On November 3, 2026, Google announced an expansion of its research work in Singapore. The company linked a new local research hub to a package of more than $6 million in grants and research capacity building. Participants include the National University of Singapore (NUS) and Nanyang Technological University (NTU). The announced projects cover biomedical research, computing efficiency, regional languages and AI safety. The individual amounts identified in the announcement support specific lines of work within a broader programme.
Computing efficiency and regional languages
Google identified a $500,000 grant to NUS covering machine learning efficiency, model serving and regional language processing. Serving is the stage at which a trained model responds to actual requests. All else being equal, a system that uses less memory or handles more requests on the same hardware can reduce the cost of running it. To establish an improvement, researchers would need to measure resource use and output quality under comparable conditions. The announcement describes a research objective and provides no such measurements.
Regional languages are also a distinct area of work in the October 30 memorandum between NTU and Google. According to the university, cooperation with AI Singapore includes collecting, cataloguing and opening datasets that reflect Southeast Asian language use, along with evaluation suited to the region’s cultural context. A model can handle formal text in a language and still misunderstand a local expression or an exchange in which speakers switch between languages. Separate evaluations can reveal where more data or a different method is needed. The university also describes plans to explore open-weight models for education and healthcare.
A research environment and AI safety tests
Google said it is providing $340,000 for four studies at NTU. The topics include evaluation methods, safety frameworks, interactions among AI agents and language-model watermarking with privacy-preserving approaches. These subjects raise different questions. In a system with several agents, for example, it matters how information is passed between them and which component is allowed to take an action. A watermarking study might examine whether a mark survives ordinary text edits and how often its detector incorrectly flags human writing. The grant funds the studies. Published evaluations of their results are still to come.
The university also plans a secure research environment on Google Cloud using pre-approved configurations. NTU describes a scenario in which a team can launch an environment on demand, work on a vetted dataset and share selected information with an authorised outside collaborator. This could reduce the administrative work of setting up an environment for each research project. Its application will depend on access rules, how data is separated and what checks are built into the working system. The memorandum describes an environment that is still to be built. Google separately identified $750,000 for long-term research capacity in Southeast Asia through NTU.
What the biomedical funding supports
Among the amounts announced for NUS, Google highlighted $1.5 million under the Google.org AI for Science programme, with an emphasis on neurodegenerative diseases including Alzheimer’s. In its list of selected projects, Google.org identifies NUS’s work on genetic variation and RNA processing and links to Liu Boxiang’s laboratory. The university described the same research objective on October 29 without giving an individual award amount. The $1.5 million figure comes from Google’s November 3 announcement.
The research question in the NUS project is which genetic differences change how parts of an RNA molecule are removed and the remaining parts joined together. Computational models can help researchers prioritise possible effects for further investigation. The value of those predictions will depend on how they perform in independent checks and whether the predicted effects can be observed experimentally. The announcements establish the funding and the research objectives. Published methods, measurements and reproducible results will provide the next evidence.
