The Navier-Stokes problem – another view on scientific funding

Blackboard with a math lesson written on it

A bit late to the Navier-Stokes “party”, but as the absolute drama was unravelling, I was thinking about it in another perspective: what does this amount of compute mean and how can we compare it to human “brain-power” terms.

From the OpenAI ‘s own paper, it is said that “Across all attempted problems, the agents sent 4.9 million messages and used about 300 billion output tokens.”
Even excluding the training cost, various sources estimate the cost between $10-40 million (!).
A reasonable number: $22 million – keep this number in mind.

On the other hand, we have to estimate the scientific funding part – a back-of-the-napkin calculation.

Total funding for a portuguese university (all research units)

I took as reference my alma mater Instituto Superior Técnico and from their own sources, their research units have an estimated governmental funding for the period 2025-2029 of €610 million.

Because I have not found other publicly available documents and for reasoning purpose, I am not considering other ways of funding.

Breakdown by research unit

IST has 22 research units, so on average each unit will have available 5.54 million euros for 1 year.

The particular case of the CAMGSD - Centre for Mathematical Analysis, Geometry, and Dynamical Systems

We can see further that their funding is below the average calculated above and that for the years 2025-2026 their funding is around €2.5 million.

From their website, CAMGSD has 62 researchers. To account for an outdated count or that some are not full time researchers, let’s apply a 80% rate to get approximately 50 researchers.

Each researcher would have on average €25240/ year of research.

Adding the university professor salary

Usually researchers are also professors, so they get a salary from their university. Even if we maximize their salary corresponding to a full professor but with the caveat that just 2/3 of the time is for teaching, their gross anual salary is a little above €56k.

Bringing all together for the final number

A researcher annual “funding” is around €80k.

An already large “group” for such a specific topic as discovering analytical (approximate) solutions for Navier Stokes could consist of 3 researchers.

Let’s also consider they shared 5 PhD students (gross annual compensation of a little over €16k).

Our magic number, for one year of research for an already arguably large group researching one topic, would be €160k.

OpenAI spent 125 years of funding of a research group in 5 days.

Let that number sink in.

Picture of Nelson Eiró

Nelson Eiró

Senior Data Engineer

LinkedIn

One Comment:

  • LLMs for scientific research – Nelson Eiró
    at 2 weeks ago

    […] one of my last posts, I tried to calculate an estimate of how many years of funding for a small research group was […]

    Reply

Add comment: