Reading: Financing and Experimentation

The second important value driver is the cost of experimentation. Recent technological developments have dramatically lowered the cost of experimentation and therefore the cost of starting a new business.

Lower experimentation costs spur long-shot startups

The internet, open source software, cloud computing, and the emergence of subscription-based service models allow entrepreneurs to start firms with essentially zero upfront capital expenditures. As the well-known venture capitalist Mark Andreessen put it: 

“In the 90s, if you wanted to build an Internet company, you needed to buy Sun servers, Cisco networking gear, Oracle databases, and EMC storage systems... and those companies would charge you a ton of money even just to get up and running. The new startups today, they don’t buy any of that stuff... They’re paying somewhere between 100x and 1000x [less] per unit of compute, per unit of storage, per unit of networking” (cited in Nanda and Rhodes-Kropf, 2015).

As experimentation in sectors that benefit from these technological developments has become cheaper, it is more likely that investors back startups with a low probability of success, simply because an initial exploration into the viability of the business case requires very little upfront investment.

We can illustrate this point with the initial example:

• We have seen that the expected value of the drug project increases from -2 million with a full upfront 2^nd-stage investment to 8.8 million with the initial experiment. Put differently, the maximum rational investors would be willing to pay for the experiment is 8.8 million, as this cost of experimentation would set the overall NPV of the project to 0:
   
  \(NPV_{\text{8.8M experiment}}=-8.8 + 0.1\times 88 + 0.9\times 0 = 0\) million.
   
   
• Suppose technological advances in analytics and simulation have lowered the cost of experimentation from 10 million to 1 million in recent year. Until recently, the drug project would therefore not have been started because the information gained from the experiment did not justify its cost (if the experiment costs 10 million, the project NPV is -1.2 million, i.e., -10 + 8.8 = -1.2).  

Technology spillover effect

The first important implication of this result is that technological advances in one sector can have a far-reaching impact on the viability of projects in completely different sectors.

For example, advances in super-computing now make it possible to simulate complex chemical or physical reactions. As a consequence, experimentation in the chemical, pharmaceutical, or nuclear sector have become much cheaper: Instead of having to build test laboratories or test reactors with massive upfront investments, startups can now rely on comparatively much cheaper simulation models for an initial exploration into the viability of their business idea. While investors might be unwilling to back the construction of a multi-billion test reactor for a long shot project in the nuclear sector, they may very well be willing to invest a few million to obtain proof of concept for the same idea with a computer-based simulation. Not surprisingly, such a dramatic decline in the cost of experimentation triggers a wave of new startups and redirects risk capital into these sectors.

However, as the supply of risk capital is relatively inelastic in the short term, this trend has the unpleasant side effect of making it increasingly difficult to finance projects in sectors that do not benefit from lower costs of experimentation.

Attractiveness of large-scale projects increases

The second and related takeaway is that a lower cost of experimentation makes it increasingly attractive to explore the viability of long-shot projects that eventually require massive investments. Again, we can easily see this in our initial example:

• With a full upfront commitment, investors would be willing to invest a maximum of 10 million (instead of 12) to get a 1% chance to make a billion, as this would set the project's NPV to exactly zero (NPV = -10 + 0.01 ×1'000 = 0 million).
    
• In contrast, if the described initial experiment can be run at a cost of 1 million, investors would be willing to provide up to 90 million for the second-stage investment! To see this, consider the following:
  •  With a 2^nd stage investment of 90 million, the expected value of the drug is -90 + 0.1×1'000 = 10 million if the initial experiment produces a positive signal.
  • Since, according to our assumptions, the probability of a positive signal is 10%, the experiment can be viewed as a project that pays 10 million with probability 10% and nothing otherwise. 
  • Consequently, the expected value of the experiment is 1 million (= 0.1×10). This is equal to the cost of the experiment, so that the overall NPV is zero.
      
• Using the same logic, it is clear that a further decline in the cost of experimentation would justify exploration into even more audacious projects. For example, if the cost of experimentation dropped to 0.1 million, investors would be willing to back a 2^nd stage investment of up to 99 million, all else the same.

As learning becomes cheaper, it becomes more attractive to inquire into large unchartered territories.