How decentralized is Cardano?
Charles Hoskinson, the CEO of the technology company that is building Cardano (Input Output Hong Kong, IOHK), has recently stated that “Cardano will be a 100 times more decentralized than Bitcoin”. By saying that, he refers to the fact that Cardano’s system includes monetary incentives to diversify the number of evenly distributed stake pools in the system up to 1000. Hoskinson also implicitly assumes that the 10 largest Bitcoin mining pools control >50% of the network’s total hash rate, meaning Cardano will have a hundred times as many nodes creating blocks on the network. However, is this a fair metric for decentralization? And how does Cardano compare to Bitcoin in other aspects when it comes to network control?
What is decentralization?
When looking for a definition, I usually tend to start off with a simple web search. For instance, this is how decentralization is defined on Wikipedia:
“Decentralization is the process by which the activities of an organization, particularly those regarding planning and decision making, are distributed or delegated away from a central, authoritative location or group.” — Wikipedia
While useful, this is a rather broad definition, which could be specified a bit — particularly for the cryptocurrencies space. Many different definitions and explanations of the concept exist, illustrating it is more or less by definition not as easy to quantify, particularly in a high-contrasted fashion that allows you to call something decentralized or not. One description I find particularly useful is the description of Ethereum founder Vitalik Buterin. Buterin splits the concept into three aspects:
- Architectural decentralization —the number of (validating) nodes that the system is made up off; which is what Hoskinson referred to earlier.
- Political decentralization —how many entities control te software that these nodes run on; something often also referred to as ‘governance’.
- Logical decentralization —if you cut the system in half, will both halves continue to fully operate as independent units? As Buterin describes in his article, blockchains are logically decentralized by design; there is one commonly agreed state and the system behaves like a single computer.
Since Buterin already gave the answer for ‘logical decentralization’, let’s just address the rest of this list.
Is Bitcoin architecturally decentralized?
At the time of writing, Bitcoin’s network consists of a total hash rate of 56,029 PH/s, which is a manyfold of the second ranked network (Ethereum; 138 TH/s). Right now, it would cost $520.000 per hour to perform a 51% attack on Bitcoin. While costly, as long as the benefits (e.g., being able to double-spend) outweigh the costs, it just might be worth it. As is evident in this case, that would only be beneficial for a single actor in very rare cases where it is able to double-spend a transaction worth over half a million dollars.
As Hoskinson implied above, at this moment only a handful of mining pools control the majority of the hash rate in the Bitcoin network. To be more precise, at the time of writing, the 4 largest mining pools control 51.3% of the hash rate (see figure below). Since the operators of these pools don’t have to carry the weight of the costs of the mining equipment and electricity but can choose what to do with the newly created and valid blocks, the cost/benefit ratio is likely more favorable. If these pool operators were to get together, they could in theory perform a successful 51% attack on the network.
Such an attack won’t go unnoticed. As this article by StopAndDecrypt describes, the non-mining full nodes that validate the blockchain (e.g., the Casa node or NODL) also play a role in this. At the time of writing there are >100.000 full nodes in the Bitcoin network. These nodes form an intricate network that prevents invalid transactions from being included in the blockchain.
If a successful 51% attack would be executed by a group of collaborating mining pools, these nodes will notice that a different chain has become the valid one (called a ‘reorg’, from reorganization). In theory, this could elicit another social process in which a fork could be proposed to reclaim the original chain to be ‘the true chain’. However, such a proposal would likely be dismissed by many Bitcoin purists, that may prefer an attempt to add extra hash power to the original chain in order to outpace the attacker and ensure the original chain to stay the ‘heaviest chain’.
So far, the Bitcoin network has been up for 99.98% of the last 10 years without a known successful 51% attack being executed. As Bitcoin’s price and adoption increases, it attracts more miners and people running full nodes, making the network increasingly secure. At this point, Bitcoin is therefore arguably the most architecturally decentralized cryptocurrency in existence.
Is Cardano architecturally decentralized?
At the moment, Cardano is a federated system, where IOHK, Emurgo and the Cardano Foundation are in control of the ecosystem as a whole — including all current stake pools. When Hoskinson stated that ‘Cardano will be a hundred times more decentralized than Bitcoin’ he was referring to the number of stake pools that he expects to be running in Cardano after the upcoming Shelley release; the roadmap phase that will include the decentralization of the network architecture. The codebase of this release will include an incentive structure where stake pools are monetarily incentivized not to grow past a certain size, incentivizing the creation of up to a thousand stake pools. If you want to read more about how this works, I encourage you to read my prior long-read ‘How secure is Cardano?’.
If network decentralization is defined as the number of nodes creating blocks on the network, Hoskinson’s statement would even be a slight underestimation, as it would actually be (500/4=) 125 times as decentralized. However, since one entity can in fact run many stake pools (e.g., Hoskinson has stated that IOHK will run about 80), I would argue that the actual coin distribution is more important when it comes to measuring Cardano’s decentralization.
Since blockchain addresses are anonymous and one person or entity can actually control many addresses, it’s impossible to proof that there’s no one entity controlling a certain set of addresses. Nonetheless, investigating the unspent transaction outputs (UTxO’s) in Cardano addresses is currently the best measure we have to investigate coin distribution. The following figure shows the ADA distribution amongst all current addresses in Cardano.
As the figure illustrates, the top 1.28% of all Cardano addresses hold 71.2% of ADA’s total supply (31.1 billion). It should be noted that this includes a number of very large addresses of entities like exchanges, IOHK, Emurgo and the Cardano Foundation, as is evident based on the following ‘rich list’ describing the top-10 ADA holding addresses in Cardano.
Depending on how you interpret Buterin’s definitions in the beginning of this article, this could be considered both ‘architectural decentralization’ and ‘political decentralization’. Either way, you could argue that Cardano’s current coin distribution needs be improved upon to reinforce the argument that Cardano could be considered optimally decentralized after the Shelley release.
So, what about political decentralization?
Is Bitcoin politically decentralized?
Bitcoin is open-source software, meaning anyone can audit the code and suggest improvements proposals (Bitcoin Improvement Proposals, or ‘BIPs’). Whether this code is actually implemented in the Bitcoin software depends on a rather social process, where a certain amount of consensus is needed to convince the core developers to implement it into the software.
If social consensus can’t be found, anyone can still choose to fork Bitcoin’s codebase and create their own version of the software, which either may (soft fork) or may not (hard fork) be compatible with the current network’s rules. Miners then decide which version of the software they choose to run. In the case of a hard fork, this essentially means that they ‘vote’ which version of the chain will be ‘the true state’, and thus ensuring logical decentralization that was mentioned earlier.
Forking the Bitcoin codebase in order to actually change ‘the true version’ of Bitcoin therefore only makes sense if you are confident there’s actually social consensus under the miners that your version of the software will be supported. This is why Bitcoin is partially a social innovation, rather than just a purely technological innovation. Hasu’s seminal piece ‘Unpacking Bitcoin’s social contract’ describes this as well.
Due to the necessity of social consensus under developers and/or miners to implement any meaningful change to the system, you can argue that Bitcoin is politically decentralized. However, it should be noted that to meaningfully participate in this governance mechanism, either technical skills and/or the possession of (a large amount of) computational power is needed.
Furthermore, as a result of the complexity of this social process, the actual pace of technical innovations on the base layer of Bitcoin is very slow. For many people, this is one of Bitcoin’s most important features; immutability, in this case from a social perspective. They want to keep Bitcoin’s base layer simple and secure, and hope to see eventual ossification of the protocol, meaning the base layer will remain unchanged, while innovations occur on different layers of the system (e.g., the lightning network).
For others, this lack of adaptability has been a reason to move on and start new projects, aiming to develop infrastructure layers based on a different perspective. Cardano is one such project.
Is Cardano politically decentralized?
As mentioned before, Cardano is currently federated by IOHK, Emurgo and the Cardano Foundation. IOHK does fundamental scientific research and develops the technology itself, Emurgo focusses on attracting businesses to the ecosystem and the Cardano Foundation targets community building, amongst things. While this is more decentralized than a single entity controlling everything (as was evident when the ecosystem as a whole was able to overcome a non-functioning Cardano Foundation chairman), it is currently clearly not ‘politically decentralized’ in the spirit of what the cryptocurrencies space aims for. However, IOHK has some very ambitious plans for Cardano when it comes to ‘solving’ its political decentralization.
One of Cardano’s key aspects is that it will have its own treasury, combined with an on-chain governance system that will allow stakeholders to decide which developments will be funded from this treasury, as well as which improvement proposal’s will be implemented in the system. The on-chain governance mechanism will be based on a liquid democracy (which I described in more detail in the latter part of this article).
Similar to Bitcoin, Cardano’s codebase will be open-source and anyone can submit improvement proposals to the system and social consensus will decide which proposals will actually be implemented. Unlike Bitcoin, the decision-making power in this case lies directly in the hands of the people that hold the coins. Due to the liquid democracy setup, people can either choose to directly vote themselves or to delegate their vote to someone that they feel is more knowledgeable on the topic. This means that to meaningfully participate in Cardano’s governance mechanism, one must either have the technical knowledge to construct an improvement proposal or hold (a lot of) coins.
Both governance mechanisms therefore are based on social processes, with the most striking difference being the role of miners and stakeholders.
What are the pros & cons of on-chain governance?
Cardano’s envisioned governance system, where stakeholders get to vote which improvement proposals will be implemented and what the treasury funds will be spent on, has several advantages:
- The system is self-funded, meaning a monetary incentive to create improvement proposals with a high likelihood of finding social consensus is built-in.
- The incentives of stakeholders are aligned with the long-term best interest of the system itself, as accepting proposals that damage the system’s value will also damage the stakeholders’ net worth.
- Anyone coin holder can participate in governance decisions, lowering the threshold in comparison to needing to acquire specialized mining equipment (ASICs) in the case of Bitcoin and most PoW systems.
- Using on-chain governance, finding social consensus and implementing the outcomes will likely be more efficient, leading to more rapid innovation and a lower risk of chain splits (and thus community splits).
This all sounds appealing, but on-chain, stake-based voting also has downsides. In my opinion there are two that are particularly important here.
First, while stakeholders are incentivized to make decisions that are in the long-term interest of the system, the stakeholders themselves do not necessarily have the knowledge to make the actual best decisions, meaning they could be persuaded to make illicit decisions. I think you could call this a ‘social attack’. Even though such an attack is possible in any system that depends on social consensus and the decisions can likely be reversed in a future vote, doing it in a system where the result of the on-chain voting is automatically enforced has more direct consequences than in a system like Bitcoin. At the time of writing very little detail on how Cardano’s actual on-chain voting system will work is available (this is planned to be implemented in late 2020), so time will have to tell to what degree this concern is valid.
The second downside I’d like to discuss actually can be inferred based on the fundamental design of PoS systems in general. As I also described in ‘How secure is Cardano?’, once an attacker (or a group of attackers) is able to collect >50% of all the stake in the system (which would be very expensive and likely take a long time to accumulate), there’s essentially no way for the honest minority to recapture control of the system — at least in a non-violent way. After all, taking away the attacker’s coins by forking the system would mean you’re basically breaking the system’s immutability. In PoW systems like Bitcoin, it’s also not possible to take away the attacker’s hash rate, but unlike in PoS systems, where stake is needed to create new stake, the honest minority could add new hash rate in order to regain its majority.
Is Cardano’s coin distribution a problem?
As described in ‘How secure is Cardano?’, IOHK, Emurgo and the Cardano Foundation had a voucher sale between September 2015 and January 2017. Due to regulatory uncertainties, this was done in a very specific geographic area. In total, 20% (~5.19 billion ADA) of the total supply (~31.12 billion ADA) was distributed to IOHK, Emurgo and the Cardano Foundation. According to the distribution audit, 94.45% of the remaining ADA was sold to Japanese citizens, 2.56% to Koreans, 2.39% to Chinese and the remaining 0.61% to citizens of 5 other Asian countries.
Besides the question whether this narrow geographic distribution and the current overall coin distribution impose a problem for the decentralization of Cardano (which I’ll leave unanswered), it’s also interesting to consider if this distribution method itself could proof to be a limitation when it comes to the social acceptability of ADA as a form of money.
The reason for this, is that some people in the space — particularly Bitcoin and PoW purists — seem to believe that just creating coins and selling them to the public cannot result in a form of money that possesses fundamental value. This belief stems from the principle of ‘unforgeable costliness’ that was introduced by Nick Szabo’s 2002 article ‘Shelling Out: The Origins of Money’. In his article, Szabo described that for a form of money to have value, it needs to be difficult to create and this difficulty to create should be verifiable by the other party. For some Bitcoiners, it’s therefore the actual mining energy consumption that ensures Bitcoins have value based on this principle.
If you assume this to be true, I think you can argue that the vouchers that were initially sold itself didn’t possess this property, but derived its value from the expected functioning of the future Cardano ecosystem. Whether it is ‘fair’ that these entities could create these vouchers, keep 20% of them and and sell the rest for $63 million in order to essentially crowd-source Cardano’s future development is also subjective to answer (I’ll also leave that unanswered).
However, based on the 40+ scientific papers (of which 20 peer reviewed and mostly open access available) that were produced so far and add value to the entire space and creation of a new, very rigorously developed cryptocurrency, I think you could argue that in hindsight there’s actually quite a lot of verifiable effort that went into creating these coins. Furthermore, once the Shelley release is completed and new coins can only be minted into existence after fairly participating in the consensus protocol, I’m not sure the concerns regarding unforgeable costliness would remain valid. Personally, I therefore don’t disregard the possibility of ADA potentially becoming a socially accepted form of money based on this principle — which after all, is just theory as well.
In comparison to PoW systems like Bitcoin, PoS systems like Cardano have benefits like lower energy consumption and hardware requirements, but also have tradeoffs like not being able to regain control after a successful 51% attack without forking.
PoS systems like Cardano will always depend on a different set of benefits and tradeoffs compared to PoW systems like Bitcoin. In my opinion, saying ‘Cardano is better than Bitcoin’ or vice versa is therefore nonsensical; they’re basically different beasts with different genetic properties.
For instance, it’s well possible that Bitcoin’s base-layer properties and already built-up network effects will ensure it’s function as a universal store of value and unit of account, while second-layer solutions will actually be the technologies that will capture the majority of the day-to-day use. From that perspective, it’s well possible that the lightning network will actually end up being Cardano’s true competition instead of Bitcoins’s base-layer.
It’s also very early when it comes to the evolution of the space in general and deciding which technologies will actually end up being the infrastructure layers we will continue to use for the next decades and beyond. Time will tell which infrastructure layer or layers will be used for which purposes, and to what extent PoW and PoS systems will coexist and/or supplement each other.
Edit 7–5–2019: Corrected the number of Bitcoin nodes from ~9.444 to >100.000 based on this tweet, and corrected the number of Cardano nodes that are expected to control the majority of the network from 1000 to 500.
Edit 14–5–2019: Olivier pointed out in a response that my description of the validating nodes wasn’t optimal and could be interpreted wrongly. I edited the two paragraphs below the figure with Bitcoin’s hashrate distribution to better explain the role of the validating nodes.