Industry News

News in Brief: Glasswing takes flight in telecoms, AI RAN runs without GPUs and turns out AI is not responsible for mass firings

In May, Verizon became the first telco to join Anthropic’s ‘good fight’, MTN Consulting debunked myths around AI and reducing headcount, T-Mobile ran AI RAN without GPUs as D2Ds drive NTNs. By Contributing Editor, Annie Turner

Verizon became the first telco to join Anthropic’s Glasswing Project which was launched in April. Project Glasswing was created after Anthropic announced its most advanced AI model, Claude Mythos Preview, can identify complex vulnerabilities in critical software and operating systems at unprecedented speed, including code that is decades old.https://www.anthropic.com/news/expanding-project-glasswing

Through Project Glasswing, Anthropic grants access to Claude Mythos Preview to a vetted, trusted group of companies dedicated to protecting essential infrastructure and committed to sharing information and best practices across industries. At launch it had about 50 partners including Amazon Web Services, Microsoft, Apple, Google, Nvidia, CrowdStrike and Palo Alto Networks.

Telcos AT&T, SK Telekom and BT have also joined the partners whose ranks how now swelled to about 200 after Anthropic announced it would expand the initiative in early June.

T-Mobile trials AI-RAN without a GPU

T-Mobile announced that tests of Ericsson’s link adaptation software in its live 5G Advanced network boosted network performance and spectrum efficiency. This is the first large-scale trial of Ericsson’s software on live, production network traffic. It was carried out across some 43 sites in locations, including Los Angeles, New York, New Jersey and Salt Lake City.

The software uses a neural network, running directly on the Ericsson hardware to predict changing radio conditions in real time. T-Mobile said spectral efficiency increased by up to 10% while downlink throughput was up to 15% higher versus “rule-based methods”. The Uncarrier says it aims for “commercialization of the technology” in the third quarter of this year.

T-Mobile 5G Advanced network achieves world first with Ericsson AI RAN innovation

Grant Castle, SVP of RAN Engineering & Emerging Technologies at T-Mobile, said, “Following our milestone as the first US operator to deploy 5G Advanced nationwide in 2025, we’re continuing to push the boundaries of RAN innovation.

“Our work with Ericsson on AI-native Scheduler with Link Adaptation demonstrates how real-time, AI-driven optimization can enhance spectral efficiency and throughput while delivering a more consistent experience for customers at scale.”

All of which is chewy food for thought for NVIDIA, the biggest maker of AI chips, whose CEO has been talking up physical AI and the case for GPUs in the RAN for months. Last October Nokia announced that NVIDIA is to invest $1 billion in it, to accelerate AI-RAN innovation and lead transition from 5G to 6G.

Commentator Sebastian Barros’s post, To GPU or not to GPU, That is the question, is a good read [subscription needed to access the full article].

AI doesn’t justify mass layoffs which often don’t improve margins

MTN Consulting published some eye-opening analysis that contained some surprises. The first is that in early 2011, telcos globally employed nearly four times more staff than the webscale sector. Since the end of 2025, hyperscalers’ headcount is 3.5% higher.

MTN Consulting: Telco Talent Tracker, 4Q25: Telco headcount erosion continues at ~2% per year. Don’t blame AI (yet?).

The strongest players are shifting from unrealistic growth targets to aggressive cost control, leveraging automation, autonomous networks and, more recently, AI, according to MTN Consulting’s Telecom AI & Automation (TAIA) module. This is backed up Verizon saying on its 4Q25 earnings call that it aims to be the industry’s “most efficient telecom company” [an ambition shared by the Vodafone Group, see this FutureNet World article] and an AI-first company deploying AI at scale.

However, MTN Consulting finds that AI does not justify mass sackings like Verizon promised late last year – to cut 15% of its workforce – or those of AT&T, BCE, T-Mobile US, and Charter/Cox in the Americas; BT, Telefonica, and Vodafone in Europe; and Telstra in Asia-Pacific AI.

Further, although operators often justify these cuts as necessary for competitiveness and profit, MTN Consulting’s data shows no direct link between headcount reductions and greater margins, even after multiple quarters, whether measured by EBIT or EBITDA.

Some operators, including KPN, Telenor and Deutsche Telekom, did improve EBIT while reducing headcount, but those gains also involved asset restructuring or revenue stabilisation, not just firing people.

Matt Walker, Chief Analyst at MTN Consulting stated, “Analysts should stop accepting claims that layoffs are needed to protect margins without clear evidence afterward”. The analyst house warned that indiscriminate cuts can erode morale, institutional knowledge, service quality, and brand equity, hurting long-term profitability.

Telcos that rush to cut staff in response to AI may also create talent gaps that increase cybersecurity risk, churn, and lost innovation.

D2D is driving the non-terrestrial networks

The Global mobile Suppliers Association’s (GSA) report on Non-Terrestrial 5G Networks and Satellite Connectivity found that by April this year, there were 275 publicly announced operator-satellite partnerships in 101 countries and territories.

(Source: T-Mobile)

Some 57 operators in 35 countries and territories have launched commercial offerings, and of them, 21 partnerships in 17 countries have launched satellite-to-phone services – widely known as direct-to-device or D2D – which are driving NTN momentum.

5G NTN is widely adopted to provide fixed broadband, predominantly in rural and remote areas. It makes up 38% of all identified partnerships, down from 45% in the previous edition because of the growing momentum of satellite-to-smartphone technology.

As of mid-April 2026, GSA reports that 23 partnerships have launched services; two more than in February 2026. Another seven are in the evaluation, testing, trialling or licensed phase, and 75 more have services planned.