Ask HN: How do you start a research based company? | Hacker News Reader

Ask HN: How do you start a research based company?

73 pointsmnky9800n38 comments

Looking around hacker news it seems like everyone everywhere has their new AI company whose main goal is to develop some kind of new algorithm and then find customers later. Where do people get funding for such initiatives? I believe I'm a bit naive but it also seems like this could be a better way of doing research for the time being than continuing on in academia. But how do you get money to start a company whose goal is "make AI and worry about customers later"?

View on HN

The term you need is focused research organization (FRO). Eric Schmidt funds some via https://www.convergentresearch.org. There are other similar funding entities devoted to specific domains.

Nature has written a few times about FROs, see https://www.nature.com/articles/d41586-022-00018-5.

It is worth noting that more conventional for-profit startups which are heavy on R&D can also benefit from government grants and tax rebates.

I am also trying to spin off from Academia. If you are interested in connecting, see email in my profile.

This is true, and the thing I didn't realize before starting a company is that you get lots of money back through these tax rebates if you are pre-revenue since you have expenses and no retained earnings.

I believe you get money back from tax credits. But tax rebates only occur when you deduct expenses from current income and rebate money you've already overpaid in as taxes, which, if you have no current income you would not have paid in these tax amounts, and means expense amounts go in a "savings account for expenses" to use whenever in the future you have income. (that's a little too complicated but at a certain point I had to quit)

I think you're talking about tax credits.

Ah, yes sorry I am talking about tax credits.

Years ago I became acquainted with a company that was entirely research driven, so had the same question as you - how the heck do they actually make money? I came to the following conclusions:

-- Have a founder / CEO / lead type who is well-connected and can play the part of "mad scientist." For this company, it was a guy with a phd in something data related who was very good at storytelling.

-- Find customers who can benefit from your research. In this company's case, it was government (which can take a lot of forms - cities, school districts, etc), non-profits, university researchers... from what I could tell, they were good about letting someone else find the grant dollars and they would provide the research. Over time they were able to take their reputation from those projects and then start selling the same process to corporate America.

-- This company had very little technology or even secret sauce. Sure, they tried to make it look like they had some magic, but once you started peeling away layers, it was clear that they were mostly just doing a lot of data crunching that no one else quite had the desire or skills to do. They delivered a result rather than a lot of worry about how they got there. That same message likely applies today - lots of young companies focusing on some magic new algorithm, but paying customers generally care a lot more about a result that has value to them over the tech you used to get there.

A research-led consultancy? Presumably that only takes on clients that align with the product they want to build?

Speaking from Graphistry experience and now louie_ai, both of which are real tech:

* Initial Graphistry funding was from showing compelling 10X technology demos started in my PhD through my extended network, which got us angel & seed and first early adopter customers, which let us hire a tiny team to build the real tech (which pivoted at our first 'big' customer) and start on our product discovery journey . Before then, ran out of my savings, and during a funding dry spell, same. All new R&D for last few years has been customer funded.

* Louie.AI: We have been reinvesting revenue here and signing on paying design partners to co-build with . (If you are on Splunk, OpenSearch, Databricks, AWS Neptune, or a graph DB, and want to use genAI for investigation & analyses, lmk!.) Skipping VC $ has felt much better as we are much more aligned with our design partners.

If I was going to do it from scratch again, I would either do a PhD and spin out when ready, or do a consultancy and build out from that. Using VC dollars and hoping you march into product market fit is horrible, a lot has to work out just right, yet you are trying to cook with a flamethrower and short fuse.

seems like this could be a better way of doing research for the time being than continuing on in academia.

I'm not sure that's an elevator pitch to wow investors. I mean "I want to do research" sounds more like a career objective than a business. There's nothing wrong with that.

Startup seed money is a bet on potential.

If you really want to start a research company, start a research company and figure out the money later. If you want to escape academia, you might just get a job in the industry.

Good luck.

A slick "investor" deck, a plausible-sounding technical deck, and a few connections goes a long, long way.

Correct

Got any examples of these?

Theranos?

Y Combinator calls these companies "hard-tech". They have a great video here: https://www.youtube.com/watch?v=C1DlZWfI6rk

If you were already well-known researcher in the field with good publications, then it’s possible to get significant funding to continue that research in an applied manner.

But if you don’t have a track record of excellent research productivity, then it’s tough to get funding to do research because it’s highly competitive and it’s winner take all payouts.

> But how do you get money to start a company whose goal is "make AI and worry about customers later"?

To a first approximation, you don't.

What is it that you want to achieve? Post-doc like research nirvana[1] with a longer time horizon? Solving a particular problem?

[1] YMMV, but in the best case, that's some version of optimal

You can do it if you can build a clear narrative of why applied research in some area now will build you a uniquely strong position to monetize later. It's not a "funding a research based company" so much as "buying an option on unique talent's work".

Maybe i'm cynical but my impression is that most productless companies with lots of storytelling around some phd's are almost always about scamming investors / grant money / public funds.

If anyone have great non sensational examples of the opposite i'm all ears.

I remember becoming aware of this during the crypto wave where there was always a bunch of "researchers" with PhD's around paper-thin whitepapers - and these researches were often from very prestigious universities, and of the hundreds of these companies no one produced anything valuable - and i've seen the same thing in many sectors like nutrition, "healthcare", financial products etc.

Had a friend leave academia for a research based company, same bullshit (begging for grants, gaming metrics, plowing hours and hours into areas of research that are pretty much guaranteed to be fruitless but sound sexy) different day.

One path is pursuing SBIR grants from the federal government (if you are based in the US). There is a wide variety from defense to agriculture to medical - but the grant process can be competitive!

I may be dating myself but back when I was involved a program normally consisted of a ~6 month/$100k feasibility study and a 1-2 year/$1-2M work plan.

https://www.sbir.gov/

I've worked for a few startup that were grant funded and its not a great approach for funding. The grants have their own goals, and while there is some overlap, they usually don't align with the companies ultimate goal. You end up spending a lot of money but not reaching to goal post. Then there is the issue of rolling the dice in getting the grant.

A paying customer is much easier to manage and predict.

I totally agree with you. It can be hard to chain together grants to make a cohesive business. But if you want to have funding for pure research, it is one option available to anyone (US citizens at least)

Don’t start a research based company for money, start it for clout. You can say you are researching cutting edge AI stuff, but not really. And no one is going to validate your claims. It’s research. It’s mysterious, exclusive, inaccessible, and sometimes it’s expected for nothing of value to be produced. A research based company could be as simple as a legit looking website with some relevant and highly technical blog posts pushed out to social media.

Best part, you can use this clout to pick up a few contracts along the way and do some work for clients with actual interest in the field. There’s your money.

While studying my post grad some friends and I started a very prestigious computer vision research company. It was all presented as a very serious co with computer scientists who knew what they were doing. We had $0 in revenue, and our demos were mostly just us playing around with various off the shelf libraries behind the scenes and presenting it to the masses. This gave us glowing resumes and credibility in the industry. We later got real jobs with big tech companies.

This is really interesting.

Can you please elaborate? How were you able to write highly technical blogs without seemingly doing anything that technical (playing with libraries)? Did you have to reach out to big companies or did they reach out to you?

It'd be great if I could contact you to ask you more.

If you already have a background in Academia or FAANG doing research, (some) investors will happily fund you, if and only if, you can convince them that the research you're doing will be monetizable. AI has enough precedent that you're seeing this in the news, but most already have a research background.

Chicken or the egg problem, unfortunately

Nobody is actually raising money on "make AI and worry about customers later." Every company that raises significant funds has either a monetization or exit strategy. You're not going to get a callback without it.

Maybe there's a niche market that they're targeting and they're just not being public about it until they're ready to enter that market, so as to not tip off the competition.

Or maybe they know that AI is important to many companies with deep pockets, and they convince investors that they have a better strategy than that being pursued by Google, Facebook, OpenAI, etc. In which case the plan is not to get customers, but to pull an OpenAI and leapfrog the state of the art, thereby becoming a prime acquisition target.

Yeah if you think investors are just throwing hundreds of millions of dollars at companies with zero growth or exit strategy, then you are naîve.

Usually the founders have a strong AI/ML background and a big vision. Combined with the current FOMO of investors they get a lot of money based on that.

I am sure most of these companies will not receive any funding beyond their seed and series A because they weren't able to identify a product that they can sell.

There's a few factors. Are there that many, or is there just a lot of news? And related to a lot of news, there are a lot of orgs chasing AI whether they understand it or not. So to that end, as a founder, one could convince the naive (about AI) money to invest. Or more likely, the founders have a track record in some way, and then it is not much different than raising for a traditional startup.

I've thought a lot about these issues - I am actively working on creating new research and then commercializing it. However I think the incentives of investors (VCs and likely angels) are not aligned well with research development. As a result, I've landed on the bootstrap/self fund side of the argument, much like Midjourney. Find the low hanging fruit on the research side that can be monetized, and build off that.

There was a user here who wanted to do AGI research. They left their job and went all-in studying then tried networking and applying to labs. After being unsuccessful in those applications, they went all-in building an algorithmic trading business and made enough money to start their own research lab.

Link?

Many times, government money. The US government could credibly considered the largest, longest running, and most successful VC fund in history. They're also one of the only organizations that regularly funds truly blue-sky research.

Otherwise: venture capitalists, but I wouldn't consider those companies "research" in the traditional, academic sense. Most times, at least in AI and the present climate, VCs are betting on that company to develop some novel tech that will be acquired a big markup by a larger company who deems the technology valuable. Not sure that is so much "research" as "niche product development with a very targeted exit strategy".

One thing to look at here is Small Business Innovation Research grants. I know a lot of companies who exist just to do SBIRs and trying to productize other SBIRs.SBIRs are also a lot of fun, tend to be more research focused, and often you can get a decent amount as part of the grant.

Varies by dept to dept but some of the phase 1 grants can be up to a quarter or a half million dollars.

They’re grants, too. If your research flops, you’re not on the hook for it.

If you make it to phase three, congratulations! You’re now a supplier of a government procurement contract, and the government paid for your entire R&D effort, plus your salaries along the way.

Remember Jack Barker's wise adage, "the stock is the product."

Very hard to always be grant driven. That is a massive burden.

Could you please expand to share experiences? Thanks!

We had a startup that ran out of income generating projects / sales. We tried to bridge the gap - extend our runway with research grants. We got some money, and were successful in some smaller grants. It was necer enough to let us breathe and relax. In the end the business failed when the money was out.

What grant resources online do you use?

I was having similar thoughts last week: academia makes no sense for any well grounded intellectual, and I will start a research foundation. But how to make it economically sustainable?

Reply on news.ycombinator.com