Background
The Internet Computer’s novelty in the marketplace (a replacement for the internet technology stack!) and the drastic price discovery volatility of blockchain technology in general, makes it extremely challenging to model and forecast a potential market opportunity for the Internet Computer. This article will suggest a mental framework that might be useful to the community and investors.
Comparison Model
In order to utilize existing financial models and publicly available metrics, it’s useful to define the Internet Computer into an already understood comparison model. For these purposes, I tend to think of the Internet Computer as a cloud-computing service, similar to Amazon Web Services or Microsoft Azure, albeit one with unique competitive advantages and disadvantages. By doing so, I am able to apply basic business financial models and I am able to challenge model assumptions against real world data.
Market Size Model
The simplest (and laziest) way to model the Internet Computer’s market opportunity is through a simple market size calculation. This formula is simply S * M where S is the total size of the cloud-computing services industry (revenue per year) and M is the market share percentage that the Internet Computer captures. This is convenient because there are many good sources for analyst predictions of S, for example, here is one example of an analyst predicting the total size of the cloud-computing industry to be ~$1.4T by 2030 and another predicting $1.6T. However, the IC’s market share is much harder to forecast.
If you have the total market size and a good estimate of the Internet Computer’s market share, you can easily predict a price for the ICP token by simply multiplying those two numbers by a revenue multiple (r) and dividing by the total ICP tokens (T). Let’s see how this formula might be used to predict the price of ICP in 2030…
S = $1.5 T (average of the example research reports provided above)
M = 3% (this assumes that the Internet Computer has captured 3% of the total cloud computing value generated in 2030)
r = 8 (revenue multiple for a growth company)
T = 760M ICP tokens in 2030 (based on a conservative model I have that assumes the IC is burning more ICP than node rewards starting in 2024)
Using these values, ICP would have a price of $473 in 2030. However, this model, as stated above, is a lazy model… revenue multiples vary widely across and within industries. For example, at the time of this writing Amazon’s revenue multiple is 2.4, while Tesla’s is 11.5. Revenue multiples for fast growing companies can be as high as 100. We want a more understandable model. Let’s get more nuanced…
Gross Earnings Model
If we define the Internet Computer’s “gross earnings” as simply the number of ICP burned for computation minus the number of ICP minted for node rewards (treating governance rewards as dilution or dividends and not operational) then it is simple to predict future earnings as simply n * Op, where n is the total number of nodes on the Internet Computer and Op is the average node profitability (in ICP terms). There exists a technical hardware limit to the amount of cycles a node can can burn in a given time. A while back I calculated that technical limit to be between 7500 XDR and 12500 XDR per month, but it’s worth pointing out that these values could be drastically wrong because it is outside of my area of competence. It’s also very likely that the Internet Computer network will not run at full capacity, allowing excess computational capability for peak demand periods and network growth. I estimate the second component of Op, node rewards, at 750 XDR per month per node. It’s worth noting that current node rewards are around 1250 XDR per month per node, however it’s likely this value will drop significantly as (a) nodes move to more cost efficient locations, (b) risk premiums on the IC decrease and (c) hardware prices drop. If we assume that an average node on the IC burns 4000 XDR worth of ICP a month in 2030 (using 12000 XDR as the technical limit and the IC executing at 33% of capacity) and costs 750 XDR a month, then the Op would be 3250 XDR in 2030.
The number of nodes on the Internet Computer can be forecasted using market size and market share. Currently Microsoft Azure has about 4M “nodes” and has approximately 21% of the market share. This implies that there are approx. 19M nodes in the current cloud computing services industry. To forecast the number of nodes on the Internet Computer at a future point in time, simply use the formula n = 19M x [market growth multiple] x [IC market share %]. A useful value for the market growth multiple for 2030 forecasts could be 4.
Assuming a 3% market share for the Internet Computer in 2030, it’s feasible for the Internet Computer to be generating ~ 7.4B XDR/month in 2030 (~$9.8B at today’s exchange rate). This is simply 76M industry nodes * 0.03 market share * 3250 XDR/month Op.
Forecasting gross earnings is a great exercise because it allows us to back into ICP price with some level of consistency across industries by using price to earning ratios (P/E). A P/E ratio of 18-20 is consistent for growth companies between industries and throughout modern history. If the Internet Computer generates $9.8B in earnings each month in 2030 ($117B for the year) and there are 760M ICP tokens in existence and we apply a P/E of 18, the implied price of ICP would be $2700 in 2030 (18 * 117B / 760M). In this analysis the term “gross earning” is identical to “net earnings” because there are no R+D and overhead costs to the Internet Computer.
ICP Equilibrium
The above analysis assumes that IC demand is driving the growth in the number of nodes, which allows the IC to maintain a robust, positive Op. In a later article I’ll explore whether this demand growth is feasible. However, if the above assumptions are valid then the implication is that the Internet Computer is operationally profitable (ICP burn > node rewards) in 2024 and deflationary (burn > node + governance rewards) in the middle or late 20s (depending on ICP price).
If you’d like to explore a future price for ICP under different assumptions, you can use the formula: $p = P/E * n * [$ burn per month per node - $ node rewards per month per node] / T.
Next Article
I want to keep this article short and I plan to write another article in the next few days in order to explore the feasibility of assumptions in the above calculations. Namely:
If the cloud-computing industry grows by 4x in the next 8 years and the Internet Computer captures 3% of market share, that would mean the IC network would have 2.3M nodes… is that even possible?
Is the demand growth required to reach the above assumptions in 2030 reasonable?
Does the above analysis produce a market capitalization that is reasonable?
Is the above calculations for Op reasonable given the competitive landscape of the cloud-computing industry?
Is 760M ICP tokens in 2030 likely? Is it appropriate to use total tokens in existence, given that ICP tokens have other purposes (like liquidity pools) other than ownership of the network.
Community Request
I’m interested in challenging all of my assumptions above, particularly around the burn capacity of a node. If you have an understanding of hardware limitations and the how they could limit the cycles burned by a node, please reach out to me via Twitter @kylelangham or comment below. Also, I’d love for the community to discuss all of the assumptions made above and produce it’s own analysis for comparisons. What might a more bearish analysis look like? What might a more bullish analysis look like? What other comparison models should have been used for analysis?
Modeling ICP Long Term
Great analysis Kyle, very helpful! If the system turns deflationary, then wouldn't the terminal ICP outstanding in 2030 be way less than the assumed 760MM tokens? Hence implies a significantly higher target price? Many thanks