AWS for M&E Blog
How Prime Video delivers AI-powered real-time insights for NASCAR fans
In the high-stakes world of NASCAR, fuel management can make the difference between victory and defeat. Recent races have demonstrated how critical fuel calculations are, with multiple top drivers experiencing last-minute fuel shortages, dramatically altering race outcomes.
NASCAR broadcasts have historically provided limited visibility into fuel consumption—one of racing’s most crucial strategic elements. While announcers may discuss fuel strategy and its impact on race outcomes, viewers had no way to visualize or understand real-time fuel usage across different drivers. This impacted the ability for NASCAR production teams to examine important storylines as the race entered its final stages.
In May 2025, Amazon Prime Video introduced its newest AI-powered Prime Insight, Burn Bar, to their NASCAR on Prime coverage. The Burn Bar is a meter that predicts if a driver is driving hard or driving conservatively—typically achieved by drafting off other cars or getting a push from another car. The proprietary model continuously evaluates each driver’s fuel consumption and efficiency throughout the race. It provides meaningful insights as pivotal stages unfold.
Innovations like the Burn Bar enable unique storytelling opportunities, enhancing the viewing experience for fans by adding excitement and complexity. Fans engage with the technical aspects of fuel management, following the race-within-the-race as race teams weigh their options in real time on different approaches.
The Burn Bar provides unprecedented visibility into drivers’ fuel burn rates during crucial race moments. It helps viewers understand:
- How aggressively each driver is racing
- The fuel efficiency trade-offs of different racing strategies
- The impact of track position and driving style on fuel consumption
- A visualization of real-time strategic decisions between speed and fuel conservation
The fuel management challenge
Unlike everyday vehicles, NASCAR stock cars don’t have fuel gauges. Without gauges, NASCAR teams rely on precise manual measurements and calculations performed by crew chiefs to track fuel consumption.
However, this process remains manual and tedious, leaving teams and audiences blind to a competitors’ fuel consumption pattern. Additionally, this process becomes increasingly complex during races due to modern pit stop dynamics. For example, teams often choose not to wait for the need of a complete refill, creating a compounding uncertainty about exact fuel levels as the race progresses. However, a miscalculation can mean the difference between the victory lane or running dry on the final lap.
There is no standardized ground truth data for measuring fuel consumption in these elite racing machines. Previously, critical fuel data like MPG, burn rates, and accurate fuel levels, have not been shown on NASCAR broadcasts. The revolutionary fuel insights solution changes the motorsports viewing experience by providing this information, giving fans visibility into a racing strategy that was previously available only to teams.
This variability of data presents a fascinating challenge for anyone attempting to build predictive models for fuel consumption. Without a single source of truth to calibrate against, developing an accurate fuel prediction system requires multiple data sources and accounting for the inherent uncertainty in each methodology.
The Burn Bar solution
The Burn Bar uses raw data that feeds into a sophisticated machine learning model. The model calculates each vehicle’s unique fuel consumption patterns, accounting for the various factors that affect fuel usage. These calculations are made available through both a custom dashboard and REST API, providing an analytical resource for the NASCAR on Prime announcers (Steve Letarte and Dale Earnhardt Jr.).
The Prime Video production team use the same dashboard data to create custom graphic overlays that enhance the viewer’s understanding of strategic race decisions. This approach overcomes the traditional challenge of measuring fuel consumption without physical gauges.
Figure 1: The Burn Bar as presented during a broadcast.
Technical overview
The following diagram illustrates the solution architecture and the key stages.
Figure 2: High-level architecture.
The solution processes multiple complex data streams to calculate accurate fuel metrics across all manufacturers (Chevrolet, Toyota, and Ford), enabling dynamic visualization through broadcast graphics and different digital platforms. To enable this sophisticated real-time analytics capability, the solution architecture is built on several key components.
- Data ingestion layer
The solution’s core functionality centers on its data ingestion architecture. Live race data is captured through containerized NAT clients that run on Amazon ECS with AWS Fargate—a serverless, pay-per-use container compute service. This serverless approach provides automatic scaling capabilities that verify optimal performance and maintain high availability throughout race events. These specialized clients operate continuously during races, capturing and processing various streams of essential racing telemetry and information.
The architecture processes multiple data feeds encompassing vehicle telemetry, accurate positional data, detailed timing metrics, and essential pit stop information. This comprehensive multi-feed strategy confirms complete capture of all data points necessary for accurate fuel consumption analysis. The Amazon ECS scheduler automatically maintains optimal task counts and, when paired with container health monitoring, seamlessly replaces any failing tasks. This approach provides uninterrupted data collection throughout the entire race while minimizing operational complexity.
Amazon Kinesis Data Streams is a scalable and durable real-time data streaming service, collecting feeds from NATS clients and distributing them to subsequent processing and analysis workflows. The service’s robust throughput capabilities provide dependable buffering for the substantial data volumes produced throughout racing events.
- Processing layer
The primary analytics engine leverages Amazon Managed Service for Apache Flink to execute advanced near real-time data processing. This intelligent processing tier manages several essential operations concurrently, including real-time consolidation of diverse data feeds while facilitating accurate time-based synchronization.
Data processing occurs at three-second intervals, balancing speed with analytical accuracy. The architecture incorporates the broadcast state pattern that enables streamlined parallel computation of fuel analytics. This design is particularly valuable for accommodating the variable data velocities from different sources while sustaining optimal processing performance. Processed results from Apache Flink flow into a dedicated Kinesis data stream, where AWS Lambda functions handle the formatting and preparation of analytics data for downstream consumption.
- Storage Layer
Real-time race analytics data is stored persistently in Amazon DynamoDB and made available through Amazon API Gateway. These analytics, including fuel consumption metrics for all drivers on the track, are visualized in custom dashboards. Static elements of these real-time dashboards are hosted on Amazon Simple Storage Service (Amazon S3), providing efficient delivery of the user interface.
- Data access layer
The solution delivers data through two main channels. First, a live dashboard provides broadcast teams the ability to instantly visualize competitive fuel strategy dynamics. Second, REST APIs hosted on Amazon API Gateway give production teams a way to develop custom visual elements that help viewers better comprehend racing tactics and strategy decisions.
- Security layer
The solution uses Amazon Cognito to deliver robust authentication and authorization controls, restricting dashboard access exclusively to verified broadcast personnel with appropriate credentials. Security is further enhanced through AWS WAF, which protects the dashboard application by filtering web traffic and defending against common web exploits that could affect availability and security.
The architecture delivers exceptional performance and reliability—critical requirements for live race broadcasting environments. By leveraging the elastic infrastructure of AWS, the system efficiently processes high-volume telemetry data streams while maintaining consistently low latency. This robust design enables the solution to deliver precise fuel consumption analytics that meet the high production quality standards of professional broadcasting, delivering reliable data to both broadcasters and viewers.
Outcomes
The Burn Bar is a mainstay of NASCAR on Prime coverage. It was featured 14 times throughout the inaugural five race campaigns in 2025. During the Coca-Cola 600, Steve Letarte utilized the Burn Bar (at 4:42:13) to tell the story of Byron pushing hard to overtake Hamlin, the race leader, near the end of Stage 3. Byron ultimately passed Hamlin to win the stage.
Conclusion
The Burn Bar represents a groundbreaking solution to the longstanding challenge of making fuel strategy visible to viewers. By transforming previously hidden fuel management data into compelling visual insights, Prime Video has enhanced NASCAR broadcasts while giving fans unprecedented visibility into race strategy.
Through a sophisticated AWS architecture leveraging services like Amazon ECS, Amazon Managed Service for Apache Flink, and Amazon Kinesis Data Streams, the solution processes complex telemetry data in near real-time with broadcast-quality reliability. As demonstrated throughout the 2025 season, the Burn Bar has become an integral part of NASCAR on Prime coverage, transforming how fans experience racing while showcasing the power of AWS to revolutionize sports broadcasting.
Contact an AWS Representative to know how we can help accelerate your business.
Acknowledgement
This was a joint effort from various teams across AWS, Amazon Prime Video and NASCAR.
Matt Stark, Tulip Gupta, Rachel Kelley, Rick Goodman, Mohit Mysore, Kevin Graziadei, Anshuman Das, Scott McAuliffe, Peter Talbot, Taylor York, Chandan Dash, Kamal Jagga, Dilshad Raihan Akkam Veettil, Yingmao Li, Lalitha Sindhura Raviprolu, Ram Yennapusa, Dashiell Flynn, Steven Miller, Aravind Kodandaramaiah, Mona Fathollahi, Kevin Oleniczak, Sadhana Tare, Mohan Kumar Guruswamy, Hongyou Lin, Sam Schwartzstein, Alex Strand, Dan Pawlowski, and Anthony Da Cruz.
Further reading
- Accelerating Motorsports: How NASCAR delivers real-time racing data to broadcasters, racing teams, and fans
- How Formula 1® uses generative AI to accelerate race-day issue resolution
- Formula 1® unlocks the most competitive season yet with AWS
- F1® sets new standard in sports streaming with F1 TV Premium using AWS