Deutsche Telekom executives used the operator's AI investor day in Bonn to commit to €2.5 billion ($2.8 billion) in indirect cost savings by 2030, alongside €800 million ($899 million) in AI-related revenues outside the United States — a plan that left the share price up just 1% in Frankfurt on the day, having fallen 7.4% over the previous twelve months.

What did Deutsche Telekom actually show investors?

CEO Timotheus Höttges opened the session with a deepfake video that placed his likeness at the fall of the Berlin Wall, the operator's 1996 stock-market listing, and the 2007 iPhone launch. The film even showed him abseiling from a magenta-painted telecom mast.

"You see what is possible with AI," Höttges told the audience. "The language, the pictures, the movements — everything was copying me already. To a certain extent, it's scary as well."

How big are the savings — and how realistic?

The €2.5 billion target equals less than 3% of Deutsche Telekom's full 2023 operational costs, including interest, tax, depreciation and amortization. The €800 million revenue goal sits below 2% of the group's non-US 2025 turnover.

The operator finished June with fewer than 192,000 full-time employees, having cut roughly 34,500 jobs since 2020. Personnel costs hit €19.8 billion ($22.3 billion) in 2024.

The savings target represents about 12.6% of that payroll bill, meaning up to 25,000 roles could be at risk if automation absorbed the entire reduction.

Deutsche Telekom's investor presentation makes no overt reference to headcount targets. But the emphasis on "automation" implies that manual work — and the staff doing it — would carry most of the burden.

Where will the money go?

Deutsche Telekom and Nvidia are committing €1 billion ($1.1 billion) to a single AI factory in Munich. The operator also plans to use AI to streamline IT systems, shrink its real-estate footprint, and trim energy use across network infrastructure.

What about the cost of the AI itself?

Deutsche Telekom's €2.5 billion is a gross figure that excludes token costs — the per-query fees charged by AI model providers. A 1,000-word essay consumes roughly 1,500 tokens, and current user payments fall short of the energy and data-centre expenses required to generate them.

A presentation slide suggests the operator believes it can "manage" token costs by avoiding "model lock-in" and applying "tiered token budgets," capping tokens at a "low double-digit percentage of gross savings."

Even at that ceiling, at least €250 million ($281 million) would come off the net savings number.

Management consultancy Bain expects global annual AI capital expenditure to reach $1.5 trillion by 2031. To fund that investment, the industry would need roughly $6 trillion a year in AI-related revenue.

"Consider the scale of investment and the gap between that and the revenue model necessary to fund it," Bain said in a recent report.

Why are employees uneasy?

A KPMG survey last year found that only a third of Germans trust AI-generated information. Höttges framed himself as "an AI optimist" on stage, a position that sits awkwardly with a workforce that has already watched 34,500 colleagues leave since 2020 and now confronts a CEO who has cloned his own image.

The Forrest Gump-style video illustrates the tension. Producing the clip likely saved production time, but Deutsche Telekom is not about to automate its chief executive, so the experiment did not replace anyone.

Workers lower in the organisation may prove harder to replace than the slideware assumes. AI tools still hallucinate, make errors, and rely on a handful of US-controlled foundation models. Swapping tens of thousands of domain experts for systems prone to mistake is a gamble the operator has not yet priced.

The next test arrives with Deutsche Telekom's full-year 2025 results, when management must show whether the Bonn plan translates into measurable cuts to indirect spend, and whether token budgets stay inside the envelope the operator has set for itself.