• VMO2 Business and Daisy merge to better target UK B2B sector
• KDDI enters into strategic relationship with DriveNets
• Macquarie is taking full control of Denmark’s TDC Group

In today’s industry news roundup: B2B services specialists Virgin Media O2 (VMO2) Business and Daisy Group form a new company to address the UK’s enterprise services sector; Japan’s KDDI is to work even more closely with disaggregated data networking infrastructure specialist DriveNets; Macquarie Asset Management is taking full control of Danish telco TDC Group by buying out its pension fund co-owners; and much more!

The UK enterprise communications services sector just got a jolt with the news that Virgin Media O2 (VMO2) Business, the enterprise services arm of British mobile and broadband giant VMO2, and specialist B2B service provider Daisy Group are to merge their enterprise services operations to form a new company, with the deal set to close in the second half of this year. VMO2 will hold a 70% stake in the new company, while Daisy Group will own 30%. The new company, which will have a customer base of hundreds of thousands of (mainly small) UK businesses, will have annual revenues of about £1.4bn and adjusted EBITDA of £150m and is to be headed up by Daisy Group founder Matthew Riley, who will be chairman, and Jo Bertram, the current CEO at VMO2 Business, who will take the same role at the new company. According to VMO2, “the new entity will offer significant economies of scale and a range of best-in-class digital-first connectivity solutions and managed services all under one roof. This will include cloud-based communications tools, 5G private networks, IoT [internet of things] connectivity, security solutions and AI-powered products, such as O2 Motion, catering to a broad mix of new and existing customers. The company will be supported by fixed and mobile connectivity wholesale agreements with Virgin Media O2, and supplier arrangements with Telefónica and Liberty Global [the parent companies of VMO2] to leverage high-growth products and services from across the portfolio of those wider shareholder groups.” The mission statement for the new company is to “drive further growth through greater scale, efficiencies and a combined set of products,” which translates as ‘getting bigger and cutting a lot of costs’ to boost margins. The companies expect to realise cost savings of about £600m and that’s likely to mean quite a few job losses. Overall, though, the move makes sense, at least on paper. Daisy has long been regarded as an excellent provider of communications services to business customers, thanks in part to its approach to customer engagement and experience: If VMO2 Business can build on that then BT Business and the enterprise arm of the about-to-be-merged Vodafone UK and Three will have much tougher rival to deal with.  

Japanese telco KDDI is strengthening its ties with DriveNets, the cloud-native virtual routing platform specialist with which it has been working on its next-generation network infrastructure for a few years already. KDDI and DriveNets have struck an agreement “aimed at accelerating the adoption of open network architecture” – also known as disaggregated network technology (the separation of hardware and software) – to “build a flexible network architecture that can quickly adapt to changing market conditions and technological advancements, in order to meet the growing data demands of the AI era.” The telco stated: “This agreement aims to expand the deployment of disaggregated routers, improve the management efficiency of these routers, and optimise the capital investment and operating costs of high-scale network deployments,” noted KDDI, adding that it will “start the deployment of DriveNets Network Cloud solution in backbone core routers at four key locations with the goal of reaching commercial operations by the end of FY2025.” That’s actually a bit later than expected but this is a major transition that takes time and, ultimately, it will be worth it, as KDDI believes the distributed disaggregated backbone router (DDBR) systems that are based on DriveNets software will help it to reduce its power consumption by about 46% and rack space by about 40% compared to the traditional routers currently used in KDDI production networks. You can get the full background to the engagement between KDDI and DriveNets, and the telco’s plans for its “Scale-Out Network” that supports its fixed and mobile traffic, in this TelecomTV article from February. 

Macquarie Asset Management has struck a deal to acquire the 50% of Danish telco holding firm TDC Group it doesn’t already own from Danish pension funds PFA, PKA and ATP, which together took a 50% stake when TDC Group became a private company in 2018. Financial terms were not disclosed. Nathan Luckey, head of digital infrastructure for EMEA at Macquarie Asset Management, stated: “Digital infrastructure and services play a critical role in improving connectivity, driving economic growth and competitiveness, and TDC Group has been at the forefront of delivering this in Denmark. We would like to thank our co-shareholders for their partnership over the last seven years, and we are excited to further increase our stake to full ownership of TDC Group. We look forward to continuing to deliver on our strategy to accelerate the digitalisation of Denmark as TDC Group invests in its networks and technology to further enhance customer service, and network access, reliability and efficiency”. TDC Group comprises TDC Net, which owns and operates a national fixed and mobile network, and Nuuday, a communications services company: When the company was privatised seven years ago it was expected that Nuuday would be sold and that TDC Holdings would become a pure infrastructure company, but that didn’t happen. What did happen though, according to Danish telecom sector analyst John Strand, is that the value of TDC Group has been eroded and the NetCo/ServCo split hasn’t worked out as planned: You can read his lengthy and very negative analysis of the management and ownership of TDC Group since 2018 here. 

SK Telecom (SKT), which has been in the headlines recently as it deals with the fallout from a major data breach, has reported its first-quarter financials, with revenues slightly down year on year to 4.45tn Korean won ($3.14bn) but operating profit up by 13.8% to 567bn won ($400m). The profit improvement came from the company’s AI operations, it noted in this announcement, but this is the calm before the storm as the financial impact from SKT’s data breach is as yet unknown, but it is going to be significant. We’ll have more on the security challenges and fallout from the incident later this week. 

The value of the market for AI chips for datacentres and the cloud will be more than $400bn within five years, according to Cambridge, UK-headquartered research house IDTechEx in its new report, AI Chips for Data Centers and Cloud 2025-2035: Technologies, Market, Forecasts. AI (including machine learning) is incredibly efficient in quickly finding patterns in data that would take many human operatives thousands of hours to even begin to recognise. AI is also so powerful it just about reduces to nil the potential of human errors being introduced and rendering complex datasets into a worthless jumble of gibberish. Indeed, a recent report by McKinsey (we know, we know…) found that it can now automate between 64% to 69% of the total time spent on the collection of data and its processing. However, the trade-off for this remarkable ability is that AI datacentres need enormous power, storage and cooling. The new IDTechEx report forecasts that continuing and growing deployment of AI datacentres, the further commercialisation of AI and the increasing performance demanded by large AI models will further fuel growth of the AI chips market but will also have to evolve quickly to keep up with the enormous demand for more efficient computation, lower costs, higher performance, massively scalable systems, faster inference, and domain-specific computation. Graphics processing units (GPUs) currently dominate the AI chips market and have powered the performance of the best AI systems by providing the compute power needed for deep learning within datacentres and cloud infrastructure. However, as the capacity of global datacentres accelerates to reach hundreds of gigawatts (GWs) of power over the next few years – a gigawatt is a unit of power equal to one billion watts – and investments in AI continue to grow, pressure is on to improve energy efficiency and drive down the costs of chips and other AI hardware. Hitherto, high-performance GPUs from the likes of Nvidia and Arm have been integral to training AI models but they are very expensive and come with a high total cost of ownership. They also come with the risk of vendor lock-in and the danger that they might be under-utilised in some AI-specific operations. Because of these potential downsides, an emerging hyperscaler strategy is to adopt custom AI application-specific integrated circuits (ASICs) that are  designed for a specific task. ASICs are produced from specialist designers, such as Broadcom and Marvell, and have bespoke cores for AI workloads that are cheaper per operation than GPUs, are specialised for particular systems and are very energy efficient. Meanwhile, companies including Samsung, SK Hynix and Micron Technology are providing high-bandwidth memory (HBM) technology designed for high-speed data transfer and lower power consumption. They are particularly suitable for high-performance computing and AI applications. As AI matures it is evident that alternatives and rivals to the currently all-powerful GPUs will emerge and find their niches, but nothing is settled yet.

– The staff, TelecomTV