In today’s industry news roundup: UK fibre altnets Zzoomm and FullFibre have agreed to combine their businesses and assets; more big names are putting forward their massive AI datacentre plans; DeepSeek’s low-cost approach to GenAI causes chaos in the AI and chip world; and much more!

UK fibre broadband altnet Zzoomm, which last September ​​hired advisors to help it find potential merger partners in a bid to gain scale, has struck an M&A deal with fellow altnet FullFibre. The combined company will have a network that reaches 600,000 UK properties, more than 65,000 paying customers, and will “be well positioned for further M&A,” according to this announcement from FullFibre, which has previously expanded as a result of a merger with another fibre network builder, Digital Infrastructure. Financial details of the deal were not shared. Zzoomm’s CEO, Matthew Hare, will become executive chairman of the combined company, while James Warner, currently CEO of FullFibre, will be group CEO. “The merged entity combines two complementary networks and operating models,” noted FullFibre. “This will create significant opportunities to accelerate customer growth across a larger footprint, secure funding for new builds and future mergers, and achieve greater operational and financial efficiencies through economies of scale,” it added. The deal is subject to “final documentation execution and regulatory approval” but this is expected in the next few months. Zzoomm’s Hare noted: “We have consistently stated our strategy is to grow organically and by M&A in this fragmented market. [As] an enlarged business with the operational and financial infrastructure, benefitting from the economies of scale, as well as a management team with an exceptional track record, we will have an excellent platform to combine with other altnets in the near future as well as driving organic growth faster.” FullFibre’s Warner added: “With a shared commitment to delivering transformational full fibre connectivity with exceptional customer experience, this deal strengthens our collective ability to grow even faster and seize new market opportunities. For FullFibre, this is our second merger and another milestone in our mission to create a 21st-century digital backbone for the nation. Following our recent successful integration with Digital Infrastructure and BeFibre, this next merger further accelerates our ambitions to lead the market and provide unmatched connectivity to homes, businesses, and wholesale partners alike.”

In what looks like a response to the Stargate Project AI infrastructure plans announced last week by OpenAI, SoftBank and others, Meta’s chairman, CEO and majority owner Mark Zuckerberg, has taken to his company’s Threads social media platform to announce a significant increase in infrastructure investment this year and plans for a datacentre that with processing capacity of more than 2 gigawatts “is so large it would cover a significant part of Manhattan”. In this Threads post, Zuckerberg stated that Meta will “bring online [about] 1 gigawatt of compute” and “end the year with more than 1.3 million GPUs. We’re planning to invest $60 to $65bn in capex this year while also growing our AI teams significantly, and we have the capital to continue investing in the years ahead. This is a massive effort, and over the coming years it will drive our core products and business, unlock historic innovation, and extend American technology leadership.” Forbes noted that the capex figure is way above the $51bn expected by Wall Street analysts and a giant leap up from the projected 2024 full year capex of $38.4bn. 

Meanwhile, reports continue to emerge of how Stargate’s backers intend to fund the giant project. OpenAI and SoftBank are believed to be pumping an initial $19bn each into the planned deployments, leaving the stakeholders some way short of the $100bn they stated was ready to be deployed with immediate effect. In addition, Nikkei Asia reported that Masayoshi Son, the CEO of SoftBank (which is responsible for Stargate’s funding strategy), is to approach investment firms Apollo Global Management and Brookfield about funding the initiative’s AI datacentres. 

While that saga continues, India’s Reliance Industries is keen to play a role in the global race to build staggering volumes of datacentre capacity. According to Bloomberg, Reliance plans to build the world’s largest AI datacentre in Jamnagar with a capacity of 3 gigawatts, a move that will cost up to $30bn. According to Bloomberg, the world’s largest datacentre currently is a 600 megawatts site in Virginia, US. Reliance is one of a number of Indian tech giants (along with Tata Communications and Tech Mahindra) that have already developed relationships with AI chip giant Nvidia, which noted during its AI Summit event in Mumbai last October that Reliance is working with Nvidia to build “AI factories to automate industrial tasks and transform processes in sectors like energy and manufacturing”.

But all of these plans might be subject to revision, in one way or another, as the AI world takes stock of the potential impact of DeepSeek, an open-source, generative AI (GenAI) platform from China that appears to match the capabilities of existing GenAI apps, including OpenAI’s ChatGPT, while reportedly running on far more cost-effective compute infrastructure. The DeepSeek app was launched last week and is already top of the Apple App Store’s ranking of free apps, one ahead of ChatGPT. DeepSeek’s capabilities and its ability to match other apps while running on less advanced chips than its US rivals sent to stock markets into panic mode, reported Bloomberg, with Nvidia’s share price slumping by almost 13% to $124.47 in early trading on the Nasdaq exchange on Monday. As a result of that decline, Nvidia’s market value is only just above $3tn (so it’s not in any danger of collapse…). 

While the launch of new AI apps and models is only going to continue, and potentially cause ructions in the AI tech world, all of these developments spell good news for the telecom sector. All that data traffic needs to be shunted from A to B across data networks, and while the hyperscalers have some of their own network infrastructure, they, as well as the world’s enterprises, will need to rely on data network operators to underpin their operations, which is the exact point Verizon made when it launched its AI Connect strategy at the end of last week. 

Having recently reached a binding agreement to form a fibre-to-the-premises (FTTP) joint venture, two of Spain’s main telcos, MásOrange and Vodafone Spain (now owned by Zegona Communications), are reportedly in talks to form a mobile network joint venture that would combine their radio access network assets, reports Expansion. Any resulting company would then seek to sell a stake to an external investor, which is the same model the operators are deploying with their FTTP venture.   

UK broadband network operator TalkTalk is investigating a potential data breach after a hacker offered access to the details of 19 million current and former customers, according to The Register. 

– The staff, TelecomTV