What’s up with… Vodafone, Nokia, LLMs for telecom

• Vodafone reignites ‘fair share’ debate
• Nokia completes $2.3bn Infinera acquisition
• GSMA Foundry evaluates LLMs for telecom tasks

In today’s industry news roundup: Vodafone is calling for a ‘Framework for Responsible Use of Networks’; Nokia has added Infinera to its optical portfolio; the GSMA is helping to establish the applicability of large language models (LLMs) to telecom sector tasks; and much more!

Ahead of the MWC25 show in Barcelona, Vodafone has reignited the so-called ‘fair share’ debate that is focused on whether or not big tech firms – or ‘large traffic generators’, as the parlance has it – should be making a contribution to telco capex budgets. To fuel the debate that will once again rise to the surface in Barcelona, the giant telco has issued a call for a new Framework for Responsible Use of Networks that would drive action in three areas: 

1. A new industry code of conduct that sets clear, consistent guidelines on optimising internet traffic. If content services are better designed to optimise bandwidth and minimise ‘waste’, this would reduce pressure on networks for everyone’s benefit.
    
2. More flexible network management rules so that operators are allowed to dynamically manage traffic on their networks. This would include applying fair use policies and deploying technical traffic management tools that prevent congestion and deliver a high-quality experience.
    
3. A legislative framework, including a dispute resolution mechanism, to underpin the negotiation of commercial terms between network operators and large content providers. This would create the right economic incentives to improve internet traffic inefficiencies. It could also ensure network operators are fairly compensated for costs incurred in providing traffic conveyance services, including peering, caching, and transit.

Vodafone notes: “Whilst this proposed framework is new, the broader idea is not. It’s been done before. During the Covid-19 pandemic, for example, content providers, telecoms operators and governments pulled together to ensure that society could continue to function. We just need to formalise this concept, and adapt it for the critical situation Europe currently finds itself in. The measures proposed by Vodafone would bring together policymakers, service providers, telcos, and end users to address this complex challenge. Working together, we can realign the incentives that drive traffic growth and ensure a more responsible use of network infrastructure. In return, this would create a more virtuous circle where less traffic growth creates opportunities for telcos to refocus their investment into the transformational digital services that Europe really needs.” You can expect this topic to be raised in public and on stage during the first day of MWC25 when Vodafone Group’s CEO, Margherita Della Valle, is due to take to the podium, along with the CEOs of Deutsche Telekom, Orange and Telefónica, to discuss Europe’s digital future.   

Following clearance from European regulators, Nokia has completed its $2.3bn acquisition of San Jose, California-based optical networking vendor Infinera, a move that “creates an optical networks powerhouse with the scale to accelerate product roadmaps, further expanding Nokia’s ability to help network operators – whether service providers, webscalers or enterprises – unlock the opportunities and meet the network and power demands of the AI era,” the Finnish vendor noted in this announcement. The deal was first announced in June 2024. “I am delighted we have been able to quickly and successfully complete the acquisition of Infinera,” noted Pekka Lundmark, president and CEO of Nokia. “This transaction will significantly improve our scale and profitability in optical networks, and allows us to speed up the pace of innovation to meet the requirements of the AI era. The Infinera acquisition will accelerate our growth strategy in datacentres and strengthen our presence both in North America and with webscale customers,” he added. Federico Guillén, president of network infrastructure at Nokia, stated: “The speed with which the transaction was approved is very positive for Nokia, as is the strong support the deal has received from customers. In welcoming our new colleagues – and the talent and expertise they bring with them – we are creating a new organisation that will be a pace-setter in innovation, offering capabilities across a wide range of optical networking technologies, underpinned by the cutting-edge research of Nokia Bell Labs. Innovation benefits from scale, and the expansion offered by the acquisition means that we will be able to bring more to customers, faster.” The Infinera team will join Nokia’s Optical Networks business headed by VP and general manager, James Watt. Meanwhile, Infinera CEO David Heard will join Nokia’s Network Infrastructure business group as Northern Ireland chief strategic growth officer. “In this position, he will help to set and oversee the implementation of the business group’s growth plans, including specific customer segment strategies, product and market mix, and go-to-market approach across the business group,” noted Nokia. “I am delighted to welcome David to Nokia and to Network Infrastructure. His extensive experience in technology and business strategy implementation will play a leading role in helping our business group seize opportunities in the market and achieve our ambitions across all our markets and business areas,” added Guillén. The incoming CEO said: “From strong growth in the webscale space to service provider successes spanning metro, long-haul and subsea networks, the proven accomplishments of the Infinera team make for an ideal complement to Nokia’s recognised optical network leadership and innovation. I’m excited about the widely expanded opportunities this new chapter opens up and what it means for Nokia and its Network Infrastructure business, and delighted to be joining the team to help accelerate its growth across all customer segments worldwide.” Nokia CEO Lundmark, who announced recently that he is stepping down in April, has been highlighting the growth potential that Infinera will bring to Nokia as it focuses more on the technology needs of datacentre operators: In October last year (during Nokia’s third-quarter webcast) he noted that the market for optical and IP routing technology for the “datacentre fabric” is potentially worth $20bn a year. He stated at the time that Nokia’s focus on the datacentre market, which had already resulted in a key contract win with AI hyperscaler CoreWeave, is now “the very core of our strategy… datacentres will be our number one growth target for the coming years.” Expect to hear more about the impact of this deal during MWC25.   

The sparks of innovation have been flying at the GSMA Foundry as a team of the best blacksmiths from the organisation’s Innovation Hub hammered out an industry-first open-source framework upon which to examine and evaluate the relative merits of large language models (LLM) in real-world telecom use cases. The GSMA Open-Telco LLM Benchmark has been designed to advance AI in telecom-specific applications as the digital service provider (DSP) sector builds its future operations around AI-based solutions. Some LLMs have been shown to be less than optimal when it comes to the complexities of telecom technologies, regulatory compliance and network troubleshooting. Indeed, in recent tests, OpenAI’s GPT4 scored less than 75% on TeleQnA, a comprehensive dataset designed specifically to assess the knowledge of LLMs in regard to the finer points and intricacies of telecom operations. GPT4 did even worse when it came to 3GPPTdocs Classification, a dataset based on 3GPP standards documentation, scoring just 40%. Meanwhile, Microsoft’s PHi2, a very much smaller model, recorded a truly feeble score of just 10% against a benchmark of 500 general maths questions. Louis Powell, head of AI initiatives at the GSMA, commented: “These results highlight the current limitations of AI models in addressing telecom-specific queries… often producing inaccurate, misleading or impractical recommendations. By creating an industry-wide set of benchmarks, we’re not only improving model performance but also ensuring AI in telecoms is safe, reliable and aligned with real-world operational needs.” The launch is supported by several entities, including host company Hugging Face, which might sound like something out of the film ‘Alien’ but is, in fact, a US outfit that develops computation tools for building applications using machine learning. Another supporter is Khalifa University, a research, innovation and enterprise-intensive public university located in Abu Dhabi, United Arab Emirates. Elsewhere, LG Uplus, the South Korean mobile network operator owned by LG Corporation is another, as is the highly respected non-profit organisation the Linux Foundation. Among the mobile network operators and vendors involved are Deutsche Telekom, LG Uplus, SK Telecom and Turkcell, together with giant tech vendor Huawei. Eric Davis, head of the AI tech collaboration office at SK Telecom, noted:  “The introduction of GSMA Open-Telco LLM Benchmarks marks a pivotal milestone for the telecommunications industry in its pursuit of tangible AI benefits. By establishing a standardised evaluation framework, we’re simultaneously driving innovation and ensuring AI solutions deliver the robustness, reliability and precision that our rapidly evolving sector demands.” According to the GSMA, mobile network operators, AI researchers and developers are able to register their interest and submit use cases, datasets and models for evaluation to aiusecase@gsma.com. A standardised benchmarking framework ensures that all AI models are evaluated against real-world challenges in areas such as telecom domain knowledge, mathematical reasoning, energy consumption and safety.  

Elon Musk’s elevated status in the Trump administration is causing friction in lots of different ways as the world’s richest person throws his weight around in an apparent effort to further line his own pockets. One of his latest wheezes is to try to discredit Verizon in order to wrestle away a lucrative government deal for the US telco. In 2023, Verizon Public Sector announced a 15-year critical infrastructure contract with the Federal Aviation Administration (FAA) worth more than $2bn “to design, build, operate and maintain the FAA’s next-generation communications platform”. Under the terms of the Federal Aviation Administration Enterprise Network Services (FENS) contract, Verizon is to “build the FAA a dynamic, highly available and secure enterprise network to support all of the agency’s mission-critical applications across the National Airspace System (NAS), which includes providing Air Traffic Management (ATM) to more than 45,000 flights and 2.9 million airline passengers travelling across the more than 29 million square miles that make up the US national airspace system.” But according to The Washington Post, the FAA is on the verge of cancelling the deal with Verizon and handing it instead of Musk’s SpaceX company, of which the Starlink low-earth orbit (LEO) communications satellite constellation is a part. As The Washington Post notes, “the move to cancel a major contract in favour of a venture led by Musk – who is leading President Donald Trump’s disruptive overhaul of the federal government through the US DOGE [Department of Government Efficiency] Service – would represent a significant test of protections against conflicts of interest in government projects. It would be an especially extraordinary step for the typically cautious FAA, whose systems are vital to the safety of millions of air travellers every day.” Musk added fuel to the fire by posting on X (which, of course, he owns as well) that “The Verizon system is not working and so is putting air travellers at serious risk”. He followed up with another post stating, “To be clear here, the Verizon communication system to air traffic control is breaking down very rapidly. The FAA assessment is single digit months to catastrophic failure, putting air traveller safety at serious risk.” But as Verizon pointed out, the current legacy system is run by L3Harris Technologies and that is the one that Verizon’s system is due to replace, a fact that forced Musk into a correction (but without an apology), stating in this follow-on post – “Correction: the ancient system that is rapidly declining in capability was made [by] L3Harris. The new system that is not yet operational is from Verizon.” How magnanimous. For more on this saga, see this Reuters report.  

Meanwhile… Carlos Slim, who once upon a time was the richest person in the world, has reportedly decided he no longer wants to do business with Musk. According to the Mexico Daily Post, Slim, who founded and still controls Latin American telecom giant América Móvil, took umbrage at an X post by Musk that suggested the Mexican billionaire has ties to narcotics cartels, and has reportedly severed América Móvil’s ties with Starlink, noting that he intends to switch his satellite communications engagements to companies from Europe and China instead of the US. Claro Colombia, part of the América Móvil empire, struck a deal with Starlink last November and Slim’s empire was mulling whether to extend that relationship to other markets. It’s worth noting also that new US tariffs are set to be imposed on Mexico in the coming days, so tensions between the two countries’ big hitters are building. 

With cat-like tread and a happy purr, Amazon Web Services (AWS) has jumped aboard the quantum computing juggernaut with the unveiling of its Ocelot quantum chip, which uses scalable architecture to reduce error correction by “up to 90%”. Error propagation currently plagues the development of real-world quantum computing applications and a global quest is underway to build fault-tolerant quantum computers that can operate for long enough to solve problems of commercial and scientific importance without interference causing the quantum state to decohere and the computation to break down. The Ocelot chip features a novel architecture based on error correction as its absolute basis and on the use of the “cat qubit” named, whimsically but memorably, after Erwin Schrödinger’s 1935 cat-in-a-box thought experiment, wherein a cat exists in two states, alive and dead, simultaneously. That notion led physicists to think about quantum states and the rest, as they say, is history. The new AWS chip, unveiled only days after Microsoft announced its Majorana 1 quantum chip, intrinsically suppresses certain forms of errors and that significantly reduces the number of resources that (currently) have to be devoted to quantum error correction. Amazon says it has succeeded in placing cat qubit technology (existing in two quantum states at the same time) and additional quantum error correction components onto a new form of microchip that can be manufactured to scale via the use of processes commonly seen in the microelectronics industry. The fact is that major advancements in computing have frequently been the result of fundamental changes in hardware (such as microprocessors) rather than software. After all, the transistor made the vacuum tube redundant and, eventually, the size of computers was reduced from the equivalent of a container to the handheld devices of today. Oskar Painter, AWS director of quantum hardware, commented: “In the future, quantum chips built according to the Ocelot architecture could cost as little as one-fifth of current approaches, due to the drastically reduced number of resources required for error correction. Concretely, we believe this will accelerate our timeline to a practical quantum computer by up to five years.” He added: “We didn’t take an existing architecture and then try to incorporate error correction afterwards. We selected our qubit and architecture with quantum error correction as the top requirement. We believe that if we’re going to make practical quantum computers, quantum error correction needs to come first.” He also said that scaling Ocelot to a “fully fledged quantum computer capable of transformative societal impact would require as little as one-tenth of the resources associated with standard quantum error correcting approaches.” Now that is revolutionary. In detail, Ocelot is a prototype non-commercial quantum computing chip, designed to test the effectiveness of AWS’s quantum error correction architecture. It consists of two integrated silicon microchips each with an area of 1cm squared. They are bonded one on top of the other in an electrically connected chip stack. On the surface of each silicon microchip are thin layers of superconducting materials that form the quantum circuit elements. The Ocelot has 14 core components: five data qubits (the “cat qubits”), five ‘buffer circuits’ for stabilising the data qubits, and four additional qubits for detecting errors on the data qubits. The cat qubits store the quantum states used for computation and rely on oscillators to generate a vital repetitive electrical signal with rock-steady timing. The Ocelot oscillators (try saying that after three Negronis!) are made from a thin film of tantalum superconducting material. AWS is, of course, a global force in cloud computing services and would dearly love to be able to add quantum computing services to its portfolio. The Ocelot chips and its successors could also be of great benefit to Amazon’s vast retail business. As Painter observed: “You know, a company like Amazon, you make a one percent improvement… and you’re talking large dollars right? Quantum computers could enable you to do that more effectively, more real time – and that’s the real value.” The Ocelot chip is a major advancement but AWS did not invent cat qubit technology – a company in France called Alice and Bob has been working on them for a few years. The company has manufactured small but commercially available cat qubits and development continues. The AWS peer-reviewed cat qubit research is published in the academic journal Nature. Such developments are, of course, fascinating in the context of quantum computing’s evolution but also of great importance to communications networking companies that are developing, and will need to deploy, quantum-safe infrastructure to protect networks and data from next-generation cyberattacks – see Why telcos need quantum-safe networks.

– The staff, TelecomTV