Nokia’s AI-RAN Push: Can Operators Really Move at Software Speed?

Why Should a Skeptical Exec Care About Nokia’s Latest AI-RAN Messaging?

In a market already saturated with AI-RAN announcements, Nokia’s July 2026 comments from CTO and AI Officer Pallavi Mahajan stand out for their emphasis on practical execution over hype. Speaking with Fierce Network on July 21, Mahajan framed AI-RAN as fundamentally about decoupling hardware from software—allowing operators to innovate at the pace of software updates rather than waiting for new silicon cycles. This directly addresses operator pain points around capex cycles, brownfield upgrades, and the need to extend existing AirScale deployments into the 6G era.

For execs evaluating multi-vendor strategies, the message is clear: Nokia is positioning its platform as a software-first evolution that leverages existing infrastructure, including GPU acceleration via its Nvidia partnership, without forcing hardware rip-and-replace. Early field trials with T-Mobile in the U.S. and collaborations with SoftBank, Deutsche Telekom, Telia, Orange, BT, and Indosat provide real-world signals worth scrutinizing.

How Does Decoupling Hardware and Software Actually Deliver “Software Speed” Innovation?

Mahajan’s core thesis: once hardware and software are decoupled, innovation accelerates because software can iterate independently. “We are providing innovation at software speeds,” she stated during Nokia’s recent AI-RAN vision presentation. This mirrors server industry dynamics where software advancements outpace hardware refreshes.

In practice, Nokia argues this means delivering measurable RAN improvements—such as spectral efficiency gains—through software algorithms and AI models rather than custom ASICs. Udayan Mukherjee, who leads Nokia’s RAN and Core technologies and joined from Intel, highlighted how AI enables “quantum leaps” in areas like multi-user MIMO that traditional manual tweaking could never achieve at scale. Bell Labs has reportedly worked on these capabilities for four to five years, providing the foundational IP behind current claims.

Skeptics might ask whether this is just marketing. Mahajan counters that the approach already supports brownfield upgrades: operators can add compatible cards to existing racks without altering power or cooling systems, thanks to Nvidia’s RTX 4500/Blackwell-derived solutions fitting within current thermal and power envelopes.

What Specific Spectral Efficiency Gains Is Nokia Claiming, and Are They Proven?

Nokia is touting immediate 20-25% spectral efficiency improvements today, scaling to 50% by the end of 2027 and a full 2x (100%) improvement by the end of 2028—all on the same hardware that will support 6G. These figures stem from a combination of multiple AI algorithms and models tested independently in Bell Labs, not a single silver-bullet solution.

Validation includes lab work in Dallas and field trials in the Chicago area. While exact trial parameters remain high-level, the claims tie into broader AI-RAN goals of optimizing resource allocation and interference management in real time. Analyst Joe Madden of Mobile Experts called the 2x target “very exciting” but noted it remains “completely unproven,” suggesting Nokia has significant work ahead to demonstrate real-world performance at scale. Ian Fogg of CCS Insight emphasized that software upgrades could help operators delay or avoid new cell site additions, directly impacting opex.

What Role Does the Nvidia Partnership Play, and Is It All About GPUs?

Nokia’s $1 billion investment from Nvidia is central, but Mahajan stresses the relationship goes beyond hardware. The partnership enables GPU-accelerated workloads that complement Nokia’s software stack, particularly for Layer 1 and Layer 2 functions that can run without real-time kernels in some configurations. The key differentiator highlighted by analysts is Nvidia’s CUDA software ecosystem, which provides broad industry support and portability.

Fogg noted that focusing solely on GPUs misses the point: operators want cost-effective platforms where software performance boosts can extend the life of existing sites. This aligns with Nokia’s Open RAN and disaggregated strategy, where software updates can deliver value across multi-vendor environments.

How Does This Fit Into 6G Readiness and New Use Cases Like ISAC?

AI-RAN serves as a bridge to 6G by enabling capabilities such as Integrated Sensing and Communication (ISAC) on 5G infrastructure today. Mukherjee noted that sensing demonstrations are already underway, with particular interest from defense and public safety sectors for applications like drone detection and perimeter security.

The same hardware foundation supporting current 5G AI optimizations is positioned to carry operators into 6G without major overhauls. This software-centric path could accelerate feature rollouts, from advanced MIMO to agentic AI workloads at the edge, while maintaining compatibility with Open RAN interfaces and RIC frameworks.

Which Operators Are Most Likely to Adopt This Approach?

T-Mobile is Nokia’s lead U.S. trial partner for Cloud AI-RAN. Internationally, SoftBank, Deutsche Telekom, Telia, Orange, BT, and Indosat are cited in various stages of engagement. Analyst Madden suggested strong candidates include T-Mobile, SoftBank, Elisa, and Indosat, while noting skepticism from Verizon (whose CTO has publicly questioned AI-on-RAN architectures) and AT&T’s shift away from Nokia in some areas.

The appeal lies in potential opex savings through better spectral efficiency and the ability to run AI inference workloads alongside RAN functions on shared infrastructure—opening doors to new revenue streams without proportional increases in site count.

What Are the Realistic Risks and Next Steps for Operators Evaluating This?

Key risks include the unproven nature of the higher-end spectral claims and the need for robust multi-vendor interoperability testing. Integration with existing RIC and rApp ecosystems will be critical for Open RAN operators. Operators should demand detailed trial data, including before-and-after metrics on real networks, and clarity on how software upgrades will be delivered and validated over time.

Next steps could involve participating in ongoing trials, engaging with Bell Labs research outputs, or piloting specific AI optimization use cases on current Nokia deployments. As Mahajan put it, the goal is to let operators “move at the pace of software” while the underlying hardware remains stable through 6G.

This software-first framing differentiates Nokia’s narrative in a crowded AI-RAN field and merits close monitoring by technical and procurement teams planning 2027-2028 network evolutions.

Sources

• https://www.fierce-network.com/wireless/nokia-cto-ai-ran-lets-operators-move-software-speed
• https://www.nokia.com/radio-access/ai-ran/
• https://www.fierce-network.com/wireless/nokia-unveils-ai-ran-platform-big-promises-spectral-gains
• https://ai-ran.org/
• https://www.rcrwireless.com/20260717/carriers/red-hat-outlines-ai-ran