Tadej Pogačar’s Tour de France team is leaning hard into artificial intelligence ahead of the 2026 race, using algorithms to turn a flood of rider data into sharper training calls and cleaner race-day decisions.
The pitch isn’t sci-fi. It’s practical: reduce uncertainty, spot trouble earlier, and avoid the small misreads that can snowball into lost time in the mountains. In modern cycling, raw power still matters, but so does how fast a team can interpret what the numbers are really saying.
Inside UAE Team Emirates-XRG’s data machine around Pogačar
Sommaire
Pogačar rides for UAE Team Emirates-XRG, one of the sport’s best-funded, most tech-forward operations. The team already collects massive streams of information from on-bike sensors and structured testing, power output, heart rate, cadence, speed, and how those metrics change over time.
They layer in context that can swing performance: temperature, altitude, route profile, travel load, and sleep quality (reported by riders or measured by wearables). The AI comes in when the volume becomes too big, and too noisy, for even experienced coaches to process quickly and consistently.
The goal, team insiders stress, isn’t to replace the coach. It’s to rank what matters. Models can flag mismatches between planned workload and what a rider actually absorbed, or detect subtle trends like slower recovery, unusual heart-rate drift at familiar intensities, or a gradual drop in responsiveness that might not jump out day to day.
That “industrialized” analysis also extends beyond the star. Pogačar’s numbers are compared against internal benchmarks, teammates with similar profiles, past altitude camps, and multi-season training blocks, to find patterns that help staff spend less time sorting data and more time making decisions.
With that comes a new pressure point: data governance. Who gets access, when, and under what privacy rules? In a sport where physiological data can reveal form, and vulnerability, teams have to balance analytical speed with tight confidentiality, from account security to controlling exports from connected devices.
And AI is only as good as the inputs. A miscalibrated sensor, inconsistent testing protocols, or sloppy self-reported sleep can send a model in the wrong direction. That’s why teams typically standardize measurements and treat algorithmic recommendations as prompts, then validate them with human expertise.
Predictive models to fine-tune workload, recovery, and injury risk
The real promise is dynamic personalization. Instead of sticking to a rigid training plan, staff can adjust workload based on how a rider is responding in real time, estimating residual fatigue, predicting whether the athlete can handle a hard session the next day, or swapping in a shorter, higher-quality workout when recovery markers slide.
Teams already use established metrics like Training Stress Score (TSS), Chronic Training Load (CTL), Acute Training Load (ATL), and heart-rate variability (HRV). AI aims to connect those numbers more tightly to real-world conditions. The same training load can hit differently depending on travel, altitude, heat, time of season, and a rider’s recent history, variables models try to integrate into probabilities: How likely is this session to land well? What’s the risk of a flat performance if we push through?
Injury and illness prevention is another selling point, carefully framed. AI can aggregate weak signals like declining power at the same heart rate, fragmented sleep, persistent fatigue sensations, or unusual variability in effort and generate alerts. But it doesn’t diagnose; it ranks risk. Final calls still sit with medical staff and coaches, who can verify with additional tests or simply back off before a small issue becomes a race-ending one.
These tools can also change the athlete-staff dynamic. Pogačar is known for racing on instinct and feel, but AI adds a second opinion. When the rider says he feels great and the data says otherwise, the conversation gets more structured: re-check the sensor, repeat a test, or choose caution even when confidence is high. When both align, the team can greenlight ambitious blocks with more certainty.
The biggest limitation is interpretation. A model trained on past seasons can stumble when equipment changes, nutrition protocols shift, or a crash alters biomechanics. Cycling is messy, weather, stress, tactics, mechanicals. AI can narrow uncertainty, not erase it, which is why the best teams treat it as an iterative tool: test, compare, correct, repeat.
Simulating stages and race scenarios for Tour de France 2026
AI isn’t just for training. Teams are also using it to prepare for the race itself, building richer simulations from GPS files, stage profiles, and reconnaissance rides to anticipate where time can be won or lost.
Models can blend historical data, speeds on certain gradients, wind effects, peloton dynamics, and estimate the likelihood of key events like splits in crosswinds. That can translate into concrete plans: where to position teammates, when to move up, which sectors to avoid, and when it’s smarter to concede a few seconds rather than burn the whole squad.
Pacing is another target, especially on long climbs and time trials. A rider might aim for a target power number, but real roads demand constant micro-adjustments, surges, steep ramps, brief recovery pockets. AI can propose effort-distribution strategies tailored to a course profile, then get checked against weather and the rider’s feel on the day.
For Pogačar, famous for sharp, decisive attacks, the value is timing: identifying moments when an acceleration has the best payoff relative to the risk of blowing up later.
Teams also run “what if” responses to rivals’ moves. If a competitor attacks a certain distance from the summit, what’s the likely cost to follow, and what does that do to tomorrow’s legs? Those calculations can shape whether a team saves a domestique for the valley chase or turns the screws early to isolate a contender.
the Tour laughs at perfect plans. Crashes, flats, sudden weather shifts, and unexpected splits can make a strategy obsolete in minutes. The advantage of AI, teams argue, is faster recalculation, estimating the cost of a chase, the benefit of waiting, and the best option when chaos hits.
A competitive edge, with privacy, fairness, and rule questions close behind
AI could widen the gap between cycling’s haves and have-nots. Wealthier teams can hire data engineers, buy better sensors, and build custom platforms, advantages that already exist in the sport, but could compound if AI becomes a force multiplier for planning, prevention, and tactics.
Data security is now part of competitive defense. Physiological trends, nutrition protocols, and indicators of form are valuable intelligence. A leak, accidental or deliberate, could tip off rivals about a leader’s condition heading into a mountain week. That’s pushing teams toward tighter digital hygiene: limiting exports, encrypting data, segmenting access, and training staff to avoid phishing, especially with constant travel increasing exposure.
Public transparency will likely remain thin. Teams love talking about “innovation,” but rarely disclose models or methods because the details are competitive. That secrecy can fuel myths that AI “creates” performance. In reality, most gains are incremental: better recovery, smarter fueling, cleaner pacing, fewer bad days.
Regulators haven’t banned AI outright, but its use intersects with areas that are already sensitive, real-time communications, technological assistance, medical data collection, and privacy. As tools get more powerful, cycling’s governing bodies and race organizers may face pressure to clarify what’s allowed during competition versus what belongs in pre-race preparation.
For Pogačar’s team, the balancing act is clear: use AI to sharpen decisions without losing what still wins bike races, instinct, tactical reading of the peloton, and the ability to improvise under pressure. In a Tour where the yellow jersey can hinge on seconds, the edge often comes from stacking dozens of small, correct calls, made fast, with reliable information, by people who know what they’re looking at.
