Recent Research: Exploring nationwide distribution of AI-focused Phase II SBIR projects

States with top-ranking university AI research programs garner a greater number of Phase II AI-related SBIR awards, according to a working paper from the Department of Economics at the University of North Carolina at Greensboro. Researchers there investigated state variations in the distribution of Phase II SBIR research projects focused on artificial intelligence (AI). The authors of the paper hypothesized that the state-by-state variations are related to the presence of a research university with a “Top 10” AI program in each state. Analysis showed that three out of the five states receiving the most funding for AI-related Phase II SBIR projects had a top-ranked AI research university. Although proximity to a top research university may be beneficial to Phase II SBIR applicants with AI-focused projects, it is not the only path to success in capturing SBIR funds.

The study of artificial intelligence has grown exponentially in the past four decades. While the study of AI first began in 1955, according to the paper, funding for AI research was minimal until the 1980s, when researchers began expanding machine learning algorithms. Despite the growth of AI-related research in previous decades, these research projects are primarily concentrated around research universities with a high-ranking AI program. Federal funding for AI-related projects follows a similar pattern of being located close to research universities, the researchers found.

The Small Business Innovation Research (SBIR) program is a competitive awards-based program that invests federal research funds into small businesses to address critical American priorities. It requires participating federal agencies with extramural research budgets exceeding $100 million to allocate 3.2% of that budget to support research and development in small businesses. The program features three phases: Phase I to determine technical feasibility, Phase II to develop the technology further, and Phase III to support commercialization. In this paper, the researchers concentrated on Phase II projects – two-year projects focused on creating new technologies with funding up to $1 million. After filtering through Phase II projects, the researchers found that about 9.4% of the SBIR Phase II projects could be considered “AI-focused.”

The researchers assumed that access to a highly-ranked AI research university program impacted the decision to award AI-related Phase II funding to small firms in FY 2020. In the paper, the researchers used the U.S. News and World Report ranking of AI research programs in the U.S. The top universities included: Carnegie Mellon University (PA), Massachusetts Institute of Technology (MA), Stanford University (CA), University of California-Berkley (CA), University of Washington (WA), Cornell University (NY), Georgia Institute of Technology (GA), University of Illinois-Urbana-Champaign (IL), University of Texas-Austin (TX) and University of Michigan-Ann Arbor (MI). The researchers then compared the distribution of AI-related Phase II projects in states with or without a top-ranking AI research program.

This analysis revealed that, on average, states with a top-ranking AI research program had more AI-focused Phase II projects. According to figures from the report, the average percent of total AI-focused Phase II projects in the United States was 1.1% in states without a Top 10 AI research university and 6.3% in states with a Top 10 AI research university. However, there were states that received a large percentage of Phase II SBIR awards that did not have a Top 10 AI research university. According to the data, the top five states receiving the majority of AI-related Phase II SBIR awards are California (46 awards), Massachusetts (18 awards), Maryland (12 awards), New York (9 awards), and Ohio (9 awards). Of these states, only California, Massachusetts, and New York have what the paper defined as a Top 10 AI research university program. 

The researchers found that the availability of research resources to support AI-related projects is associated with public support of AI research in small firms. To encourage more diversity in the spatial distribution of AI-related research in smaller firms, the researchers recommend increasing the availability of AI-related research resources across the nation. The authors noted that the National Science Foundation has acknowledged this need with the establishment of seven National AI Research Institutes in FY 2020 and 11 in FY 2021. Additionally, the paper noted that the National Artificial Intelligence Research Resource Task Force is developing a roadmap to further develop the symbiotic relationship between AI-related firm resources and AI-focused research universities. Seven of the 21 NSF AI Research Institute awards were made directly to states with Top 10 AI Research Universities to encourage firm-university partnerships. The locations of these AI Research Institutes are as follows:

• AZ: Arizona State University (planning grant)
• CA: University of California – Davis
• CA: University of California – San Diego
• CO: University of Colorado at Boulder
• GA: Georgia Research Alliance (2)
• GA: Georgia Tech Research Corp
• IL: University of Illinois at Urbana-Champaign
• IN: Indiana State University (planning grant)
• MA: Massachusetts Institute of Technology
• MO: Washington University (planning grant)
• NC: Duke University
• NC: North Carolina State University
• NM: Santa Fe University
• OH: Ohio State University (2)
• OK: University of Oklahoma, Norman Campus
• PA: Pennsylvania State University (planning grant)
• PA: University of Pennsylvania (planning grant)
• TX: University of Texas at Austin
• VT: University of Vermont (planning grant)
• WA: University of Washington

This research provides an examination of the relationship between research programs at higher education institutions and AI-related Phase II SBIR awards. However, it does not confirm that these institutions' presence directly impacts the SBIR awards received. Several other factors could contribute to the distribution of AI-related and non-AI-related SBIR awards nationwide. Access to local organizations providing SBIR support, the strength of a proposal, and the evaluation metrics used by each participating agency to select its award recipients could also impact how likely a small business is to be selected for a Phase II award.

The working paper, The Spatial Distribution of Public Support for AI Research, by Farhat Chowdhury, Albert N. Link, and Martijn van Hasselt, is available here.