AI Efficiency Challenge ​

Supporting Translation-Ready
AI Efficiency Solutions

The rapid expansion of AI infrastructure presents an opportunity to strengthen long-term competitiveness

Significant AI efficiency gains are achievable with existing technologies. But a gap remains between these innovations and their large-scale deployment.

The AI Efficiency Challenge, led by STRIDE Ventures, operated by Start2 Group, and supported by the National Science Foundation, is closing that gap — accelerating commercial adoption of translation-ready AI/ML solutions to dramatically improve efficiency at scale.

The AI Industry Is Building Toward an Efficiency Breakthrough
With Unified Benchmarks Set to Drive Innovation and Define a Global Standard

High power demands and rising costs are hindering data center construction. And while researchers are developing powerful solutions to optimize AI/ML training, inference, and end-to-end performance, translation and adoption of these innovations have been slow.

The Solution

Two Types of Teams. One Mission

The AI Efficiency Challenge funds two types of applicants in a structured model, each with a distinct role in improving AI/ML efficiency at scale.

Solution Teams

Teams developing software-based technologies to improve efficiency throughout the layers of the AI/ML stack.

Pitchers

Technology Developers

Researchers, innovators, algorithm designers, and systems engineers with translation-ready efficiency technologies that are deployable at scale without longer development lead times.

Catchers

Problem Owners

Organizations operating large-scale AI and data center systems with the technical capacity to deploy efficiency-improving solutions directly into meaningful environments at scale.

Required: Solution Teams must include at least one Problem Owner (“Catcher”) capable of and willing to implement efficiency-improving solutions at scale. 

Benchmarking Teams

Teams developing industry-standard benchmarks to assess efficiency across the AI/ML software stack – and may also validate the efficiency gains achieved by Solutions Teams.

Umpires

Benchmarking Experts

Teams with deep expertise in benchmarking AI/ML software-based systems, who can define industry-facing efficiency measures and independently validate performance.

Priority Focus Areas

Software-driven Solutions with Measurable Impact

The Challenge prioritizes software-driven solutions capable of delivering measurable improvements in system performance and resource utilization. 

Software implementation of AI/ML systems

Tools to guide the creation of efficient code

MLOps and distributed system software

Edge computing

Energy-aware system management

Efficient agentic orchestration 

How the Challenge Works

Three Stages. Two Years. Up to $3.5 Million Per Project

The AI Efficiency Challenge runs over three milestone-driven stages spanning up to two years, with awards between $1.75 and $3.5 million per project. 

1

2 months: Identify inefficiencies, establish baseline performance metrics, and define the deployment roadmap for Stage 2 and Stage 3.

2

10 months: Execute iterative  deployment-development cycles in operational environments to validate efficiency improvements at scale.  

3

12 months: Continue spiral development-deployment cycles to further improve efficiency in larger, more impactful at-scale deployments.

Fast Track Available! Teams ready to move directly to implementation can choose an accelerated pathway, completing each stage in half the time.

TWO FUNDING LEVELS AVAILABLE

Two Award Levels. Built for Your Scope.

Teams can choose their desired funding level: Large ($3.5M) or Medium ($1.75M). For either level, a meaningful portion of the budget must be applied directly toward development and deployment-related activities.

Large: $3.5M

For teams with broader deployment scope and resource requirements. A meaningful portion of the budget must be directed toward deployment-related activities such as integration, instrumentation, and at-scale operation. 

Medium: $1.75M 

Teams at this level may supplement external funding with in-kind resources, such as staff time or cloud services contributed by commercial team members who prefer to shoulder their own costs. 

Bring Your Solution Forward​

Details

Solicitation Details

Read the solicitation details for more information on teaming structure, project scope, eligibility requirements, and application instructions.

Deadline

Application Closes July 13, 2026. 

Submit your application via the application form only.


Live Session

Join an Info Session

Our team hosts live info sessions where you can hear directly from the program teams, get a detailed walkthrough of the challenge, and ask questions during a live Q&A. 

FAQs

Still have questions? Browse the FAQs below. 

Is there a minimum team size?

No, the team size is flexible. What’s important is that all necessary skills and expertise are covered. The work plan must demonstrate that the team can carry out the tasks.

A team of experts from STRIDE Ventures will make a pre-selection. The final decision is based on the application and a pitch to an external jury consisting of scientists, industry experts, and/or investors. Teams will pitch between July 30 and 31, 2026. Final award decisions are subject to approval by NSF.

The application deadline is July 13, 2026, at 11:59PM Pacific Time. The Challenge starts in September 2026 and runs for 24 months, divided into three phases.

The intellectual property rights to the results created by the teams during the Challenge remain with the teams. Details are set out in the participation agreement, which will be published at or shortly after the application opening and can be accessed via the “Proposal Details” link above.

Though Solutions Teams and Benchmarking Teams apply through the same link, they are distinct applicant types with different questions to answer. At a high level, “Solution Teams” are developing and testing the efficiency technologies, while the “Benchmarking Teams” are creating standardized guidelines to measure efficiency gains in AI/ML systems.