Singapore, the national university gives ChatGPT Edu to all students and faculty
The agreement between the National University of Singapore and OpenAI has been in effect since August 31 and also includes Codex; a survey of more than 500 university students in the city-state indicates that 94% use artificial intelligence several times a week.
As of August 31, 2026, ChatGPT Edu and Codex are available to the entire National University of Singapore community: enrolled students and faculty alike, with no distinction by department. The availability stems from an agreement with OpenAI, made public through a joint announcement and reported by AI News.
A university’s choice to adopt a tool for its entire student population, rather than for individual courses or research groups, shifts the question from whether to how: once access is universal, usage is no longer a variable that the institution can decide course by course, but a starting condition that teaching and assessment must reckon with.
That this starting condition already exists independently of the agreements is suggested by a survey conducted among more than 500 Singaporean university students: 94 percent report using artificial intelligence several times a week. This is a self-reported measure referring to a limited sample from a single city-state: it indicates how widespread the habit is among respondents, not how it affects study outcomes or the quality of learning, aspects the survey does not measure.
The other front: the chip
In the same week, OpenAI moved on a different front, that of hardware. At the Hot Chips conference on August 25, 2026, the company presented the first test results of its inference chip Jalapeño, developed with Broadcom. According to data shown by OpenAI on the public InferenceX benchmark by SemiAnalysis, the chip records 1.5 to 1.9 times more work per watt compared to an Nvidia Blackwell system, reports AI News.
The data should be read for what it is: these are measurements presented by the company that produces the chip, based on a public third-party benchmark, with no independent verification of the results reported in the available source. The commercial scope is equally relevant: deployment is planned for late 2026 in limited volumes, with expansion in 2027, and the chip will not be sold to external customers. It therefore does not enter the market in direct competition with existing suppliers: it serves to reduce the energy cost of the infrastructure of the company that produces it.
| Element | Data |
|---|---|
| University access active since | August 31, 2026 |
| Tools included in the agreement | ChatGPT Edu, Codex |
| Weekly usage reported by students | 94% |
| Reported efficiency of Jalapeño | 1.5-1.9 times per watt |
| Chip deployment | late 2026, limited volumes |
What we don’t know
The announcement of the agreement with the university does not clarify the duration, the costs borne by the university, the handling of student data, or the rules of use adopted for exam assessment. Regarding the survey, no information is available on the sample selection method or the survey date. Regarding the chip’s results, independent verification is lacking, as is detail on the configurations being compared.
The verifiable fact that remains is the scale: as of August 31, 2026, the entire student and faculty population of an Asian national university has access to the same tool, provided by the same company.
← Archive · Front page · Past editorials · Report an error · Original article (in Italian)