1kx: From hardware to software, how to estimate the cost of the DePIN project?

Compiled by Elvin, ChainCatcher

Summary

Framework for estimating costs:

• Step 1: Identify network contributors
• Step 2: Evaluate the components of the cost
• Step 3: Evaluate cost structure differences and summarize

Case Study

Key Points

• In order to ensure the continuous participation of nodes in the Decentralized Physical Infrastructure Network (DePIN), network administrators (founders, DAO members, etc.) must consider the costs incurred by operators when operating nodes.
• In some cases, key decisions about cost optimization are obvious. For example, Livepeer's switch from Ethereum to Arbitrum in 2022 was a good choice without controversy, thus reducing settlement costs by over 95%. In other cases, DePIN administrators might need external assistance to evaluate the cost of operating nodes when resources for research and development are limited.
• If the node continues to incur losses, the operator will stop running the node, resulting in a decrease in overall node supply. Understanding the operating costs of the DePIN network and its main driving factors can enable network operators to initiate governance discussions; at the same time, cost estimation can provide information for research and development work before the decline in network service supply begins, in order to reduce the cost of node operators.
• For protocol managers, estimating network operating costs can be challenging because of the anonymity of contributors (these networks are usually permissionless, meaning anyone can contribute and leave at any time) and the lack of public data related to costs.
• To guide managers' decision-making, we have proposed a three-step framework for cost estimation:

1. Define network contributors and assign them to specific roles.
2. Identify the cost structure related to nodes
3. When evaluating the combination of 1 and 2, consider the differences in cost structure.

In addition to the overall estimate of the current cost, this framework also provides:

• By segmenting roles and cost components, help identify the largest cost driver
• The estimated changes under different assumptions and scenarios of increased demand/network capacity

The case study will demonstrate how to apply the framework. For example, a joint investigation with the POKT network revealed the ongoing efforts of node operators to expand their service nodes. Nevertheless, decentralizing their gateways has addressed the remaining obstacles in terms of economic scalability, including demand generation.

Introduction: What is DePIN and why discuss costs

DePIN is a series of decentralized networks that provide hardware resources (physical infrastructure) for a wide range of use cases, such as computing, storage, wireless networks, or data measurement. DePINs utilize the Web3 incentive model (i.e., token reward system) to incentivize the construction of physical infrastructure networks. As of May 2024, the total market capitalization of all DePIN tokens is $29 billion.

DePINs contributes to both digital and physical resource networks:

In the Physical Resource Network (PRN), contributors deploy location-specific hardware to provide (non-fungible) services. This includes:

• Wireless networks (such as Helium, World Mobile, XNET, Nodle)
• Sensor networks (such as Dimo, Hivemapper, Silencio, Onocoy)
• Energy networks (such as Starpower, PowerLedger, Arkreen)

In the Digital Resource Network (DRN), contributors guide hardware to provide (alternative) digital resources, where physical location is not the primary criterion. This includes:

• Calculation (e.g. ICP, Livepeer*, Akash Network, POKT Network*, Covalent*, Lit protocol*)
• Storage (e.g. Arweave*, FIL, Sia)
• Bandwidth and privacy (e.g. NYM, Hopr, Orchid, Mysterium, Fleek)
• AI (e.g. Bittensor, Fetch.ai, Modulus Labs*)

The early DePIN project generated a lot of initial interest due to its token framework design. For example, Helium rewards contributors with HNT tokens for helping to operate wireless networks through hotspots, while Filecoin allows users to rent out their excess storage space. Although this is enough to get many DePIN projects off the ground, token distribution may not be sufficient to ensure long-term participation of nodes in the network.

If running nodes becomes unprofitable, node operators will no longer have the incentive to operate the DePIN infrastructure. Therefore, the DePIN founding team must help node operators optimize costs.

DePIN Flywheel

The typical flywheel of the DePIN token economy is as follows:

• Establish the supply side of the service, such as storage or 5G antennas
• The inflationary token rewards incentivize nodes operators to provide the necessary infrastructure, although the demand is not yet sufficient to cover the costs
• As time goes by and the demand rises, monetizing network activities may increase the income of node operators, even if token rewards gradually decrease.
• Continuous currency network activities and increasing node operator revenue further incentivize supply to create the DePIN flywheel

The visual presentation of the DePIN flywheel is as follows:

As described in our analysis of the reward issuance schedule, the USD value (token price) of these token rewards is greatly influenced by overall market sentiment. Therefore, they may look like this:

Or depending on when you entered the bull market, it may be like this:

So, what is the relationship between reward issuance and cost?

As mentioned above, if the token rewards and income from user demand are not sufficient to achieve a balance of income and expenses, node operators may decide to stop supporting the network. A large portion of DePIN's operating expenses are paid in fiat currency, which makes the USD value of token rewards important and tied to the overall market performance. Despite any well-planned token issuance measures, the situation may turn out like this in the worst case scenario:

This will result in node operators exiting, leading to higher latency, lower reliability, and worse user experience. Ultimately, stagnating demand will shut down the flywheel.

The good news is that there are many ways to deal with this situation. One way is to make token issuance more flexible to align with the monetization of the network (see KPI-based issuance here). Another way is to address the cost issue to make the overall network more efficient, so it is less sensitive to token price drops. Our dynamic chart will be as follows:

Key proposition: If you know the cost of operating the DePIN network and its main driving factors, you can initiate governance discussions and research to reduce node operator costs before the reduction in network service supply.

Given the decentralization and permissionless nature of DePIN, assessing the cost base is not easy. Although token rewards and user demand revenue are usually tracked on-chain, other costs involved in running nodes are not publicly available (e.g., infrastructure expenses). This means we need to use assumptions and estimates about available data points.

In this article, we will address this challenge and introduce an estimation framework.

• Step 1: Network Contributor
• Step 2: Components of Cost
• Step 3: Evaluate the cost structure of network contributors

Framework

We propose the following framework as a methodology for managers of the DePIN network to assess the operational costs associated with operating infrastructure nodes.

By using this framework, the cost estimation of DePINs is decomposed into three steps:

• Identify Network Contributors
• Evaluation cost components (such as hardware, labor)
• Evaluate the above cost structure and summarize to obtain the overall cost estimate

Step 1: Identify Network Contributors

Although DePINs provides a variety of services (such as computing, network coverage, mobile data, etc.), the roles required to provide these services are the same (see an overview of the roles of DePIN suppliers in more than 30 networks here):

• Service Node / Producer: They provide services and the physical infrastructure required for them (such as servers, antennas, dashcams, etc.). For example, Filecoin's storage providers, Helium's hotspots, or Livepeer's transcoders.
• Validators/Observation Nodes/Phishers: They check the work completed by service nodes, directly or through the accounting layer. Then, the results of these checks are sent to the accounting layer. For example, Filecoin's storage providers (as they also validate storage proofs from other providers) and Helium's hotspots and Oracles (performing coverage proofs for other hotspots).
• Calculation Layer: Tracks the flow and status of the work/service provided and the corresponding payment. Please note that the protocol defines its own computation logic, such as how to track and store work and payments on the blockchain (we will discuss this in more detail in another article). For example, Livepeer's Arbitrum or POKT-chain on the POKT network (operated by POKT validator nodes).
• Gateway: They serve as coordinators/balancers in user, service nodes, and manage access or aggregate services (such as data in sensor networks), also related to the accounting layer. For example, Orchestrators in Livepeer or Gateways in the POKT network.
• Entruster: can participate in the service or observe the economy of the node through mortgage.

Roles related to the demand side (such as sales teams) are currently not common, and assessing the costs associated with governance, such as governance costs, is the subject of another article.

Please note that not every DePIN has delegation and gateway, nor does it require all roles to be separated. For example, the storage provider (SP) of Filecoin is classified as both a service node and a validator, and it also operates the Filecoin chain, thus forming an accounting layer. The same goes for Arweave miners.

Step 2: Evaluate the components of the cost

Each of the above roles can be executed through a Node, the cost of which can be one of the following four components (most have multiple):

• Hardware/Infrastructure: Costs related to actual physical infrastructure, such as dashcams
• Artificial: The cost related to the time spent on setting up and operating infrastructure
• Bandwidth, electricity, and other operational expenses: Costs related to data exchange and other operational costs, such as electricity and data center rent.
• Mortgage: The (opportunity) cost of not investing elsewhere

The last point refers to the cost of capital: it is almost impossible to obtain information about the debt/financing costs associated with these operations on a wide scale. However, there is a part of the content related to the cost of capital that we can evaluate: many DePINs follow the pattern of mortgaging to obtain access to working tokens and require node operators to mortgage some tokens to be allowed to contribute. Obtaining these tokens is an investment, and even if we assume that this amount can be recovered when leaving the network, holding these tokens has opportunity costs compared to investing capital elsewhere.

Our assessment of the cost components would be incomplete without considering the costs associated with accounting layer transactions. Assessing this is not straightforward and depends on several variable factors. In general, the network determines to what extent accounting is outsourced off-chain. However, there are three options for recording settlement layer and on-chain transactions:

• Proprietary L1: Running their own blockchain. For example, Arweave, Filecoin and POKT Network. Typically, service nodes and validator nodes also cover this role, which is why related costs are included (but if possible, we will try to separate them - see the example of POKT Network).
• Proprietary L2, better known as an application chain or application-specific Rollup: The cost of Rollup infrastructure (sequencers, etc.) and adjacent infrastructure (block explorers, wallet integrations, etc.) can usually be mapped to these four components. In cases where it is not clear, such as when using a Rollup-as-a-service provider (RaaS), it will be mapped to bandwidth and other costs.
• Public L1/L2: These outsource the settlement layer, meaning there are no hardware and labor costs for the network. However, the service Node, authentication, Node (and user/payer) pay directly (based on usage). There are some challenges in assessing the network-related costs of these transactions, and therefore some limitations: not all transactions are related to the accounting layer, such as swaps or other Decentralized Finance transactions, but it is often not easy to separate these transactions. We map these costs to bandwidth and other costs.

Bringing all these elements together to create a cost estimate is a challenging task. We not only need to estimate each cost component for each role in the network, as shown in the diagram below, but we also need to consider that not all node operators have the same cost structure. Determining the overall cost estimate is more complex than simply multiplying the number of network node operators by an estimate for one node operator.

Step 3: Evaluate Cost Structure

When we talk about cost structure, we are referring to the key differences that affect costs. These key differences make assumptions crucial. Of course, this is a trade-off: making assumptions simplifies the process, but may sacrifice accuracy. In other words, considering how many factors are involved, certain assumptions must be made to arrive at a viable theory.

There are three main factors to consider when evaluating cost structure:

• Differences in settings: A typical example is an operator using bare metal servers while another runs on the cloud (purchase vs. leasing). When we know the corresponding shares in the entire network, we can usually take these differences into account. This also involves capital costs in leasing or financing agreements. Assuming no capital costs, we suggest ignoring these differences.
• Another cost difference is related to the purchase time (storage becomes cheaper over time, and purchasing H100s may not). or the operating location. We recommend considering the time aspect by using the current price. For labor costs, location is important: DePIN can recruit contributors from around the world, and there are significant wage differences at the local level, making it difficult to evaluate the time invested in these jobs. Nevertheless, we made a simplified assumption that in our framework version, the hourly wages of all node operators are the same.
• Efficiency Differences: Node operators may have identical settings, but if one runs more of the same nodes, they may have lower costs per node due to economies of scale. In our framework, we first need to assess the node distribution of each node operator to address these impacts. Then, in order to understand and estimate the cost impact, surveys are needed with larger and smaller operators or other available data points (e.g., bulk discount promotions).
• Another example is long-time supporters of the network who progress faster on the learning curve and are therefore more operationally efficient, compared to those who are new to joining. Unless we have direct data points from the survey, we ignore this aspect.
• Differences in attribution and calculation: Although Node operators are equal on the first two points, they may view their contributions on a different cost basis, so the final cost will vary. For example, one person treats their involvement as part-time and does not track any time spent, while another treats it as a major business, paying wages based on time spent on projects. We account for this difference by providing a wider margin of error for the "part-timer" side (as they are often underestimated), but assuming the same time investment for each Node operation (see also economies of scale).

This is related to the benefits of the sharing economy, which is very common for DePIN: operators can use the same settings in multiple networks (and therefore also hardware, labor, bandwidth, power, etc. operating expenses), such as Livepeer with Ethereum and Filecoin operations, io.net with Render, Filecoin, and other GPU networks. For cases where hardware is crucial to operations, we do not consider cost savings related to the sharing economy. They are not only difficult to identify, but also difficult to quantify which network benefits the most in terms of costs, and how the savings are allocated. In terms of accounting, we will need to break down the total costs into monthly amounts. To simplify, we assume that we amortize the total amount over the same period throughout the lifecycle, and allocate the same amount to each month for all node operators.

Of course, there are more subtle differences, which we will explore in more detail in the DePIN repository.

This adds a third dimension to our 'execution plan', creating 60 different combinations to consider:

In general, although this formula is very comprehensive and provides various options for cost structures, the most useful is to apply it to multiple different time points rather than a static time point. The most powerful model is the one that links operational costs to network capacity. This allows us to understand the extent to which costs vary with changes in capacity or utilization. The network's capacity is related to the services provided by the network, such as the number of RPC requests in Pocket, the storage volume of Arweave or Filecoin, or the percentage of road network mapping in Hivemapper.

Please note that this formula requires a large amount of publicly available information, and we recommend obtaining this information through documents, forum/Discord posts provided on the internet, and if possible, through investigation.

Conclusion and Next Steps

As DePIN develops at an increasing rate, estimating the cost components of various DePIN is challenging. In addition to the known power law of hardware costs and changing capacity over time, estimating specific costs of cryptocurrencies, such as gas on the settlement layer and throughput capacity, is not a simple matter.

It is useful to understand the relationship between the current cost and the issuance of rewards and the income of the demand side, how the major cost drivers change with the assumption changes, and how the cost increases with the increase in demand.

In order to help guide governance decisions about the economic design of DePIN, cost estimation needs to be associated with reward issuance and revenue usage. While I plan to provide more examples of cost estimation for DePINs, I welcome feedback on the proposed framework, its assumptions and summaries, as well as potential improvements to the provided cost estimation.

Appendix - Example Illustration Framework

Livepeer

Livepeer provides a decentralized video infrastructure for real-time and on-demand streaming. Recently, Livepeer has started enabling idle GPU resources for AI model training use cases (see here for details).

Here is the process of gradually applying the framework. Most cost estimates are based on surveys with node operators (i.e. Orchestrators) conducted in the summer of 2023 and community information (such as here).

The total estimated cost of operating the Livepeer network is approximately $85,000 per month. A detailed breakdown of the average cost shows that hardware and labor account for roughly the same share (about 40%). If the uncertainty of the labor cost estimates described in the table is taken into account, the monthly cost of the network's 100 Orchestrators, their transcoders, and settlement costs on Arbitrum is approximately $40,000, at the lower end of the estimated range. It is worth noting that the monthly cost of $40,000 is not far from the current monthly revenue of 5-10 ETH (corresponding to an ETH price of $3,000-4,000). However, Orchestrators do not have a negative profit, as a larger part of their income actually comes from stake rewards.

It is worth noting that, as Livepeer's transactions are settled on Arbitrum, the cost of settlement layer is within the range of 0.5-2 ETH per month. Compared to the situation in the first quarter of 2022 before the Arbitrum migration, this saves over 95% of the cost. In addition, as of today, transactions on Livepeer have increased 2-3 times. Relatively speaking, the settlement layer now accounts for about 5% of the total cost, while before the migration (it accounted for about 80% of the total cost) it was a major cost driver.

Recently, it adjusted the algorithm that determines the work allocation method, placing more emphasis on the price per pixel provided by the Orchestrator. This puts downward pressure on transcoding prices and may help stimulate demand, but discussions on the forum indicate that price levels need to be further reduced. On the other hand, the recently launched AI-subnets may help add further monetization channels for the network.

One potential scenario in estimating spreadsheets is that tripling the demand for transcoding minutes would only increase overall costs by 20%. It is worth noting that bandwidth is the main driving factor for cost increases.

If we assume a similar price level (1ETH for $3,000), this should be enough to bring the network into the breakeven zone. However, if the transcoding price drops by 50%, the network-level fee income will be about $45,000 per month, which is below the lower limit of the cost estimate. The dynamics of costs and income on the Livepeer network will change as new use cases (such as AI video generation) emerge, thus increasing monetization opportunities. This remains to be observed.

POKT

At its core, the POKT network provides decentralized remote procedure call (RPC) endpoints. Recently, the POKT network announced its expansion into more use cases involving AI model inference. The framework for phased implementation is as follows. Most cost estimates are based on surveys conducted in the summer of 2023 with node operators, as well as subsequent interviews with these node operators and gateway operators.

Based on approximately 15,000 nodes and four gateway operators providing RPC endpoints, we estimate that the POKT network currently costs around $200,000 (+/- $80,000) per month to serve approximately 500 million relays per day. The largest portion currently is the service nodes (accounting for about 75% of the cost).

Since we can get historical data on the number of active nodes in the network, and there are data points with different cost components over time, we can put the network cost estimate on a timeline that shows three points in time when the larger cost cuts were addressed:

• After entering the bear market in mid-2022 and reducing token rewards, especially USD-based token rewards, node integration
• Improvements have been introduced within the network, such as Geomesh and LeanPOKT, which significantly reduce operational costs and personal improvements for node operators' configurations.
• By adding simpler gateway settings, the decentralized gateway role reduces bandwidth costs.

As our cost framework ties cost estimation to network capacity and demand, we can assess changes in cost structure. For example, if the demand increases from the current 500 million per day to, for example, 2.5 billion relays per day, then the gateway will account for 60% of the total cost base, approximately $400,000 per month (currently around $200,000). Please note that this is double the cost, while the demand growth is fivefold. This is because the service nodes can improve their setup and meet the growing demand on a essentially the same cost basis.

If we further assume that the share of new gateways operating on a lower-cost basis in the total number of relay services increases to, for example, 50% (currently 30%), then the overall network cost will be $300,000 per month.

With the decentralization of the gateway, gateway operators can define their own price points. Assuming an average price of $4 per million requests, the overall scenario of the POKT network would earn $300,000 per month, thus achieving a basic break-even point.

Dfinity/ICP

Dfinity/Internet Computer Protocol (ICP) is designed to be the "blockchain of blockchains," providing computational resources for executing smart contracts (called canisters), which are organized in subnets (details). The backbone is to provide storage, computation, and bandwidth to replicate all canisters, states, and the node machines of their subnet calculations.

The progressive application framework is as shown here. Most cost estimates are based on data from documents and forum posts.

ICP is one of the few networks that incorporate the cost based on fiat currency into the token reward mechanism, which makes cost evaluation easier. Currently, there are about 85 operators running approximately 1400 node machines. We don't have data points for larger operators, so our overall estimate range is quite broad: the monthly cost of operating the ICP network is approximately $400,000 to $900,000, with an average of around $600,000.

Although a proper income assessment is worthy of a separate article, we estimate the current monthly income to be around $25,000. Compared to the estimated costs, this seems low, but it is due to low utilization: with only 559 active node machines, we estimate the current demand (expressed in cycle burn rate) to be about 2% of the total capacity. This means the network can withstand, for example, a 25-fold increase in demand without increasing the current cost base. A forum post actually estimates that demand in the next two years will reach 15-25 times, which would then (under the same conditions) lead to ICP earning these fees per month.

DIMO

DIMO is a decentralized network that empowers drivers to manage their vehicle data. At the same time, DIMO enables businesses and developers to build innovative mobility-related applications (and profit from them). Data measurement is done through special devices (Autopi, Macaron) or applications. While the above DePIN example is a digital resource network, DIMO is the first physical resource network included in this analysis.

The framework for gradual application is as follows. Most cost estimates are based on online (device) price information, Dune data, and forum posts.

For the settlement layer, we assume that half of the average cost of connecting a car in Q1 2024, ranging from $0.6 to $1.5, can be attributed to the operation of DIMO. For the gateway, we assume a monthly hardware cost of about $4,000, and labor costs related to the above operation are about $11,000 per month. Overall, this adds up to about $180,000 in monthly expenses, as shown in the table below. Most of the costs are related to bandwidth and other costs, about 1/3 of which are related to settlement costs on Polygon, and the remaining 2/3 are related to the monthly cost share of smart car integration.

We have no clues about the actual network revenue, but the estimate of the global automotive data market and related automotive data revenue shows that the current revenue per car is about $150 to $185, and may rise to $500 to $600 by 2030. If DIMO can obtain 10-15% of the revenue, the generated revenue range will be $110,000 to $180,000 per month, thus covering operating costs.

However, data monetization itself does not seem to be the actual goal of the protocol; instead, DIMO focuses on providing infrastructure for building applications on top of the network (), as reflected in the recent discussions about DIMO nodes and token upgrades. The changes discussed may impact the aforementioned cost structure.

Special thanks to my contributors: Mihai (Messari), Raullen (IoTeX), Nodies Team, Grove Team, Pocket Network Foundation, DIMO team, Diana Biggs and Christopher Heymann for their feedback and contributions.

The standard project is a 1kx investment portfolio.