Spectrum is finite. Mobile data demand is not. That arithmetic has driven up the cost of licensed spectrum for two decades, with U.S. auctions regularly clearing in the tens of billions. The less-discussed shift underway right now is that AI is beginning to dissolve that constraint from the inside, without new auctions, new towers, or new hardware.

What the Data Actually Shows

SoftBank’s field trial of Ericsson’s AI in RAN, conducted on a live 5G commercial network, delivered gains of up to approximately 25% in spectral efficiency and up to approximately 50% in downlink user throughput versus conventional technology. Ericsson launched the commercial software subscription for AI in RAN on June 11, 2026, claiming up to 20% higher downlink throughput and up to 10% better spectral efficiency, based on more than 15 deployments and trials using existing baseband silicon.

Nokia launched its AI-native RAN platform in July 2026, combining Nokia anyRAN software with NVIDIA’s Aerial AI-RAN platform. Nokia says its AI-driven radio technologies have already demonstrated more than 20% improvement in spectral efficiency, with a roadmap targeting 50% by 2027 and more than 100% by 2028, effectively aiming to double the capacity available from operators’ existing spectrum assets.

The market noticed. Nokia stock surged about 6% to around $10.39 after it said its AI-RAN platform advanced to lab and live field trials with eight global operators, spanning North America, Europe, Asia-Pacific, and the Middle East. The move was Nokia-specific. Ericsson was essentially flat on the same session.

The CBRS Angle Traders Are Underweighting

While attention has clustered around Nokia and Ericsson’s licensed-spectrum plays, a parallel market is moving faster. Federated Wireless announced Spectrum AI in June 2026, describing it as the first Physical AI platform purpose-built for shared spectrum planning and coordination. The company said field-validated results across hundreds of live commercial CBRS deployments show operators achieving up to 5x gains in network capacity without adding spectrum, towers, or other infrastructure.

The platform carries a stated 10% to 11% premium on subscription pricing. Federated has said it supports roughly 397,000 CBRS devices and represents approximately 83% of the CBRS market. The company is private, but its commercial scale matters for the listed names: any operator upgrading CBRS efficiency reduces the business case for licensed mid-band capex from the major network vendors.

Sector and Valuation Context

Dell’Oro Group expects the global RAN market to remain broadly flat in 2026, with the top five suppliers by worldwide revenue being Huawei, Ericsson, Nokia, ZTE, and Samsung. In a flat-revenue environment, AI-driven spectral efficiency is not a growth story so much as a margin and market-share story. Rather than requiring operators to find third-party demand for AI processing, AI for RAN generates value from an asset the operator already owns, with immediate return coming from improved network economics and potentially deferred investment in additional sites or spectrum.

Qualcomm’s TTM revenue is about $44.5 billion, with a gross margin around 54%. It trades at a P/E near 20. Its partnership with Ericsson on 6G radio innovations, highlighted around MWC 2026, included cmWave exploration in the 6 to 8 GHz range to highlight enhanced uplink performance in new spectrum.

What Traders Should Watch

Three levels matter here. First, Nokia has framed these engagements as moving from early evaluation into lab and live field trials, with commercialization positioned as the next step. Whether the operator trials convert to contracts with disclosed revenue is the near-term catalyst. Second, Nokia has not put a figure on what the AI-RAN pipeline is worth, which means upside from any initial pricing disclosure is unpriced. Third, the CBRS market’s efficiency gains, if they structurally reduce the demand for licensed spectrum, represent a long-run headwind for auction-dependent capex cycles at Ericsson and Nokia alike.

The cognitive radio bet is not one trade. It is a compression of two different revenue timelines: the immediate margin improvement available through software subscriptions today, and the valuation re-rating that follows when 100% spectral efficiency gains stop being a roadmap target and start appearing in operator financials.