as a software layer bolted onto conventional infrastructure, AI-RAN proposes that baseband processing and AI workloads share the same silicon at cell sites and central offices.

Why GPUs in the RAN?

The argument for placing GPUs inside radio networks rests on two ideas. First, AI models can manage radio resources — scheduling, beam steering, interference suppression — faster and more finely than rule-based software. Second, the same compute capacity could host third-party AI inference services near users, opening a revenue stream that today's RAN equipment cannot touch.

For Nokia, the calculus is commercial as much as technical. Its traditional RAN business faces stagnant operator capex in mature markets and fierce price competition from Chinese rivals. GPU-based products could carry higher margins and align the vendor with the datacenter buildouts now absorbing a growing share of the AI economy.

A Gamble, Not a Guarantee

Light Reading's framing of the move as a "high-stakes gamble" captures the risk. Operators have spent the past five years consolidating RAN vendors and pressing down equipment prices through the Open RAN movement. Asking them to add accelerator-class hardware to cell sites reverses the cost-reduction logic that has dominated network procurement since 5G deployment began.

Power consumption is the second open question. GPUs draw far more energy than the purpose-built baseband chips they would supplement or replace, and site power budgets are already strained across European and North American networks.

A third uncertainty concerns the ecosystem. AI-RAN has no settled standards posture yet, and operators will want proof that GPU-equipped networks deliver measurable gains in spectral efficiency or operating cost before they commit to volume orders.

What Comes Next for Nokia

The report positions Nokia as the first major Western RAN vendor to commit to distributing GPUs as part of its network portfolio. That first-mover status carries exposure: if operators decide AI workloads belong in centralized datacenters rather than distributed cell sites, the investment in GPU distribution channels and integration work could sit idle.

Nokia's challenge over the coming quarters will be converting the AI-RAN concept into reference deployments with named operator partners, measured throughput or efficiency results, and a credible power story. Until those elements appear, the GPU initiative remains a strategic bet whose payoff depends on when — and whether — the industry's AI-native network ambitions translate into purchase orders.