Telco Workforce Tracker, 2Q26: Headcount Still Falling By 2% Per Year, Even as Telcos Accelerate AI Efforts

Why you need this report

Global telco headcount fell 2.1% year over year in 2Q26. That is not new - it has fallen every quarter since 2019, and this is in line with the historic decline. What changed is the reason: operators are now citing AI and automation deployments directly when they explain the cuts, not just cost discipline. Whether the cuts are paying off is a separate question, and MTN Consulting’s own data says the answer is no: across 72 operators, there is no measurable link between cutting headcount and expanding EBIT margin (r = -0.01). Revenue growth predicts margin change; headcount cuts don’t.

We write this report because what telcos say about their workforce is far different than how they act. This edition adds the industry’s first hard comparison of telco headcount against hyperscale headcount; hyperscalers are now running 12.7% above telco employment. The buildout is changing who does the technical work, and this report is the only one tracking both sides of that shift quarter by quarter.

The report is updated every quarter using the same panel and methodology, but with new findings each time. It isn’t last quarter’s numbers with a new date. That’s what makes a standing subscription useful: the value comes from seeing the next change as it develops, rather than looking at one quarter in isolation.

Scope

This study monitors global employment dynamics within the telecommunications operator sector. MTN Consulting covers 145 telcos in its research, including 120 active companies. This “talent tracker” report provides a deep dive analysis of 72 key telcos, who represent roughly 85% of the global market. Data coverage spans from 1Q11 through 2Q26. With the 2Q26 report, we have added one company: SpaceX, with coverage starting in 1Q23.

Introduction: The automation imperative

With global telecom revenue still flat, operators keep shifting from growth targets to cost control. Automation, autonomous networks, and AI remain central to that shift. MTN Consulting’s Telecom AI & Automation (TAIA) module tracks this transition.

Telco headcount has fallen every quarter since 2019, other than the anomalous COVID quarter of 2Q20. The declines have been driven by layoffs, retirement, and attrition. The workforce profile keeps changing too, with demand shifting toward software, cloud, AI, and quantum-computing skills.

Operators are increasingly describing automation progress in specific, measurable terms rather than general AI enthusiasm. Verizon disclosed more than 70 million autonomous network configuration changes in 2025 and said 33,000 staff use Claude Code (Anthropic’s coding tool), targeting Level 4 autonomy under the TM Forum’s Autonomous Networks framework while retaining human oversight for exceptions. T-Mobile’s Chief Network Officer, Ankur Kapoor, claimed Level 4.5 autonomy for Dynamic CX, its AI-powered event-traffic management system.

Other operators are more guarded about how far that progress actually goes. Telstra’s global head of pre-sales, Regan Ireland, said network autonomy “must be earned, not assumed,” and that data quality remains the hardest unsolved problem in network AI. A Fierce Network Research report identified four barriers still blocking operators from an AI-native operating model, starting with network inventory accuracy running at only 50-60%. That’s a concrete reminder that AI-first language from the C-suite runs ahead of what most networks can support today.

Training and upskilling remain essential regardless of how far autonomy has progressed. BT launched an AI-upskilling initiative with Avanade and Microsoft; BT is also the panel’s #8 largest headcount decliner in 2Q26, down 9.5% YoY.

Success now depends on balancing retraining with selective hiring for digital-first roles, the same conclusion as the prior edition - what’s changed is that operators now have a full year of concrete autonomy-level claims to be checked against, not just stated ambitions.

The layoff paradox

Large layoff announcements keep making headlines, increasingly paired with an explicit AI justification. AT&T said in June 2026 that its AI-driven OSS/BSS token-optimization architecture, now processing more than 27 billion tokens per day, is on a path to $4 billion in savings by 2028. Amdocs, a telecom vendor rather than an operator, disclosed AI-related layoffs as part of its OSS/BSS restructuring while also unveiling an AI-RAN blueprint. The same automation story operators are applying to their workforces is now playing out inside the vendor itself.

Among operators, Charter added 1,200 new job cuts following its Cox acquisition, and SK Telecom’s dedicated AI subsidiary offered staff voluntary retirement amid restructuring just weeks after launch. That’s notable because SK Telecom’s group-level headcount still grew 8.0% YoY in 2Q26. The AI unit is shrinking within a growing workforce, rather than driving an overall headcount decline.

MTN Consulting’s own 2Q26 data shows no reliable link between the size of a headcount cut and the size of a margin gain. Telefonica cut headcount 19.3% YoY in 2Q26, the panel’s largest reduction, and its EBIT margin fell 4.3 percentage points over the same period, from 6.7% to 2.4%. That’s the opposite of what a cost-savings narrative would predict. By contrast, Tata Communications, Telecom Argentina, TPG Telecom, KT, and BT all cut headcount and saw EBIT margin rise, led by Tata Communications (+11.8pp) and Telecom Argentina (+7.9pp). But those gains cannot be attributed to workforce cuts alone: across the full 72-telco panel, the correlation between headcount change and margin change is effectively zero (r = -0.01, n=72) , but revenue growth is a good predictor of margin change (r = 0.34, n=72).

This matters for telecom CFOs and labor unions alike: headcount cuts do not reliably raise EBIT margin, even when paired with an explicit AI-savings target. A stated AI-driven cost target, like AT&T’s $4 billion 2028 figure, is an intention, not a result – and companies are rarely held to their commitments. MTN Consulting’s data says the base rate for that intention translating into a margin gain is close to a coin flip: of the 47 panel operators that cut headcount YoY in 2Q26, 26 saw EBIT margin rise and 21 saw it fall or hold.

Analysts should keep asking for evidence, quarter by quarter, before accepting that a given round of layoffs protect margin. By contrast, MTN Consulting’s TAIA research suggests operators with the highest EBIT margins and EBIT per employee tend to reinvest in workforce upskilling rather than cut indiscriminately - indiscriminate cuts risk morale, institutional knowledge, service quality, and brand equity, each of which can hurt profitability over a longer horizon than one quarter’s margin print captures.

Key findings: 2Q26 analysis

1. Global headcount fell 2.1% YoY in 2Q26, more than double the rate in MTN Consulting’s own 72-operator panel

Global telco employment was 4.286 million in 2Q26, down from 4.378 million in 2Q25, a decline of 2.1%. The 72-panel, paired on companies present in both quarters, fell a smaller 0.93% over the same period (3.638M in 2Q26 vs 3.671M in 2Q25).

The gap suggests headcount reduction is running deeper among the roughly 68 smaller and mid-scale operators in the group of 145 (120 active) telcos that sit outside the 72-telco panel.

2. Telefonica cut the most jobs in absolute terms; TPG Telecom cut the most in percentage terms

Largest absolute headcount reductions, 2Q25→2Q26: Telefonica -17,769; BT -7,914; Deutsche Telekom -7,254; AT&T -6,680; Etisalat -4,822; Mobile Telesystems -4,808; BSNL -2,536; Verizon -2,400.

Largest percentage declines, 2Q25→2Q26: Telefonica -19.3%; TPG Telecom -15.6%; Spark New Zealand -13.8%; KT -13.2%; Telecom Argentina -11.2%; Telenor -10.6%; Telkom Indonesia -9.7%; BT -9.5%.

Largest percentage gains, 2Q25→2Q26: KDDI +13.2%; Turkcell +9.3%; SoftBank +6.1%; Singtel +5.9%; Millicom +5.5%; Globe Telecom +4.8%; Telus +4.4%.

3. SK Telecom’s headcount declined 6.2% as its AI subsidiary offered voluntary retirement

SK Telecom’s headcount declined 6.2% YoY in 2Q26, consistent with broader workforce changes underway at the group, including restructuring of its AI operations. In October 2025, SK Telecom’s dedicated AI subsidiary offered voluntary retirement as part of a restructuring, according to Light Reading. The AI restructuring therefore appears to be part of a broader reduction in the group’s workforce, rather than occurring alongside group-level headcount growth.

4. Labor cost per employee rose 8.4% YoY in the panel even as headcount fell

Mean labor cost per employee across the 72-panel was $71.7K (annualized) in 2Q26, up 8.4% YoY. Total panel labor costs (annualized, trailing 4Q) were $228.0B in 2Q26 versus $216.2B in 2Q25, up 5.5%. At the global (G-140) level, labor cost per employee was $62.5K in 2Q26 versus $58.5K in 2Q25, up 6.8%.

Analysis: a shrinking, more expensive-per-head workforce is consistent with either wage inflation and a shift toward higher-paid technical/AI roles, or simply the mechanical effect of a smaller, more senior remaining headcount. The data does not tell us which is driving the increase.

5. Headcount cuts show no measurable link to margin gains - the correlation is effectively zero

Across the 72-panel, the correlation between YoY headcount change and YoY EBIT-margin change is r = -0.01 (n=72) - no relationship, positive or negative. Of the 47 operators that cut headcount YoY, 21 (45%) saw EBIT margin fall or hold rather than rise. Revenue growth is a far stronger predictor of margin change over the same period: r = 0.34 (n=72).

At the aggregate level, global EBIT margin was essentially flat: 15.81% in 2Q26 versus 15.82% in 2Q25, despite the 2.0% YoY global headcount decline.

Implication: headcount reduction is not a reliable driver of margin expansion, either for individual operators or across the panel. Revenue performance has a much stronger relationship with margin changes. This reinforces the finding from the previous edition, with 2Q26 data showing an even weaker relationship between headcount and margins.

6. Labor is a smaller share of the telco cost base than depreciation

Mapping trailing-4Q global revenue ($1,894.7B) to its cost components : labor costs $270.1B (14.3% of revenue); D&A $350.4B (18.5%); other opex $974.7B (51.4%); EBIT $299.5B (15.8%). These four shares sum to 100.0%, confirming they’re mutually exclusive and complete.

Implication: D&A is a larger share of the revenue dollar than labor. Headcount cuts therefore address only a relatively small part of the cost structure, while high capital intensity leaves operators carrying a large ongoing cost burden from their network and other assets.

7. EBIT per employee rose 6.6% YoY, tracking labor cost per employee almost exactly

Global EBIT per employee (annualized) was $69.3K in 2Q26 versus $65.5K in 2Q25, up 6.6% - nearly identical to the 6.9% rise in labor cost per employee over the same period. Whatever is driving the cost side is showing up on the output side too, in roughly equal measure.

The Hyperscaler Crossover

In 1Q11, the telco sector employed nearly four times as many people as the webscale sector. Following years of rapid hyperscale growth and telco consolidation, the two sectors reached parity in 2Q24. As of 2Q26, hyperscale headcount is now 12.7% higher than that of the global telco sector.

Telcos tend to hire lots of people in two groups: network/IT engineers, and sales & customer support staff. Telcos will continue to need people in these areas for many years to come, but the needs are declining. Geographic and scale efficiencies, automation, autonomous networking, and now AI all are allowing the telco workforce to do more with less. AI may facilitate some of these changes, but it is not the main driver. Telcos have been using automation to do more with less (staff) since well before the first Lucent 5ESS digital switch was deployed in 1982 in Seneca, Illinois.

By contrast, hyperscalers continue to branch out and have more diverse hiring needs. They do hire plenty of software engineers, but that’s not all. Some hire lots of logistics and fulfillment staff; some hire retail specialists. All key hyperscalers spend heavily on R&D, and in a number of different areas: robotics, drones, aerospace, quantum computing, gaming. Nowadays there is high demand in areas like chip and DC infrastructure design, cloud platform development, AI model training, etc.

At the same time, the hyperscalers have always aspired to downsize their workforce when possible. That’s why, for instance, Amazon has been investing in robotics since its 2012 acquisition of Kiva. Now there is more of a push to downsize, for two reasons. First, hyperscalers are spending so much on capex, that they need to cut operational expenses. Second, they need to show that they can “take their own medicine”. After all, they are all pushing the world to adopt AI as fast as possible, and they need to show that this approach can work. Meta’s big May 2026 layoff announcement is an example. There is a chance that Meta’s aggressive layoff strategy will be as successful as its rebranding to Meta in 2021.