What’s up with… T-Mobile US, optical sector growth, Perplexity

• T-Mobile US flexes its 5G SA muscles with L4S
• AI-related investments are fuelling optical tech sector growth
• AI search specialist Perplexity strikes more telco partnerships

In today’s industry news roundup: T-Mobile US is making the most of its 5G standalone capabilities by deploying L4S to support and enable low-latency services; increasing investments in datacentre interconnect (DCI) infrastructure, fuelled by increasing AI workloads, will help drive steady growth in the optical networking tech sector, reckons Dell’Oro; Perplexity strikes partnerships with Telecom Italia and Telefónica Tech; and much more!

T-Mobile US, the ‘uncarrier’, delights in burnishing its reputation as an industry iconoclast as it takes a pop at its venerable domestic competitors AT&T and Verizon. Thus, in a new blog, Unpacking 5G advanced, authored by the operator’s CTO John Saw, it gleefully rubs the noses of its rivals in the dust when he claims that T-Mobile US’s 5G capabilities leave them lolloping along in its wake as tired also-rans. Saw makes no bones about it, saying that T-Mobile US was savvy enough to deploy a cloud-native 5G standalone (5G SA) from the very outset of the availability of the technology, and is now reaping the benefits in a way its rivals are not. 5G SA permits the full potential of 5G to be realised, including network slicing (where network operators can dedicate network segments to specific needs), ultra-low latency and enhanced capabilities for massive IoT (Internet of Things), faster speeds and better security. As part of T-Mobile US’s nationwide rollout of 5G-Advanced services, it is deploying Low Latency, Low Loss, Scalable Throughput (L4S) capabilities. L4S does not increase 5G download speeds – instead, it consistently delivers low latency, minimal packet loss and real-time responsiveness, even when traffic loads are heavy. AT&T and Verizon are unable to do the same because the full panoply of L4S benefits can only be deployed by operators with a 5G SA core platform. Saw writes: “Because everything flows through a single, optimised 5G core, we can deploy programmable features like L4S at scale while laying the groundwork for advanced network slicing and differentiated service tiers We’re also testing L4S to demonstrate how exposing real-time network data can help our partners and other innovators fine-tune their app experiences. That’s where the full value of this work starts to show.” What L4S does is very quickly to detect imminent network congestion whilst it is building, but before any data is lost or delays start to become evident. It does so by marking some data packets with ECN (Explicit Congestion Notification) bits that warn delays are likely to occur. Devices and servers then respond to this warning to ensure that data continues to reach users quickly and evenly, without stuttering, without dropouts and without freezing. Forewarning ensures that fluctuations or jitter in data flow are minimised and the connection remains intact and stable. Among the services and applications that benefit from L4S are cloud gaming, video conferencing, augmented reality (AR), virtual reality (VR), teleoperated driving (remote control of vehicles), and real-time communication in industrial settings. T-Mobile reports that it will soon be making the technology available to enterprise customers across a wide range of industrial sectors.

Increasing investments in datacentre interconnect (DCI) infrastructure and capacity will help the optical transport networking technology sector return to growth and make it a $19bn sector by 2029, according to a new market forecast from Dell’Oro Group. The research house notes that the optical transport networking technology sector, which shrank in value by 9% last year, is set to grow at an average rate of 5% for the next five years and be worth about $19bn in 2029. “We are anticipating that the time has come to interconnect all those new AI datacentres being built,” stated Dell’Oro VP and analyst Jimmy Yu. “We are forecasting datacentre interconnect {DCI] to grow at twice the rate of the overall market, driven by increased spending from cloud providers. The direct purchases of equipment for DCI will encompass ZR/ZR+ optics for IPoDWDM, optical line systems for transport, and DWDM systems for high-performance, long-distance terrestrial and subsea transmission,” added Yu. Dell’Oro’s press release didn’t note any vendors that are particularly likely to benefit from the market growth, but Ciena and Nokia, which recently completed the acquisition of Infinera, as well as Huawei, ZTE and Fujitsu’s new standalone subsidiary 1Finity are well placed to grow their optical network product revenues as a result of the forecasted trend. 

We reported earlier this week that generative AI (GenAI)-enabled conversational search engine developer Perplexity has been ramping up its engagement with the telecom sector and it has continued to do so with two new partnerships with major telcos. In Italy, Telecom Italia (TIM) is offering all of its fixed and mobile consumer customers one year’s free access to Perplexity Pro, “one of the world's most advanced generative Artificial Intelligence platforms,” the operator noted in this announcement. The service would normally cost €20 per month. “This initiative puts artificial intelligence at the service of people, making it truly accessible to all TIM customers,” stated Andrea Rossini, Chief Consumer, Small & Medium & Mobile Wholesale Market Officer, at Telecom Italia. “We are proud to be the first operator in Italy to offer an advanced AI assistant like Perplexity Pro, which marks a tangible step towards a new model of the digital relationship. Innovation, for us, means simplifying and improving the everyday experience, with solutions that are useful, reliable and affordable for everyone.” In addition, Telefónica Tech, the Spanish telco’s digital business unit, is partnering with Perplexity to provide Perplexity Enterprise Pro to enterprise users in Spain, with business customers able to try out the service for three months for free as part of a promotional pilot service. 

BSS vendor Optiva is in talks to be acquired, the company revealed as it struck a deal with its lenders for a 45-day “grace period” regarding the repayment of $108.6m, plus $5.2m in accrued interest, that was due to be paid on 20 July. The grace period allows “Optiva's special committee of independent directors to conclude negotiations with noteholders and prospective merger counterparties regarding a potential transaction” with an as yet unidentified “third party” that would deal with the outstanding debts. “We're thankful for the trust our customers place in us and our team's unwavering dedication during this exciting chapter," stated Optiva CEO Robert Stabile. "The strong momentum we're seeing with new customer wins and product adoption reflects our world-class team and the powerful innovations we've delivered. As we finalize our future ownership structure, we're energized by what lies ahead and confident that our customers will benefit from an even stronger, more dynamic Optiva," he added. The vendor has $12m in cash to “ensure business continuity” in the meantime. For the first quarter of this year, Optiva reported revenues of $11.6m and a net loss of $2.3m. 

Lynk Global, the low-earth orbit (LEO) satellite firm that aims to enable direct-to-smartphone communications from its constellation of “cell towers in space”, has announced the “mutual termination of its previously announced Business Combination Agreement (BCA) with SLAM Corp., a publicly traded special purpose acquisition company [SPAC]. Lynk Global announced its intention to become a listed company by merging with SLAM in February 2024 in a move that would have valued the company at $800m, but the deal turned sour and resulted in some spiteful legal action, which has apparently all been sorted out. “The parties have jointly agreed to terminate the agreement with immediate effect and have also reached a comprehensive settlement of all pending claims and counterclaims in the Delaware Court of Chancery,” noted the satellite operator. Lynk Global CEO Ramu Potarazu noted: “With the Delaware litigation resolved and the BCA mutually terminated, Lynk is now better positioned to pursue a broader set of strategic and commercial opportunities that were previously constrained by the agreement. We remain fully focused on executing our long-term vision and, in partnership with the newly merged SES and our global MNO partners, accelerating our mission to deliver mobile connectivity to anyone, anywhere – directly from space.” 

Communications platform-as-a-service (CPaaS) specialist Sinch has reported a 2% year on year increase in organic revenues for the second quarter of 2025 to 6.6bn Swedish krona ($690m) and a 4% year on year increase in earnings before interest, tax, depreciation and amortisation (EBITDA) to SEK760m ($79m). The financial news comes shortly after Sinch, which has more than 175,000 enterprise customers, announced that it has implemented Model Context Protocol (MCP), “ensuring AI agents can successfully execute communication actions” through its global Customer Communications Cloud platform. “As MCP emerges as a standard for how AI agents perform various actions, from database queries to API calls, Sinch's implementation specifically enables these agents to handle the communication components of their tasks effectively,” the Swedish company noted in this announcement. 

News of yet more remarkable progress towards properly fault-tolerant quantum computing and quantum communications. Researchers have just demonstrated the reality of a process on logical qubits called “magic state distillation”. The theory has been around for a generation but only now has it proven possible to demonstrate it in practice. Magic state distillation is a filtering process, via which more accurate quantum states are created from a mass/mess of ‘noisy’ ones. It is somewhat analogous to the production of a very high-quality single malt whisky, but without the resultant distinctive aroma and rounded taste. Logical qubits are collections of entangled physical qubits sharing the same data. They are based on the notion of redundancy whereby, if one or more physical qubits in a logical qubit fail, the calculation is not disrupted because the same information exists elsewhere. In fault-tolerant quantum computing it is used to mitigate the effects of noise and errors that naturally occur in physical qubits. In the research, the highest quality “magic states” were filtered and “purified” into “higher-fidelity” versions to enable them to be used in the most complex quantum algorithms. Hitherto, it has been possible to apply the process to notoriously error-prone physical qubits, but now, as the demonstration showed, it can be applied to logical qubits as well. Essentially, multiple physical qubits are used to encode a single logical qubit, allowing for error detection and correction without disturbing the logical qubit's state. Commercial quantum computers with greatly enhanced fault-tolerance will be able easily to outperform even the most powerful “classical” supercomputers. The latest development was achieved by a team of researchers at QuEra Computing, a quantum computing company based in Boston, Massachusetts. The company develops unique quantum computers using neutral atoms based on research conducted at the Harvard University and the Massachusetts Institute of Technology (MIT). Neutral atom quantum computing is touted as being an important architecture of choice for future large-scale, fault-tolerant quantum systems. The results of the research are published in the 14 July, 2025 issue of the journal Nature (subscription required). Commenting on the advance, Yuva Boger, the chief commercial officer at QuEra, stated: “Quantum computers would not be able to fulfil their promise without this process of magic state distillation. It's a required milestone.” Sergio Cantu, the VP of quantum systems at QuEra, added: “For quantum computers to be useful, they need to run fairly long and sophisticated calculations. If the error rate is too high, then this calculation quickly turns into mush. The entire goal of error correction is to lower this error rate so you could do a million calculations safely.” However, the error-correction codes that are applied to logical qubits run via “Clifford gates” (which are described as “basic operations in quantum circuits”) but by exploiting the high-quality “magic” states, “non-Clifford gates” can enable true parallel processing, but require expensive and highly complex protocols such as magic state distillation and injection. So, “magic quantum states” are vital to the provision of capabilities beyond supercomputing, but there’s a cost. As Cantu puts it, “magic states allow us to expand the number and the type of operations that we can do. So practically, any quantum algorithm that's of value would require magic states.” And that will put up prices, in the short term at least.

The UK government has signed a Memorandum of Understanding (MoU) with OpenAI (the company behind ChatGP), to use its artificial intelligence technology, software and systems in the UK’s public services sectors, including defence, education, justice and security. What could possibly go wrong? The rationale behind the Labour administration’s decision is that the new regime will “drive change and economic growth”, although, of course, some civil service jobs currently done by humans will go once the machine takes over. The BBC reports that the deal was signed by OpenAI and, on behalf of the UK’s Department for Science, Innovation and Technology (DSIT), by the current Secretary for Technology, Peter Kyle MP. OpenAI will be laughing all the way to the bank. It gets wide access to government data via an “information sharing programme.” Meanwhile, the British state will use OpenAI to "improve understanding of capabilities and security risks, and to mitigate those risks" whilst developing “safeguards that protect the public and uphold democratic values.” Opponents of the government’s decision to do a deal with a US company that will permit it virtually untrammelled access to data held on British citizens, point out that the administration is making parts of the public sector technologically dependent on private companies headquartered overseas even as politicians try to place them under meaningful legislation and regulation. The UK government is making much of its intent to use AI to kick-start the nation’s moribund economy. Currently it is growing by somewhere between 0.1% and 0.2% and stagflation looms. As was to be expected, Sam Altman, OpenAI’s CEO, commenting on the signing of the MoU, airily pronounced that said the agreement will "deliver prosperity for all" because “AI is a core technology for nation building that will transform economies and deliver growth.” The news comes as Altman announced that ChatGPT will have in use more than 1 million GPUs by the end of this year. That inventory alone will be worth well over $3tn. He wants that figure to rise to 100 million GPUs during the next few years. As for the MoU? The sting’s in the tail. It states that OpenAI and the government “will collaborate to identify opportunities for how advanced AI models can be deployed throughout government… to help civil servants work more efficiently” and help “citizens to navigate public services more effectively”. Big Brother will be watching you, from a server farm somewhere in Arkansas.

– The staff, TelecomTV