Singapore Launches National AI Research Fellowship

On 1 April 2025, the National Research Foundation (NRF) opened the first call for Singapore’s National AI Research Fellowship, a five-year award worth up to S$3 million per fellow. The scheme is aimed at early-career academics whose work sits along the artificial intelligence pipeline, from core machine learning to computer vision, natural language processing, robotics and AI safety. NRF has set aside S$120 million to support about 20 fellows in the first tranche. Host institutions include the National University of Singapore (NUS), Nanyang Technological University (NTU), Singapore Management University (SMU), Singapore University of Technology and Design (SUTD), Singapore Institute of Technology (SIT) and A*STAR research institutes.

For an assistant professor who has spent the past year trying to schedule GPU time, a five-year funding line changes the arithmetic. It covers salary support, postdoctoral researchers, PhD students, compute and lab space before the first grant cycle starts, which is exactly the stretch where early-career hires stall.

What the fellowship pays for

The S$3 million ceiling is a package, not a salary. It can be spent across the full cost of a nascent lab.

• Salary support for the fellow, calibrated to the host institution’s pay scale
• Postdoctoral researchers and PhD students recruited to the fellow’s group
• Equipment, including GPU clusters and cloud compute credits
• International travel and collaboration costs
• Technology transfer and licensing expenses for projects that move toward application

Singapore’s older NRF Fellowship for early-career researchers also offers up to S$3 million over five years. The new scheme narrows the remit to artificial intelligence and asks each host university or institute to commit to a tenure-track or equivalent research position. That requirement changes the risk calculation for applicants. It turns a fellowship into something closer to a start-up package with an academic home attached.

Who can apply

The base requirement is a PhD awarded within the past ten years, with a publication record in AI conferences or journals. Singapore citizens and permanent residents, as well as international candidates, may apply. The call asks for a research statement, three letters of reference, a host institution endorsement and a one-page compute plan.

NRF expects fellows to spend at least half their time in Singapore, and the award remains tied to the host institution. That condition will matter to applicants juggling offers in two or three time zones.

Why Singapore is paying for AI talent now

Singapore’s National AI Strategy 2.0, adopted in December 2023, commits more than S$500 million to AI compute, talent and industry development. The strategy sets a target of 15,000 AI practitioners, roughly triple the existing pool. The fellowship is the academic lever in that plan, designed to seed research groups rather than make one-off senior hires.

The base rate is sobering. Public universities in the United States and Europe are delaying AI faculty searches, while technology companies keep bidding up salaries for expert PhDs. Singapore has not matched those industry salary packages across the board. What the fellowship packages differently is pre-negotiated compute, dedicated student funding, formal access to local industry partners and a clear publication pipeline.

The 15,000-practitioner target appears in the National AI Strategy 2.0. NRF lists the host institution requirements on its grants and fellowships page.

What this means for your lab

A computer scientist I’ll call Dr. L spent February 2025 asking three local data centres for GPU quotes. By the time two responded, the third had tripled its per-hour rate. Dr. L is three years out from the PhD, with a small group, a growing publication list and no institutional purchasing power. The fellowship is aimed at exactly that profile.

Early-career AI labs usually rely on shared clusters, where large training runs queue behind undergraduate courses and established groups. A five-year award changes the baseline if, and only if, the host department writes down the compute commitment before the fellow signs. Otherwise the first year disappears into procurement forms.

Before accepting an offer, ask for five things:

• Written confirmation of GPU allocation and cloud credits for the first two years
• A PhD student slot in year one, not year three
• Clarity on whether indirect costs are covered or waived
• Data-sharing terms for any hospital, defence or industry partner
• Named laboratory or office space that survives the next renovation cycle

Application mechanics

The first call closes at the end of June 2025. Candidates submit through NRF’s online grant system, and the host institution must endorse the application. A review panel that includes international AI researchers will shortlist applicants for interview. NRF has said awards will be announced in batches rather than on one day.

University administrators should read the scheme as a matching commitment, not a free-standing line. Each award requires an institutional appointment and lab space. Faculty deans need to decide early whether a fellowship application will pair with a tenure-track search or a fixed-term research post, because that choice affects visa classification, teaching load, promotion clock and access to faculty housing.

For a broader look at compute bottlenecks in research universities, see our earlier report on AI compute access and research infrastructure. The NRF call document is published alongside the A*STAR research community.

The next step

For a final-year PhD student or postdoc, the first action is not the application form. It is securing a host. In Singapore’s system, an applicant cannot submit without a department chair or research institute director willing to endorse the proposal and commit space. Send a two-page research statement to one named faculty member at NUS or NTU, or to a group leader at A*STAR, and ask whether the department will nominate. If the answer is vague, move to the next institution.

For research administrators, the next step is to map current AI hires against the department’s compute allocation. A S$3 million fellowship will stall quietly if its first year disappears into queue time and procurement forms.

Photo by Swapnil Bapat on Unsplash