Uplink growth is real, but the capex discipline holds
In sum – what we know:
- A narrower “no” – DT isn’t dismissing AI traffic, but says any uplink needs can be absorbed inside its existing RAN-modernization roadmap rather than new spending.
- Data, not panic – Ericsson finds uplink now outgrows downlink at 43 of 55 operators measured, but credits collaboration apps and cloud storage, not AI, as the drivers.
- AI spend, different bucket – At its AI Investor Day, DT held its capex guidance while targeting €800 million in B2B AI revenue and €2.5 billion in savings by 2030.
Deutsche Telekom does not expect AI-related mobile uplink demand to require incremental radio-access-network investment in the medium term. That was the message from Dhananjay Mirchandani, the company’s senior VP for group controlling and incoming CFO.
Asked whether AI could lift capital intensity after 2027 through higher uplink demand, Mirchandani said the company’s network organization had not identified a need beyond its existing RAN-modernization roadmap. As reported by Light Reading, his formulation was blunt: “I cannot recall a single conversation in which Alex Jenbar [CTO of Telekom Deutschland] … or somebody else approached us and said, ‘in addition to whatever we currently have on our roadmaps for RAN modernization, that there is an incremental requirement for an investment to be able to prepare ourselves for additional uplink/AI-related traffic.'” He added that the position reflects DT’s confidence in its mobile-capacity capex plan.
To be clear, Deutsche Telekom isn’t saying AI will have no effect on network traffic or that uplink performance is irrelevant. Its position is narrower than that. Any necessary response can be accommodated within planned modernization spending, rather than through a material increase in overall RAN capex. For an operator that simultaneously positions itself as an “AI telco,” that’s a pointed distinction — and one that pushes back on a growing vendor narrative. Equipment makers have been arguing that AI, especially agentic AI, wearables, smart glasses, robots, and other so-called physical AI endpoints, will reverse mobile networks’ historic downlink-heavy traffic profile and trigger a fresh upgrade cycle. Deutsche Telekom, one of Europe’s largest mobile groups, isn’t buying the premise that this translates into a bigger spending envelope. For RAN vendors hoping AI would unlock a new capex wave, it’s a cautionary signal. Rising uplink traffic might drive software upgrades, spectrum-efficiency features, and 5G-Advanced rollouts, but that doesn’t necessarily mean more aggregate spending.
How AI changes uplink demand
Mobile networks have traditionally been built around downlink demand. Whether through video consumption, web browsing, or app downloads — the overwhelming majority of traffic flows from the network to the device. AI interactions can flip that ratio, at least partially. Devices transmit prompts, voice, images, video, sensor data, and contextual information to cloud- or edge-based models for inference, which makes those interactions meaningfully more uplink-intensive than conventional mobile use.
Ericsson’s June 2026 Mobility Report put some numbers to the shift. Of 55 service providers measured, 43 had uplink traffic growth that exceeded downlink growth in 2025. At 17 of those 55, uplink growth was more than 1.5 times faster than downlink growth. But Ericsson also identified the principal current drivers as communication and collaboration apps, user-generated content, and cloud storage — not a broad-based AI traffic surge. The AI component, for now, remains largely scenario-based. Ericsson’s modeling suggests AI traffic could push uplink volumes in 2031 to three times or more the 2025 level, but that’s a forecast built on assumptions about adoption curves and endpoint proliferation, not evidence that operators already need materially higher uplink capex.
That gap between present-day trends and future projections is exactly where Deutsche Telekom’s position sits. Uplink growth is real, and AI will probably accelerate it over time. The question is whether that acceleration demands a new investment tranche or whether it can be absorbed through planned upgrades — things like uplink carrier aggregation, uplink MIMO, and spectral-efficiency improvements delivered via software and 5G-Advanced features. DT’s answer, for now, is the latter.
Network modernization
Deutsche Telekom’s capex discipline makes more sense in light of what it’s already spending on. The company’s German RAN vendor mix has historically been led by Huawei, with Ericsson supplying the balance. DT is now in the process of replacing Huawei equipment with Nokia Open RAN at roughly 3,000 German sites by 2027, according to Light Reading’s reporting, with about 1,000 Nokia sites reportedly live in 2025. That’s a substantial modernization program, and it’s the backdrop against which Mirchandani’s comments were made. AI-uplink requirements, in DT’s view, can be dealt with inside that existing cycle.
At its October 2024 Capital Markets Day, DT guided to investments excluding expenses for spectrum of about 21% of service revenues in 2027, for Deutsche Telekom excluding T‑Mobile US. Mirchandani noted the group was running slightly above that level because of heavy German fiber deployment. The AI Investor Day didn’t revise that guidance. The group’s 2024 plan targeted around €21 billion in adjusted free cash flow after leases in 2027, alongside revenue growth of roughly 4% CAGR for 2023–2027. AI uplink, in the medium-term financial framework, doesn’t move the needle.
What AI could change is the composition of those RAN upgrades rather than their total cost. Uplink carrier aggregation, uplink MIMO, receive diversity, spectrum refarming, and software-defined feature upgrades are all levers that improve uplink performance without requiring fundamentally different hardware investments. The Nokia Open RAN rollout, in particular, gives DT a modernized platform on which to deploy those features as they mature. It’s a story about prioritization within an existing budget, not a story about budget expansion.
T-Mobile US, which Deutsche Telekom controls, has taken a compatible approach. On its July 2026 earnings call, T-Mobile CTO John Saw cited uplink carrier aggregation, uplink MIMO, and transmit switching as 5G-Advanced features that improve uplink performance in anticipation of a possible future AI-driven traffic increase. T-Mobile also said it isn’t yet seeing meaningful AI-driven traffic pressure. Prepare through the normal technology path, without assuming an immediate capacity crisis — that’s the shared DT and T-Mobile playbook. AT&T has been more explicit about needing to “build better upstream,” linking that logic to its acquisition of EchoStar’s 600 MHz spectrum and its selection of Ericsson dual-band 600 MHz radios with eight-receiver uplink-enhancing technology. The contrast isn’t necessarily that AT&T expects an immediate AI uplink crisis while DT does not. Carriers differ in spectrum positions, network architectures, and the degree to which they frame technology choices around future uplink demand. AT&T’s low-band spectrum strategy represents a different path to the same general preparedness, but its messaging has been notably more forward-leaning on the upstream narrative.
AI strategy
The capex restraint on the RAN side becomes more striking when set against Deutsche Telekom’s broader AI ambitions. The event was, after all, an AI Investor Day — and the company used it to lay out an aggressive commercial AI strategy.
DT targets roughly €800 million in AI-related B2B revenue outside the U.S. by 2030. CEO Tim Höttges separately described a wider AI-related revenue opportunity of more than €1 billion outside the U.S. by 2030, implying over €200 million in consumer-related AI revenue potential on top of the B2B figure. On the cost side, DT forecasts AI and automation gross savings outside the U.S. of approximately €1.1 billion by 2027 and €2.5 billion by 2030, both measured against a 2023 baseline. Those are large numbers for any operator, and they signal that DT sees AI as a central part of its financial story.
Höttges also said DT plans to deploy inference data centers within its telecom infrastructure, positioning distributed compute, connectivity, and low latency as part of the operator’s AI value proposition. That’s a bet on the network edge as an AI platform — and it’s consistent with the broader industry trend of operators trying to monetize their physical footprint for AI workloads rather than ceding that value entirely to hyperscalers.
All of which makes Mirchandani’s uplink-capex answer that much more pointed. DT is bullish on AI commercialization, bullish on network-adjacent AI infrastructure, and actively building out inference capabilities. It simply doesn’t see a need, in the medium term, for additional mobile-capacity spending specifically attributable to AI uplink traffic. The money flows toward monetization and efficiency, not toward pre-emptive radio upgrades driven by vendor projections. Upgrades may still come, of course — through later phases of RAN modernization or eventual 6G evolution — but DT’s position is that they’ll arrive as part of the planned technology path, not as incremental spending forced by an AI traffic surge that hasn’t materialized yet.
Table of Contents
