banksyweb3
Posts
12 Mental Models for DePINs

12 Mental Models for DePINs

December 19, 2024

This blog will cover 12 different mental models for DePIN projects:

DePIN Classification & Decision Tree
Consumer DePIN - 1st & 3rd Party Applications
Virtual to Physical & Active to Passive
Custom Hardware Devices & Commodity Hardware Devices
No Manufacturing - Outsourced Manufacturing - Own Manufacturing
Ease of Installation
Burn & Mint Tokenomics vs No burning
Classified as a Multiple of Revenue
Location Dependant & Independent
Not a DePIN
Physical & Digital - A Different Term for DePINs
Decentralised & Not Decentralised

DePIN Classification & Decision Tree

The term DePIN was coined after a 2022 Messari poll on Twitter (X), with DePIN receiving 31.6% of only 136 votes [5]. The academic research paper “Are you a DePIN? A Decision Tree to Classify Decentralized Physical Infrastructure Networks” provides a framework for assessing whether a project is in fact a DePIN, see Figure 1. The criteria for a DePIN is:

A three-sided market - has an interface for the supply side, demand side and a customer interface.
Token-based incentives for supply - uses cryptocurrencies to incentivise users to contribute to the network supply.
Physical asset placement - the service has physical components.

DePIN Decision Tree [4]

Perhaps Web3 is not about the 3rd generation of the internet, in fact, it is about the 3 sided markets, like DePINs, which have permissionless participation of the supply, service and demand side.

Consumer DePINs - 1st & 3rd Party Applications

Amanda Young’s Growing Consumer DePIN, provides a fresh framework for viewing consumer DePINs [7]. The graphic below shows which projects have consumer or enterprise powered supply and demand. The article covers the development of consumer marketplaces, with Teleport and direct to consumer products like Helium Mobile. Consumer DePINs have resulted in new behaviour creation, with Grass’ browser extension creating new demand for data for AI. Whilst existing environmental behaviour is accelerated with green energy projects like Daylight. The result of consumer DePINs is in applications:

First Party Consumer Applications - DePINs can sell directly to their consumer, like Roam’s eSIM.
Third Party Consumer Applications - DePINs can create APIs and SDKs for their projects, like Geodnet’s mobile location SDK, to be used in Animoca Brands’ gaming and with future Solana Phones [9,10].

Enterprise vs Consumer Driven DePIN Projects [7]

This follows a growing trend of what Tom Trowbridge calls DePIN Institutionalisation. Whereby DePIN supply side is being supplied by enterprises and not consumers. I believe institutional adoption of DePINs is twofold, both in terms of supply provision and demand generation.

Depin Institutionalisation [8].

As a result of this, Amanda’s chart could be modified slightly by adding arrows down the supply axis, where projects like Hivemapper are seeing supply shift from individuals towards fleets.

Enterprise Supply Shift - Josh's Updated Consumer DePIN Chart

Virtual to Physical & Active to Passive

Decentralised Co’s framework for viewing DePINs is across two axis [11]:

Virtual to Physical reflects the varied barriers to entry and investment in infrastructure, from a free Grass browser extension to thousands of pounds worth of GPUs for AI training on IO net.
Active to Passive reflects the time investment and human involvement in the day to day operation, from a set and forget Grass browser extension to mapping new areas with your Hivemapper.

Decentralised Co DePIN Framework [11]

As the graph shows, the framework is heavily skewed to passive participation in the network. This is for obvious reasons, in that the opportunity cost threat and time investment is lower, enabling easier onboarding and supply side scaling. Decentralised Co expect more value to accrue in the active part of the graph, despite its challenges.

Mason Nystrom discusses the same idea. Adding that actively supplied projects are “stickier”, because the suppliers can chose their marketplace freely [12]. They also have more resistance once they scale and generate network effects to spin the DePIN flywheeel. Actively supplied, customised hardware are the hardest to onboard, as they are single use for one project, with a (3rd generation) Hivemapper dashcam costing $500. Whilst passive projects tend to be less sticky, because of the sunk cost of fixed asset investments. With free, passive apps on commodity hardware, they can be more difficult to secure long term, due to the nature of other ‘free app’ competitors. A project like Filecoin can leverage their token incentives to be patient when building up demand, once the supply side is ready. Whereas an active project like Nosh will have to generate demand as supply builds, or participants will leave the network. To mitigate this risk, a “permissioned manner” is advised to link supply build out with a fixed yield in line with demand generation. Otherwise, DePINs will end up with the same issue they are trying to solve, where VCs have to spend billions building out projects like Uber, as covered in my previous Hivemapper blog post here.

Different DePIN Classifications [12]

The article also discusses human capital and labour compared to resources. With Braintrust and Nosh being focussed on the output of humans, adding in more subjective quality and reliability concerns compared to machine resources. The subjective quality makes it difficult to scale, especially without long-term reviews of staff, compared to chips where you know the quantitative design spec of a A100 or RTX 4090. The reliability concerns of DePINs is also present in AI modelling. One solution is provided by Reppo, who are a DePIN support project, with a Modular Reputation Framework for Machine Learning (ML) models, shown below. Also, Space and Time provide Proof of SQL, SXT’s ZK coprocessor.

Reppo’s Modular Reputation Framework [13]

Custom Hardware Devices & Commodity Hardware Devices

DePIN hardware devices are either:

Commodity Hardware
Custom Hardware

As a result, the design and strategy considerations vary greatly, as covered by Blockworks [16]. The different approaches are seen in road mapping by Natix, who enable users to map using their phone and Hivemapper who require a $500 payment for their (3rd generation) dashcam. For Natix's supply side the only barrier to entry is a mobile phone, which most crypto natives are fortunate enough to have. Therefore, its easier to onboard users to build out the supply. But, there are concerns around stickiness and community loyalty when another free app comes along. Whereas with Hivemapper, users have to commit to investing a large sume of money and will then have sunk cost bias and remain loyal. It is likely that after only a couple of years Hivemapper's 1st and 2nd generation devices are nearly redundant. This is not only wasteful, but also runs the risk of poor community engagement (without fair token rewards). The blockworks article is similar to Mason Nystrom’s mental model of fungible (GPU for Akash) and non-fungible hardware (Hivemapper Dashcam) [6].

No Manufacturing - Outsourced Manufacturing - Own Manufacturing

Another DePIN thought piece was written on the Multicoin Capital blog, by Shayon Sengupta and Tushar Jain [14]. The blog covers the different hardware design strategies of DePIN projects, whether that is sticking to software, or outsourcing your design to (approved) manufacturers, or owning the full processes. A key factor in the scalability of Bitcoin was that there were no standards or guidelines on miner devices, enabling a free market for innovation and diversification. Despite this, the majority of modern Bitcoin mining is done on ASICs. On balance, Helium have 11 approved manufacturers and scaled effectively as a DePIN [2]. Despite limiting scalability, Dimo enabled full quality control over their product (and margins).

Hardware Strategy of DePIN projects [14]

Ease of Installation

The same Multicoin blog covers the ease of installation of DePIN devices. Of course, it is easier to build out the supply side when the installation process for users is simple. This becomes more pressing as we approach “mass adoption” and non-crypto natives get involved.

Ease of installation of DePIN projects [14].

Burn & Mint Tokenomics vs No burning

The academic research paper “Burn-and-Mint Tokenomics: Deflation and Strategic Incentives” provides a game theoretical and mathematical analysis of Burn and Mint Tokenomics, used by DePIN projects such as Helium (first developed by Factom) [3]. When the product is used and revenue is generated, the tokens are burned, to reduce supply and increase token value in the network. When users contribute to the network, the tokens are minted, to increase supply and reduce token value in the network. This is done to achieve deflation and enables token price appreciation as a key incentive for network supply and continued security.

Given the global distribution of DePIN devices, the different interest rates offered by countries within the DePIN network can vary drastically, impacting the net present value and discount rate of returns. Fiat and Stablecoins are a solution for this, to ensure stability of the Burn and Mint tokenomics modelling and ecosystem stability. Proposition 3 in the paper covers economic collapse with some maths beyond the scope of this blog and figures 1-4 shows that only in the case of fiat-fixed rewards are both network value and token rewards sustained. However, sufficient deflation is not sufficient on its own to ensure a stable economy because, even sufficient deflation, the reward amount may converge to zero faster than the price can appreciate. The solution of providing liquidity to the economy would only cause hyperinflation.

Figure 1 to 4 showing fiat fixed rewards [3]

“How-ever, the choice of the ideal deflationary factor depends on many factors, most of which are unknowable to designers.”

Another difficulty of Burn and Mint, is that projects are innovating as first movers, so the factors determining burn and mint design include some known unknowns and unknown unknowns. Therefore it is difficult to design them perfectly. Particularly when each DePIN project can be very different to those in their shared vertical, never mind other verticals within DePIN. It’s not as easy as cloning a contract. Space and Tokens allows teams to simulate tokenomics and with Space and Time’s AI chatbot for dApp creation launching in 2025, users will have more tools to play around with tokenomics game theory. I would be interested to see how investors are tracking the burn and mint tokenomics to see which DePIN projects are at risk of monetary or economic collapse.

Similarly, Geodnet and XNET are using a buy and burn strategy to use 5% and 80% of revenue to buy back tokens, then burn them. As a result of Burn and Mints challenges, many DePIN projects do not use a Burn and Mint tokenomics design.

Classified as a Multiple of Revenue

Another way to view DePINs is in the context of revenue, by taking their market cap and what multiple of their revenue it is. As you can see from the table below, the Revenue/MCap varies from 8.4x (Glow) to 6,862x (The Graph) [15].

DePIN Protocol Revenue Leaders [15]

Location Dependant & Independent

The paper also classifies DePINs as Location Dependant (LD) and Location Independent (LI), shown in Table 1 below. Where LD DePINs are not interchangeable, Hivemapper’s mapping rewards are based on where the supplier is located. However, with LI DePINs, they are interchangeable if one goes offline, as Filecoin storage treats geographical location pretty much equally.

Table 1 DePIN System Classification [3]

Not a DePIN

DePINs are active in every sense (pun intended), so it is no surprise there are many people who don’t agree fully with the term. Another way to think about DePINs is by not thinking about them as DePINs. Leading DePIN blockchains like IoTeX and Peaq are not DePINs, as covered in the Decision Tree paper “Despite these DePIN-centric extras, like Bitcoin, these blockchains can be considered a two-sided market where those contributing the supply make up the platform that is given as a service to Web3 customers, thus not meeting the first criteria.“ [4].

Physical & Digital - A Different Term for DePINs

With sub-sectors such as Physical Resource Networks (PRN) and Digital Resources Networks (DRN), Michael Andrew and Mark Ballandies argue the P in DePIN is redundant. Mason Nystrom outlined a new framework of Decentralised and Token Incentivised Infrastructure Systems (DeTIS), split into Decentralised Resource Networks (DeRen) and DePINs [6].

August 2024 DePIN Sector Map by Messari

The most recent DePIN sector map has used the split of PRNs and DRNs. With new categories being added for:

Decentralized Gaming Infrastructure (DeGIN)
Compute Network
AI Data Layer
Bandwidth Network
Robotics
Mobility Network
Manufacturing
Mobility Network

DePIN Sector Map [15]

Decentralised & Not Decentralised

“This balance between decentralization and coordination remains an important issue” - Simplicity Group [2]

It is not yet the time for widespread analysis and debate on how decentralised DePINs are, especially considering that many ecosystems require years to become more and more decentralised. However, Vitalik’s blog on “The Meaning of Decentralization”, provides an important framework to start with [1]:

Architectural decentralisation - The physical distribution of the nodes and devices in a network.
Political decentralisation - The democratic governance and control of the network.
Logical decentralisation - The functioning of the entity cohesively.

Many thanks,