How Universities Can Lead AI Research
Industry now produces more than 90 percent of the world's notable frontier AI models. Roughly 70 percent of AI-relevant PhD graduates head straight to companies rather than campuses, up from about 20 percent two decades ago. The four largest hyperscalers spent an estimated $410 billion on AI infrastructure in 2025 alone.
Against numbers like those, it is fair to ask whether research universities still have a meaningful role in shaping where AI goes. In a new report from Boston Consulting Group, *Building the AI-Forward Research University*, Urban Impact Advisors Senior Consultant Daniel Goetzel and his BCG co-authors Tejus Kothari, Bob Wu, and J.R. Sullivan argue that they do, provided they stop trying to out-scale industry and start leaning into what only a university can offer.
The report lays out a four-layer framework in which each layer reinforces the next. Distinctive, world-class research attracts talent. Talent requires compute. Compute makes a university a credible partner to industry. And those partnerships fund the next cycle of research. The authors walk through the emerging models at each layer, from independent and for-profit labs with university ties, to single-institution compute powerhouses like UT Austin's NSF-backed facility and state-convened consortia like New York's Empire AI, to corporate partnerships built around the curated datasets and regional implementation expertise that companies cannot easily replicate on their own.
On talent, the report is candid: universities cannot match industry salaries, and the gap has grown more than fivefold since 2001. But freedom to publish, multi-year research endowments, flexible leave, and mentorship are competitive advantages when they are treated as strategy rather than tradition.
The piece closes with seven concrete steps for university leaders, including clarifying an institutional AI position, establishing IP frameworks before partnerships begin, diversifying compute funding, and engaging state partners proactively on shared infrastructure.
For those of us who work at the intersection of universities, regions, and innovation districts, this report is a useful map. The institutions that anchor the next generation of innovation ecosystems will be the ones that decide, deliberately, where they intend to lead.
Read the full report at BCG →
https://www.bcg.com/publications/2026/how-universities-lead-ai-research
