What makes technology research commercially viable?
It depends. We don't talk about "what on?" enough.
At Speculative Technologies, we frequently grapple with the question “When should an idea that smells like research be a startup?” If something makes sense as a startup, it should be: markets are incredibly effective at supporting commercially viable technologies. But at the same time many promising technologies have been crushed under investors’ growth or profit expectations: we want to be a home for these technologies until they can survive in the market.
Let’s start with a toy example of technology-building work that will take $100M and 10 years to get a technology to a point where it has an expected value of $200M. As an investment, that $100M will have an internal rate of return (IRR) of 7%. If an index fund will return 7% a year, that work isn’t commercially viable: you could just put your money in the stock market and make the same return with far more liquidity and less uncertainty. It certainly wouldn’t be a good investment for a venture capitalist, who needs the possibility of much higher returns.
The more general version of the question about startup viability could be framed as “when is research commercially viable?” or more jargon-y “when will the market make sure research will happen and when is it a market failure?”
(This piece will ignore the important questions “how can you tell the difference between commercially inviable research that is nevertheless valuable (market failures) and research that is just a waste of resources?” and “is there even research that is both valuable and not (yes) commercially viable?”)
Everybody has different answers to all of these questions: some people think that anything that smells anything like a product is either commercially viable or not actually valuable, others think that government should play a much larger role in supporting research. People often treat these answers as laws of nature that are invariant across time and domains.
I want to argue that like most big meta-questions the answer is “it depends.” Endogenous attributes of a technology can make iteration cycles faster or slower, demand more or less equipment, or create more or less uncertainty about whether the work will pay off. Exogenous factors like economic, cultural, and regulatory conditions shift timelines, profit expectations, and motivations. Like many situations where “it depends,” we do a poor job of breaking down what it depends on. This piece is my attempt to start unpacking those dependencies.
Going back to our toy example, there are many things that could change commercial viability:
Shorter timelines: If the work only took 5 years, it would have an IRR of 15%.
Leveraging pre-commercial work: if you could get the first five years funded through mechanisms that didn’t expect a financial return, funding the last $50M over 5 years would be commercially viable even if the valuation (or the fraction of that $100M it bought you) was lower — turning $50M into $70M over five years has an IRR of 23%.
Worse comparables: If the stock market had worse returns, say you expected it to return 5% or it was far more volatile, locking up your money for a 7% return would be commercially viable.
Cheaper work: If the work were cheaper, obviously that would increase the IRR.
Portfolios and huge upside: If the upside of the work is potentially huge, you could bundle this work with other work into a portfolio.
Decreased uncertainty: If there is much more certainty about the value of the work or it’s more likely to succeed (say you’re making a drug that has an automatic market if it gets through the FDA) the expected value of the technology will go up.
Better value capture. It is hard to capture the value of a lot of technology research. Studies suggest that inventors only capture 2% of the social value they create and the inventors of many important technologies died in poverty. Figuring out better ways of capturing that value also increases the expected value of the work to build the technology.
(Changing the sign on any of these things — longer timelines, worse value capture, etc. — will make the work less commercially viable.)
All of these factors can vary over time and between technologies. Some obvious ways things can vary:
Markets and interest rates change.
Underlying technologies are built up and tapped out.
Industries have very different dynamics. Software has very different dynamics than life sciences technologies and both are very different from materials or manufacturing.
Value capture can vary based on where a technology is created. A technology that was developed in a large corporation that can immediately deploy it and tune it to their needs can have a very different value capture profile from one that was initially developed in a university, spun out into a startup, and then needs to be integrated into a large corporation.
People gain more knowledge about what technologies do and do not make good businesses. (It’s possible that corporate research declined in part simply because it took corporations a few decades to realize that wide-ranging corporate R&D was far less profitable than they expected.)
And the list goes on!
One level down, costs and timelines are context dependent
Costs and timelines themselves are context dependent. Both endogenous attributes of a technology and exogenous factors can have huge effects on how much time and money it takes for it to reach commercial viability.
Endogenous Attributes
Different technologies and fields have “characteristic timescales” that impose limits on iteration speed, which in turn has a big effect on timelines. The spectrum runs from software, where often you can alter and test things about as fast as you can think them,1 and horticulture where you need to wait for a whole plant to grow. New technologies, like better physical or digital models, change iteration speed.
Predictability affects timelines by determining how much you need to iterate. We’ve made things like static building structures pretty predictable (at a macro scale) but to a large extent software performance is pretty unpredictable (luckily you can iterate quickly). Predictiability in turn is downstream of modelability and measurability.
The nature of scaling2 something also affects timelines and cost. Scaling can be effectively free (as in the case of distributing software that runs on other people’s local machines) or scaling can require many iterations with full scale physical systems, effectively reinventing the entire process from scratch.
Equipment affects both cost and timelines. Some work requires tons of custom equipment like femto(10-15)second laser pulses or scanning tunneling microscopes that can resolve individual atoms; custom equipment is both expensive and can drastically lengthen timelines. Some work requires expensive but standardized equipment (a lot of synthetic chemistry and cell biology falls here). And some work just requires a laptop.
Skillsets affect costs and timelines in similar ways to equipment. If you need a lot of specialized skillsets, it’s usually both expensive and hard to find the right people.
Exogenous Factors
Imposed Frictions
A lot of exogenous factors boil down to “how much friction do other people impose on doing the thing?”
In 2024, building a new software product is one end of the imposed friction spectrum: where you can do almost anything on your own computer, spin up an LLC on Stripe Atlas, and then either distribute files or stick the thing on AWS without ever talking to a person or leaving your house. Building a new kind of nuclear reactor is probably on the other end of the spectrum: you need fissile material that, without permission, will get you arrested at gunpoint3 and in order to build a pilot plant, you need to iterate with the NRC over the course of years (and pay millions of dollars for the time they spend!)
Imposed frictions can range from getting permission for a building to extensive lawyer-filled negotiations over contracts. Many imposed frictions are culturally mediated: they’re downstream of how many people have veto power over the work to invent and discover, individual gatekeeper’s sense of urgency, and their ability to unilaterally make decisions.
Not all imposed frictions are from the government! Spinning technology out of universities can involve massive bureaucracies and many private organizations dither on decisions and require reams of paperwork. The amount of time and effort a research org needs to spend figuring out what large companies would pay for and who would pay for it is a huge imposed friction.
Transaction costs
Transactions costs include imposed frictions but many other effects — from how hard it is for people to know that work is good or to find the right person to talk to. Transaction costs in technology research also vary over time and between technologies.
Expectations
People’s expectations also have a big effect on timelines and costs, both the expectations of the people building the technology and the people buying it.
Builders’ Expectations: If people with the expertise to do a thing expect to live in high-desirability places or expect to be paid a certain amount (the two or often often coupled!) that can drastically increase the cost of building a technology. People have expectations around working conditions, equity, and status. These factors are in turn affected by both culture and markets: the same work will have drastically different costs in a world where a scientist has no better options than to move to the midwest to help spin up an industrial research lab for reasonable but not great pay vs expecting to live in San Francisco and get both equity and good pay.
Buyers’ Expectations: Customers can expect very different levels of polish and reliability from technology products. Early cars broke down constantly and were little more than an engine, drivetrain, seats, and controls all encased in metal; modern cars have everything from cruise control to air conditioning, to remote start systems. Today, most people expect new products to have the polish and reliability of an iPhone, even in new categories. These expectations can add time and cost to building technology to the point where you can start capturing its value.
So what?
If nothing else, I hope you’re convinced that there is no hard and fast rule for answering whether the work to create a given technology is commercially viable. We all have patterns of “how technology creation works” from domain-specific experience or stories we’ve imbibed: from dorm-room-created software, academic-lab-spun-out therapeutics, or industrial-lab nurtured devices. None of these patterns are universal across domains or time.
We need to do a better job of asking the wonky question “what are the endogenous attributes and exogenous factors that combine with a specific technology’s state today to make the work to build it either commercially viable or not?”
You can have normative beliefs about what should be commercially viable, but the reality is that many domains today have factors that make significant chunks of work to build many valuable technologies commercially inviable. Identifying those factors and trying to reduce them so that more technology work is commercially viable is incredibly important – markets are powerful. However, that work will take time and (I suspect) there will always be some amount of technology work that is not commercially viable; we need new structures and institutions to support the valuable technology work that is not a great investment.4
Clearly there are exceptions to “you can iterate on software at the speed of thought” like software that takes a long time to compile or AI that takes time to train etc.
Talking about scaling raises the question “what does it mean for a technology to scale?” but for the sake of simplicity I’m going to use the trite definition that scaling means “creating a process that can make enough of a thing at a price that people want it.”
There are, of course, pretty good reasons why you need permission to accumulate fissile material.
Obviously, I think Speculative Technologies is one such institution, but we cannot do it alone.