Nokia and NVIDIA: AI-RAN Revolution in Telecoms

Nokia and NVIDIA: An Alliance to Transform AI-RAN Radio

On July 15, Nokia unveiled its new AI-RAN platform, marking a major turning point in the telecommunications industry. This platform, developed in partnership with NVIDIA, is built on Nokia's anyRAN software and NVIDIA's Aerial system. It promises to optimize radio spectrum usage, enabling telecommunications operators to maximize the capacity of their existing infrastructures. Nokia presents this development as a significant advancement in radio architecture, a first in decades.

Nokia claims that its platform has already demonstrated a spectral efficiency improvement of over 20%. The goal is to reach 50% by 2027 and exceed 100% by 2028. These figures, while promising, remain projections. Nokia plans to launch pilot projects before the end of the year, with commercial availability expected in 2027. This approach aims to allow operators to double the capacity of their current spectrum without requiring massive new hardware investments.

Operators will be able to access this new capacity through a software subscription, thus avoiding the need for costly hardware renewals. Three deployment options are offered: a GPU extension card for existing AirScale sites, a standalone AI-RAN node, and a cloud-based solution provided by partners.

A Crucial Challenge for Nokia

This launch is not just a simple product innovation for Nokia. Radio has long been a challenge for CEO Justin Hotard, who took the helm of the company in 2025. During Nokia's capital markets day in November, Hotard admitted that the mobile business had not generated the expected returns. In response, he integrated this business into a new Mobile Infrastructure segment, accompanied by cost-cutting measures.

The partnership with NVIDIA, announced in October 2025, is central to this turnaround strategy. NVIDIA invested $1 billion to acquire about 3% of Nokia. By leveraging NVIDIA's silicon and CUDA software, Nokia can reduce its internal R&D expenses, focusing more on software development. This transition marks a departure from the traditional hardware model.

Investors have welcomed this strategy. Nokia's shares experienced positive revaluation throughout 2026, driven by its momentum in AI and cloud. The launch of AI-RAN occurred just before the announcement of second-quarter results. According to Omdia and its analyst Rémy Pascal, the cumulative opportunity for AI-RAN could exceed $200 billion by 2030. The question remains how much of this market Nokia will be able to capture.

Competition from Ericsson

Nokia's claim to be the first to launch an AI-RAN platform requires careful scrutiny. In June, Ericsson launched a commercial software subscription called AI-in-RAN. This service promises a 20% increase in downlink throughput and a 10% improvement in spectral efficiency across more than 15 live deployments. Unlike Nokia, Ericsson's solution operates on existing baseband silicon from operators, without requiring a GPU.

Nokia distinguishes itself with its GPU-accelerated AI-RAN platform, a different approach from AI solutions layered on existing hardware. Both companies can claim innovations, but Nokia's formulation deserves deeper analysis.

The divergence between Nokia and Ericsson is not limited to timing. Nokia has chosen to partner closely with NVIDIA, while Ericsson has opted for independence from silicon, thus avoiding any hardware dependency.

Nokia uses commercial silicon from Marvell in its broader ecosystem and describes its platform as compliant with Open RAN. However, the spectral efficiency gains it offers currently rely on NVIDIA's architecture, with no equivalent alternative available. This dependence on NVIDIA is both a strength and a weakness of Nokia's strategy.

While this strategy is not necessarily flawed, it relies on outsourcing the silicon race to a dominant industry player. The subscription model provides Nokia with recurring revenue, an advantage that traditional hardware cycles could not offer.

However, the platform is not yet commercially available, and its key efficiency figures will only be achieved in two years. Moreover, Ericsson has already gained a lead in the market through a different path. For Nokia, this launch represents a comeback, but not yet a guaranteed victory. Its future trajectory will largely depend on its partnership with NVIDIA.