Not invented here

Two posts ago, I said I was worried about the Trump administration's attack on universities, but had more urgent things to write about. I want to return to that topic today. It's really important.

Fundamental research in science pays enormous dividends to society. We're surrounded by examples – your mobile phone, the GPS system, vaccines, improved treatments for deadly disease, low-cost solar energy generation and so much more come from research done by scientists investigating important problems, not sure where their research would lead, and not focused on an immediate financial return. They discover and invent. We all benefit when those ideas are taken up, made available to everyone.

How research works in America

Companies don't invest in fundamental research. At best, they carry out applied research – work on scientific problems directly relevant to their businesses, where they can foresee a new product that provides an attractive return on that investment. Google and Microsoft and IBM all operate research labs today, but they concentrate on work that are likely to become products or services that the companies can sell to their customers.

In the US, fundamental research is done almost entirely at "R1" and "R2" universities, which award doctoral degrees and concentrate on scientific research.

At these universities, professors teach, but their primary job is to train and mentor PhD students. To get a PhD, you need to discover new things – to advance the state or the art in your field. A PhD student has already completed a four-year bachelor's program at some university. The professor and student work together for four or five years more, researching a particular problem deeply. Professors generally have several PhD students at the same time.

A junior professor isn't particularly well-paid. Over time, though, with some success in her field, she becomes a senior faculty member and is able live comfortably (not extravagantly!) on her salary. The PhD students are paid small stipends that keep them in school. They share apartments with a few roommates and live quite modestly. Once a student earns his doctorate, he may turn into a junior faculty member somewhere, or may get a job in industry earning a good salary.

The system produces a steady flow of interesting ideas from smart people working on hard problems. It's surprisingly inexpensive, given how effective it is.

I attended an R1 – UC Berkeley – and worked as an undergraduate and graduate researcher on the POSTGRES project, exploring new ideas in database management. Those were my cheap-beer-and-ramen-noodles years. I'm not complaining, because I loved the work and we did really cool and important stuff. Things turned out okay for me, eventually.

The vast majority of R1 research funding comes from the Federal government. The National Institutes of Health (NIH), the largest funder, spent nearly $40Bn on "external" (university) research grants in 2024 and almost $5Bn more supporting work by scientists it employs directly. That same year, the National Science Foundation (NSF) administered over $7Bn in research funding in a variety of scientific fields. The Department of Energy awarded grants totaling just over $22Bn. The Defense Advanced Research Projects Agency, or DARPA, spends about $4Bn annually on research, mostly in academia, though some of it in industrial labs. DARPA invests primarily in technologies that it believes matter to national defense, but those often have important commercial applications as well. DARPA paid for the research that created the internet, for example.

Professors write grant proposals for these agencies to support one- or multi-year projects. The agencies get lots of grant proposals and fund only some of them. They rely on experts, including peers of those professors who are working at other universities or who do tours of duty working in Washington DC for the funding agencies, to choose the most promising. A grant, when it's awarded, is basically a contract between the Federal government and the researcher: You'll work on this problem, spend the money according to these rules, report your results by such-and-such a date, we get to audit you and check your work.

There are other granting agencies in the US government, but NIH, DoE, NSF and DARPA are the biggies. They added up to more that $72Bn in research funding in 2024. That's real money! But all together, it's just one percent of the total US 2024 spending of nearly 7 trillion dollars.

There's simply no alternative today to R1 and R2 universities as the fundamental engine for scientific research in the US. No other institutions have the staff, the systems or the institutional expertise to do this work.

And there's no alternative today to Federal funding to support it. Public universities, like my alma mater UC Berkeley, don't have big endowments they could tap to pay for this work. There's no collection of funders waiting in the wings to pick up the tab. If Berkeley's NSF grants disappear, then that work will simply stop and the people who do it will have to leave.

Private R1 universities like MIT and Columbia do have endowments. Those are used today for a variety of purposes, but generally not for research. If the endowments are tapped to pay for the research, then the endowments will be drawn down over time – the seed corn will get eaten – and eventually, that work will stop, just as at the public R1s.

And it's not just the research that will stop. Professors are teaching undergraduates as well. Getting rid of researchers gets rid of teachers. The four-year college system in America will be badly damaged.

You don't briefly interrupt a system like this one. Once professors and PhD students disappear, they can't be conjured again out of thin air. The pipeline of new faculty members is gone – altogether, or overseas, but certainly gone in America. We won't be able to rebuild the R1/R2 system in less than a generation. And the Trump administration has proposed no alternative to it.

The Chan School of Medicine at the University of Massachusetts has rescinded offers to new PhD students for the 2025/2026 academic year, due to NIH funding cuts.

Indirect costs

The administration has attacked "indirect costs" at NIH and elsewhere as particularly egregious. They cite them a reason to cut research funding across the board. It's kind of inside baseball, but I want to shine a brief light here as well.

When a professor receives a grant, there is a provision for some of the money to go to the university where she works to cover overhead costs. The students need office space and electricity to run their laptops. Administrators have to manage the facilities and budgets that the professors rely on. People need toilets and sidewalks and so forth. Indirect costs, these days, can be half of the value of the research grant.

It's useful to understand where the indirect cost provision came from. You can read deeply on the topic if you like. Just the first paper in that collection, by Robert Rosenzweig, is meaty and likely to be enjoyed only by university administrators who do battle with government and professors for a living.

The short version of indirect cost recovery is that, until the Reagan years, most grant proposals spelled out all the associated infrastructure and support costs incurred to do their work as part of their proposals. That made the financial part of the proposals long and complicated. It meant the professors had to be super careful to account for every cost, and that government auditors and university administrators had to track an enormous amount of detail. That was time-consuming and expensive.

So, in the 80s, the government began to ask for an "overhead" budget, no longer requiring a detailed accounting. Just add 15% to the basic research costs and spare us the details. Saved trouble and time!

Those overhead numbers have crept up over time. It's perfectly legitimate for funders to negotiate better rates! Professors and funders are actually on the same side of this issue – both groups would like to see the indirect costs assessed by the universities go way down.

But making that change is likely to push a lot of the cost detail back into the core grant proposal so that it escapes any indirect caps. People will still need toilets and electricity for their laptops, after all. A much longer list of direct costs will require grantors and universities to do the expensive and troublesome tracking and auditing work – the old problem, back again. It's not at all clear that grant totals would actually go down.

Surrendering global leadership

All major research in the world works on the US model. In Europe, Japan, China and elsewhere, public institutions fund researchers working in academia, very much in the American way.

But America has been particularly good at it. The very best researchers world-wide have chosen for decades to pursue faculty appointments at US R1 and R2 universities. The most promising students have applied for international admission to those same schools. We have developed a set of immigration policies and visas that permit those students to come here to study, to remain in the country when they graduate, and to spend long and productive careers here. A number of my good friends are foreign-born, naturalized US citizens who came here for grad school. I'm glad we have them.

Our global prowess in science and technology, our innovation and our leadership, are a direct result of our outstanding public support for academic research.

If we gut research funding, then the next generation of bright American and international students just won't do PhD work here. Some will do other things – get industry jobs, switch fields. Some will go elsewhere, because China and Europe are contemplating no such evisceration of their universities.

And that will cost us vastly more than any money we save. Just in my own field of computer science, we're on the cusp of tremendous advances in artificial intelligence. We're making powerful new quantum computers that will let us solve problems that are simply intractable today. Developments in theoretical computer science give us new algorithms, tools and languages for putting computers to work in new ways. We won't make those advances. Some may happen elsewhere. Some simply will not.

The story is the same for energy and climate change, for medicine, for materials science, for astronomy and fundamental physics.

"Basic research" sounds like it's expendible. In tough times, why not save a few bucks, right? But as a fraction of the money the government spends, it is miniscule. It produces enormous long-term benefits for society, and that loss has to be reckoned in any honest accounting. Killing it also guts our four-year college system.

Defunding this research will mean that the future will no longer be invented here. It may not be invented at all – but it will certainly not happen in America, because we will lose the smart people and the effective institutions that know how to do it.

Note: I updated this post to include research funded by the Department of Energy, which I omitted from my original accounting and which was pointed out to me by a researcher who reads the blog. I also added a link to the UMass Chan school's rescinded admissions.