TomTom Cuts 200 Roles as It Reorganizes Around Automated Driving and Location Intelligence
TomTom is splitting its business into three divisions and targeting €21 million in annual savings. The reorganization makes its growth thesis unusually clear: navigation pays the bills today, while automated driving and AI-ready location intelligence are expected to deliver tomorrow’s expansion.
TomTom is reorganizing around three divisions, Automated Driving, Location Intelligence and Navigation, in a move that will affect approximately 200 jobs worldwide.
The Amsterdam-based mapping company says the new structure will reduce management layers and bring decisions closer to product execution. It expects the changes to produce about €21 million in annualized savings. Yet the announcement is more than a cost-cutting exercise. It sets out a sharper answer to a strategic question TomTom has faced for years: where does an established mapmaker find growth after consumer satnav hardware?
The answer is now explicit. Navigation remains the commercial backbone, but TomTom is placing its future growth bets on automated driving and enterprise location intelligence, both increasingly shaped by artificial intelligence.
A strategic reset under revenue pressure
The reorganization arrived alongside TomTom’s third-quarter 2026 results. Group revenue fell 6% year over year to €129 million. Location Technology revenue declined 3% to €115 million, while automotive operational revenue dropped 19% to €68 million.
TomTom reported an operating result of €6 million, equivalent to a 5% margin, and net cash of €239 million. The company raised its expected full-year operating margin from 3% to approximately 4%, even after including a €15 million restructuring charge expected in the fourth quarter. Excluding that one-time charge, it expects a margin of about 7%.
The market focused on the weak top line. Reuters reported that quarterly revenue missed the company-provided analyst consensus of €131 million and TomTom’s shares fell about 5% after the results. ING analyst Marc Hesselink pointed to continued weakness in automotive and an enterprise revenue miss.
TomTom also narrowed its 2026 revenue forecast to between €505 million and €530 million, from a previous range of €495 million to €555 million. The midpoint is lower, even as the profitability outlook improves.
That combination matters. TomTom is becoming leaner faster than it is becoming larger.
Three divisions, one shared map
Under the new structure, Navigation remains the unit closest to TomTom’s established business. Chief executive Mike Schoofs described it as the company’s backbone and the source of most current revenue.
Automated Driving will concentrate the technology used to support higher levels of vehicle automation. This includes high-definition and continuously updated geographic representations that complement vehicle sensors with information about road geometry, lanes, rules and conditions beyond immediate sensor range.
Location Intelligence will serve enterprises and developers that need map data, routing, traffic analysis and place intelligence inside applications or analytical workflows. This is where TomTom expects AI to shorten the path between raw geographic data and a usable answer.
All three units rely on the same underlying assets: TomTom’s maps, traffic data, location technology and AI capabilities. The reorganization therefore does not represent a breakup of the map stack. It is an attempt to package that shared stack around customer groups with different buying cycles and product needs.
The location layer for AI agents
TomTom has started presenting its maps not only as data for visual interfaces, but as context that software agents can query.
During the quarter, the company expanded its Microsoft collaboration to make location intelligence available to AI agents built with Microsoft Fabric and Microsoft Foundry. It also launched a Maps MCP Server in Anthropic’s Claude connector directory. MCP, or Model Context Protocol, allows an AI system to call external tools and retrieve structured information instead of relying only on knowledge embedded in a language model.
For location technology, this matters because many seemingly simple questions require live and spatially precise data. A model asked to identify reachable facilities, compare routes or analyze activity around a site needs geocoding, routing, travel-time calculations and current place information. A general language model cannot reliably infer those details from text alone.
TomTom’s enterprise opportunity is to make those capabilities easy for agents to access while preserving the accuracy, coverage and update discipline expected from a production mapping service.
Schoofs told Reuters that revenue from AI-related location intelligence should emerge faster in enterprise markets than in automotive. That assessment reflects the different deployment cycles. A developer can add an API or agent connector relatively quickly. A mapping component destined for a production vehicle can face years of engineering, validation and model-cycle planning.
Automated driving is a longer bet
TomTom remains optimistic that automated driving is moving from experimentation toward commercial programs. Its recent initiatives include collaborations with Microsoft, Mitsubishi Electric and DeepRoute.ai.
The case for maps in automated driving is also becoming more nuanced. Some autonomy companies have promoted map-light or mapless systems, but vehicles still operate inside a geographic and regulatory context. Road layouts, speed rules, lane restrictions and local driving conventions change by market. A continuously updated map can provide context and redundancy even when AI perception handles the immediate scene.
However, the 19% decline in automotive operational revenue shows why this cannot be treated as near-term proof of growth. Automotive contracts are large but slow, and a partnership announcement does not guarantee production volume.
What to watch next
TomTom says it expects Location Technology revenue to return to growth in 2027. That is now the key test of the reorganization.
Investors and customers should watch for three kinds of evidence. First, TomTom needs to identify production deployments that turn automated-driving partnerships into recurring revenue. Second, it needs to show that agent-facing location tools create meaningful enterprise usage rather than experimental integrations. Third, it must demonstrate that reducing approximately 6% of its workforce does not weaken map quality, customer support or the pace of product development.
TomTom’s new organization tells a coherent story: navigation funds the present, automated driving offers a long-cycle opportunity, and location intelligence for AI could grow sooner. The third-quarter figures also show the burden of proof. Efficiency has improved, but the company still needs revenue growth to validate the strategy.
#Business
Next article
Earth-observation systems produce vast historical archives, but the models creating those archives often behave as if every year were the first.
A crop-mapping model might process the latest Sentinel imagery, infer what is growing in each field and then hand the result to a separate consistency layer. That workflow can work, but it keeps temporal reasoning outside the model. A new study asks a deceptively simple question: what if the model could directly use its own earlier maps as another input?
The research, focused on the Copernicus High Resolution Layer Croplands production setting, offers a practical version of “memory” for GeoAI. The authors do not propose a general-purpose agent or a larger foundation model. Instead, they feed thirteen years of aligned historical class predictions and a Basic Vegetation Layer signal into the existing multi-source architecture.
The missing temporal context
Most land-cover systems already acknowledge that geography changes over time. The problem is where that logic lives. In many operational pipelines, the classifier sees only the current year while temporal rules are applied later. This separation makes the system harder to optimize end to end and prevents the current prediction from learning directly from previous output patterns.
The paper reframes those prior maps as a “free input modality”: free not because they cost nothing to produce, but because an operational programme has already created them. Instead of treating yesterday’s predictions as an inert archive, the model treats them as structured context.
Turning old predictions into a new input
The added branch is deliberately lightweight. It consumes two aligned time series for each location: the product’s own class history and a vegetation mask. The model then combines those features with the baseline inputs used in production.
This matters because historical predictions encode more than class labels. They also capture the behaviour of the whole earlier production process: sensor availability, model decisions, local crop rotations and the way particular landscapes are represented. Used carefully, that record can help the new model distinguish genuine change from a one-year anomaly.
The approach is also operationally attractive. It does not require building a new continental reference database or acquiring a novel sensor. It reuses an asset the programme already owns.
Modest overall gains, larger gains where history matters
The headline improvement is not a dramatic leap in aggregate accuracy. That is important context. The value appears in more consistent predictions, class-level gains and fewer implausible year-to-year changes.
For a continental product, those changes can be strategically significant. Users often compare annual maps, measure transitions and detect trends. A small improvement in a single-year score may understate the benefit if the resulting time series contains fewer false flips.
Why consistent masks matter
Temporal stability is not simply aesthetic. A noisy transition can be mistaken for crop rotation, land abandonment or a change in management. Those errors propagate into statistics, policy analysis and downstream machine-learning systems.
By making history available during inference, the model gains a prior about what is plausible at a location. The vegetation layer provides an additional constraint, helping it avoid classifications that conflict with the land’s broad physical state.
Not every problem needs a larger foundation model
The study lands at a moment when the geospatial AI conversation is dominated by foundation models and agentic workflows. Its lesson is more prosaic and possibly more useful: some operational improvements come from better system design rather than more parameters.
A mature mapping programme has accumulated valuable internal data. Turning that archive into a model input can be cheaper, more explainable and easier to deploy than replacing the entire stack.
The danger of inherited errors
History can also mislead. Previous maps contain their own systematic errors, and a model trained to trust them may repeat or reinforce those mistakes. The authors’ result should therefore be read as evidence for controlled temporal context, not as permission to treat past predictions as ground truth.
Future operational work will need to monitor error persistence, confidence calibration and performance in places where land use changes abruptly. New construction, extreme weather and changes in agricultural practice are exactly the cases where yesterday’s map may be least reliable.
GeoAI with memory
The broader implication extends beyond crop mapping. Any regularly updated geospatial product—buildings, roads, forests, impervious surfaces or settlement layers—could potentially use its own output history as context.
The shift is subtle: a mapping model stops being a sequence of isolated annual classifiers and starts behaving like a system with continuity. That is not full agentic reasoning, but it is a meaningful step toward GeoAI that understands not only where something is, but how its mapped state has evolved.
Research source: “Multi-Year Geospatial Reasoning using Interannually-Consistent Historical Predictions as a Free Input Modality”. The paper is a preprint and its reported findings have not yet been independently replicated.
How do you like this article? Read more and subscribe to our monthly newsletter!
