Decentralized Infrastructure
Disclaimer: This is not financial advice. Anything stated in this article is for informational purposes only, and should not be relied upon as a basis for investment decisions. Chris Keshian may maintain positions in any of the assets or projects discussed on this website.
To subscribe to my mailing list, input your email here.
Satoshi created bitcoin to disintermediate financial institutions. Bitcoin is a peer-to-peer electronic currency that “allows any two willing parties to transact directly with each other without the need for a trusted third party” (bitcoin whitepaper). This initial white paper gave rise to a new paradigm of decentralized, blockchain-based applications, such as peer-to-peer lending protocols (AAVE, COMP) and decentralized exchanges (UNI, SUSHI).
To date, the majority of these products have been built “on-chain” meaning computation and settlement happen directly on the blockchain. While blockchains do convey certain benefits - most notably decentralization and distributed consensus - they also have inherent shortcomings. Blockchain networks are distributed public ledgers that are designed to grow perpetually as transactions are validated and blocks are mined. This is fine when the purpose of a blockchain is simple, such as the Bitcoin blockchain, which is designed to keep an updated state of ownership by tracking which public addresses own which bitcoin. However, when we attempt to build fully functional decentralized applications, with an entirely decentralized tech stack, relying solely on a costly and slow distributed ledger is not sufficient.
Take data storage, for example. The problem of data storage capacity becomes significant when the size of the blockchain becomes very large. Full blockchain nodes that are required to store the entire copy of the blockchain ledger will have a difficult time keeping their available data storage capacities on pace with the growth of the blockchain network. This problem is compounded when network effects are taken into consideration, where the network grows exponentially as more users are onboarded and on-chain data storage requirements increase rapidly. Therefore, storing media like digital images on-chain is prohibitively expensive.
If web 3 is to realize the vision of fully disintermediated applications and services, we must have more robust decentralized infrastructure to service these applications. A fully decentralized compute network can be broken down into three general components:
Applications - The software that submits API requests (e.g., queries, relays) meant to be routed to any public or encrypted database node.
Nodes - The decentralized servers that provide functions such as storing database indices, sending session information to an application, servicing API requests submitted by applications, and storing network states such as account balances and work reports.
Network Layer - The ecosystem that maintains the operation of the decentralized protocol, including governance, protocol rules, the actors involved, and the economic games they participate in.
Fortunately, there are talented entrepreneurs working on companies that plan to fully decentralize the entire web 3 tech stack.
In this article, I will briefly discuss seven web 3 infrastructure companies, on which I believe the future of decentralized applications will be built. Each of these projects has an associated token that accrues value relative to the growth and use of these services. As this bear market progresses, I plan to purchase a basket of these assets at depressed prices, and hold them through the next bull market. As I discussed here, I think this will likely be at least a two year holding period.
Since crypto is just early stage investing with liquidity, I view these investments through venture capital lens - expecting some to fail, but those that succeed to return multiples on my initial investment. As such, I believe a diversified basket of web 3 infrastructure companies will perform well in aggregate over a two to three year time period.
Below I will briefly discuss:
Filecoin
Arweave
Ceramic
The Graph
Render Network
Pocket Network
Chainlink
Most of these services are middleware, as shown in the image below, and serve as the connective tissue for the application layer and blockchain layer of the stack (source: Delphi Digital).
Filecoin ($FIL)
Filecoin is a decentralized storage network. The best analogy for this service is “AirBnB” for file storage. Filecoin is meant to serve as a decentralized alternative to the large, centralized cloud incumbents like AWS, Google Cloud, and Microsoft Azure.
Filecoin is built on IPFS (interplanetary file sharing system) which is a novel file sharing protocol based on the content (hashed) vs the geo-location (i.e. HTML). Filecoin is the complementary incentive layer on top of IPFS, which facilitates open markets for storing and retrieving files. The Filecoin blockchain allows parties to reach “deals” - basically SLAs - with payment terms and duration, where providers must consistently provide “proof” (Proof of replication & Proof of Space Time) of storage to network where they are compensated by transaction fees and “block rewards”.
Unlike centralized storage services, the storage cost is determined by free-market dynamics in an open market instead of a fixed pricing structure. To incentivize storage providers to participate in storage deals, the Filecoin network rewards storage providers with the network’s native token ($FIL). Storage providers earn block rewards for committing storage for future deals on top of payments for storage and retrieval. Below can see a visiual of how these participants interact.
Filecoin is the most prominent entity in this emerging market in terms of storage usage and capacity. It has been described as a “P2P version of AWS,” perhaps rightly so as it has accumulated more than eightfold the storage usage of its second-leading competitor Storj. Furthermore, it vastly exceeds its competitors in terms of market cap, as shown in the chart below (source: The Block).
The data storage market is massive. According to Fortune business insights, the global cloud storage market was ~$70b in 2021, and projected to grow to ~$375b by 2029 with a CAGR of 24% over the forecast period, accelerating from historical ~13% CAGR. The market is dominated by web 2 giants – Azure, AWS, GC – and quite concentrated with the top five players holding 77% of the market today.
Data growth (and storage demand) is likely to continue exponentially as we move more of our lives into the digital realm. An “AirBnB for data storage” which efficiently matches growing demand with “excess” supply could be a very valuable marketplace – servicing high growth areas like AI and IoT, which both require massive volumes of structured and unstructured data. At scale, the cost for decentralized solutions should undercut those of centralized providers. Moreover, Filecoin should be able to support any use case its centralized counterparts can, while offering some unique features such as immutability, permanent storage, and a public transactional ledger.
Arweave ($AR)
Similar to Filecoin, Arweave is a decentralized storage solution. Arweave is trying to solve the problem of reliable long-term data storage. The protocol matches people who have available hard drive space with individuals and organizations that need to store data or host content permanently.
From the user’s perspective, the main difference between Arweave and Filecoin is that Arweave allows users to store data permanently, sustainably, with a single upfront fee, while Filecoin has a subscription model. Arweave uses a fee structure where users pay once to store information forever instead of being bound to monthly or yearly subscription fees. There are some technical differences between Arweave and Filecoin as well, but the nuances fall outside of the scope of this article.
One of Arweave’s primary use cases is storing NFT metadata and hypermedia. NFT metadata refers to a JSON document that includes descriptive information about an NFT, such as its name, traits, a link to the associated hypermedia, etc. NFT hypermedia refers to the graphics, audio, and video representing the digital art, profile pictures, collectibles, music, and so forth that people are paying to own when they buy an NFT.
As discussed at the beginning of this article, Ethereum and other blockchains popular in NFT development are not designed for storing large data files like NFT metadata and hypermedia. It is estimated that storing just 1 GB of data on the Ethereum blockchain would cost roughly $20M. Given these costs, storing even the JSON file containing metadata is prohibitively expensive. What can be stored on these Layer 1 blockchains is a URI string pointing to the location where the user can find the NFT’s JSON metadata.
The success of these storage protocols depend on the applications built on top of them. So far, NFT and web 3 storage are the prominent use cases, but emerging new applications in areas ranging from audio/video to gaming to computation and more are making the case for global decentralized storage networks accumulating more of the cloud storage market.
Ceramic (no token yet, expected in early 2023)
The Ceramic Network is a decentralized network that supports mutable and composable data streams across an application ecosystem. The network infrastructure is centered around user data streams and decentralized identification, which allow users to be in full control of their own data while allowing different applications to build features from them.
Ceramic's public infrastructure allows participants to create signed, versioned, tamper-proof documents that act as a censorship-resistant and a universally available source of truth for important information. Because Ceramic's Universal Document Graph is public, permissionless, and verifiable, it unlocks information access and interoperability amongst all platforms and services across the web. This is a critical requirement for building interoperable systems that need to operate outside the bounds of any single wallet, application, database, service, or network silo. Ceramic allows developers to decentralize application databases and make data universally composable and reusable across web 3 apps. That means data models can be shared endlessly across different apps as long as they’re supported. For example, documents, user data, metadata, tables, video game items or anything that developers may want to share can be exchanged peer-to-peer without the need for a centralized database.
Ceramic Network is essentially a network of data streams that are maintained by individual user accounts who own the data and are verified by node operators whenever there are state changes along the data stream. A collection of data streams from different user accounts can be used to develop specific features and apps, such as building a user profile or social graph.
The Ceramic network currently operates atop 3 web 3 technologies as its main data stack: a static content address network like IPFS, a decentralized persistent storage layer like Arweave, and a blockchain like Ethereum.
The main difference between Ceramic and Filecoin/Arweave is that the latter provide only static documents. Filecoin and Arweave perform well for storing static files, but lack the computation and state management capacity for more advanced database-like features - such as mutability, version control, access control, and programmable logic. These more advanced features are required to enable developers to build fully-featured, decentralized applications, comparable to what we see in web 2 social applications. Ceramic aims to take Filecoin and Arweave a step further by providing infrastructure for mutable and composable data that is based on the static and immutable data stored in robust decentralized storage networks.
The Graph ($GRT)
The Graph is a decentralized protocol for indexing and querying data from blockchains. It makes it possible to query data that is difficult to query directly. Projects with complex smart contracts like Uniswap and NFT initiatives like Bored Ape Yacht Club store data on the Ethereum blockchain, making it difficult to read anything other than basic data directly from the blockchain.
In the case of Bored Ape Yacht Club, we can perform basic read operations on the contract like getting the owner of a certain Ape, getting the content URI of an Ape based on their ID, or the total supply, as these read operations are programmed directly into the smart contract; however, more advanced real-world queries and operations like aggregation, search, relationships, and non-trivial filtering are not possible. For example, if we wanted to query for apes that are owned by a certain address, and filter by one of its characteristics, we would not be able to get that information by interacting directly with the contract itself.
To get this data, we would have to process every single transfer event ever emitted, read the metadata from IPFS using the Token ID and IPFS hash, and then aggregate it. Even for these relatively simple questions, it would take hours or even days for a decentralized application running in a browser to get an answer. We could also build our own server, process the transactions there, save them to a database, and build an API endpoint on top of it all in order to query the data. However, this option is resource intensive, needs maintenance, presents a single point of failure, and breaks important security properties required for decentralization.
As demonstrated by the above example, indexing blockchain data is hard. Blockchain properties like finality and chain reorganizations complicate this process further, and make it not just time consuming but conceptually hard to retrieve correct query results from blockchain data.
The Graph solves these problems with a decentralized protocol that indexes and enables performant and efficient querying of blockchain data. These APIs (indexed "subgraphs") can then be queried with a standard GraphQL API. Today, there is a hosted service (Edge and Node) as well as a decentralized protocol with the same capabilities. Both are backed by the open source implementation of Graph Node.
To ensure the protocol runs correctly and efficiently, The Graph Network incentivizes several key roles within its ecosystem of participants:
Indexers operate nodes on The Graph Network (“Graph Nodes”) that index data from subgraphs.
Curators analyze subgraphs for quality, signaling which subgraphs are valuable to index.
Delegators stake The Graph’s native token GRT to indexers, earning a portion of query fees and indexer rewards without running a Graph Node themselves.
As of December 2022, only Ethereum is supported by The Graph’s decentralized protocol, but more than 30 networks are currently supported by The Graph’s centralized hosted service. The Graph plans to decentralize these 30 networks early next year, which is bullish for the protocol and the token, as increased use will instigate natural demand (i.e. buy pressure for the asset).
Render Token ($RNDR)
The Render network is a peer-to-peer distributed GPU computing platform that democratizes computing and rendering power. Render Network allows for unlimited scaling across parallel GPU nodes, enabling high speed and efficient rendering with end-to-end encryption.
The network connects users who want to perform rendering jobs with systems that have available compute power and are not in use. This allows the network to leverage idle GPUs and creates a mechanism to allow GPU computing resources to be allocated more efficiently. The use of a blockchain and the Render Network’s token, RNDR, creates the incentive system needed to bring GPU node operators to the network and to easily coordinate the economic activity required to form a global, distributed rendering network. This model also allows individual creators or large render farms to earn RNDR tokens when their GPUs are not in use and to apply the RNDR earned to their own render jobs on the network.
Render is part of OTOY inc, a leader in rendering software whose customers include all major movie studies (Disney, HBO, Warner Brothers, etc.), along with tens of thousands of digital artists including Beeple, Pak, and FVKRNDR. Octane is the most widely used 3D rendering software in the world and it’s estimated that ~85% of new 3D artists learn Octane as their primary software. The title sequence of Westworld, Lil Nas X’s music video for MONTERO, and the collections at the Frieze Viewing Room in Los Angeles were all rendered with Octane. Apple features Octane above the fold on the Metal homepage, and Apple’s new Octane X is built using Render.
Commercial projects have already begun leveraging the Render Network including FIFA FIFPRO’s World 11 player reveal, deadmau5’s music video, Robinhood’s crypto ad, and Apple’s event for the M1, featuring Jules Urbach. The Render Network produced physically accurate renderings of the Gene Roddenberry Star Trek archives, Beeple’s Everydays, and the Star Trek scenes (29:19 - 29:38), as seen in Apple’s October 2021 keynote, exponentially faster and more cost effectively than using incumbent cloud providers.
The Metaverse, movies, games, simulations (in medical, manufacturing, etc) and animations require immense computation to render trillions of digital objects in high fidelity. The amount of rendering needed in these industries has been exponentially increasing with the push to higher frame rates, higher resolution, and more realistic images. For example, Pixar’s Inside Out took 33 hours to render just one frame.
Today Render is focused on graphics rendering but has plans to support AI and physics simulations, which also require GPUs. The number of industries adopting virtual products and services is increasing exponentially while the supply of accessible compute resources has remained limited. These industries are all competing over limited GPU resources and typically rely on centralized cloud providers since building a local render farm is too difficult and costly.
The Render Network capitalizes on a multi-billion-dollar arbitrage. By unlocking hundreds of millions of otherwise latent consumer grade GPUs, the Render Network can dramatically reduce the cost of GPU-based computation, increase total number of renderings, and decrease the time required to complete rendering jobs. Render is well positioned to capture growing demand for GPUs, with the leading cost structure vs centralized cloud-based GPU offerings, integrations with the leading rendering software packages, and the largest pool of GPUs.
Pocket Network ($POKT)
Last year Moxie Marlinspike, the founder of the popular messaging app Signal, wrote a critique of web 3. One of his core criticisms is the lack of attention to the client/server interface in web 3.
“This was surprising to me. So much work, energy, and time has gone into creating a trustless distributed consensus mechanism, but virtually all clients that wish to access it do so by simply trusting the outputs from these two companies without any further verification.” — Moxie
In the modern Internet Age, the majority of online activity is filtered through a small cohort of centralized platforms that abstract away the complexities of publishing and consuming web-based content. Despite the decentralized ethos the internet was built upon, we’ve witnessed a consolidation of power toward centralized platforms.
In the context of the current web 3 infrastructure stack, Moxie likens users and Dapps to clients and blockchains to servers. Drawing parallels to the rise of web 2 platforms, he asserts that it doesn’t matter how trustless blockchains are, users still don’t want to run their own nodes. Therefore, client-server interactions are bound to be facilitated by centralized platforms that abstract away the complexities involved with blockchain connectivity. See a visual of this architecture below (source: Delphi).
Currently, there are centralized and decentralized operations that streamline blockchain data access. For example, blockchain infrastructure providers like Infura and Alchemy provide portals to blockchain data, but they are developed, owned, and operated by centralized entities.
Pocket Network is a decentralized RPC node provider. The project tackles the data dependence of dApps on centralized infrastructure providers and delivers a resilient and distributed solution to cater to their data requirement needs. A good analogy for Pocket is “the TCP/IP of web 3 node infrastructure”.
Aside from maintaining a fully decentralized stack, another advantage to using Pocket Network is that, for developers, it transforms their infrastructure costs from an Operating Expense to a Capital Expense. This is done because, in order to use the network, developers pay an upfront cost to buy POKT, rather than a monthly recurring cost to use Infura or Alchemy.
As the space matures, and more participants realize the third party security holes created by centralized RPC providers, I expect decentralized solutions like Pocket Network to increase in use.
Chainlink ($LINK)
Chainlink is a decentralized oracle network that aims to serve as a middleware between smart contracts on blockchains and external data sources, allowing smart contracts to securely access off-chain data feeds.
Chainlink greatly expands the capabilities of smart contracts by enabling access to real-world data and off-chain computation while maintaining the security and reliability guarantees inherent to blockchain technology. Chainlink smart contracts check data provider reputation, match data request orders with data providers, and then aggregate data from multiple sources to ensure accuracy.
Chainlink was the first decentralized oracle network that allowed smart contracts to securely connect to external (off-chain) data. To oversimplify the workflow, Chainlink allows someone purchasing its services to specify (1) a query and (2) the number of oracles needed to fulfill that query, among other details. This is all written into a service level agreement (SLA), and an order-matching process then selects the oracles required to execute the services written into the SLA (note that the node operators retrieving the data may not necessarily be the data providers themselves.)
The process is more complicated than that, as Chainlink also checks the track record of oracle providers and aggregates information in such a way to ensure validated data. Moreover, the services provided by Chainlink are expanding, currently including but not limited to:
data feeds (such as pricing data for digital assets) which are regularly updated and make up the bulk of the requests,
offchain API / data sources,
verifiable randomness functions (VRF) used for random number generation in gaming, betting markets and dynamic NFTs (version 2 was launched on February 16),
proof of reserves used to verify stablecoin and cross-chain (i.e. wrapped tokens) holdings off chain,
a Keeper network that helps dapp developers automate smart contract transactions that require off-chain input
Oracle networks like Chainlink are critical as they allow decentralized applications and blockchains to take reliable inputs from the off-chain world. I expect these networks to grow alongside the wider crypto market over the next 5-10 years.
Conclusion
All cryptoassets tend to correlate to 1 when there is a market-wide selloff. As such, protocols with real traction and revenue sell off as much as worthless meme coins. Given the current macro climate and the aggressive nature of the crypto cycle, this particular sell off has been extreme, with most of the above-mentioned assets down 70-90% this year. Over the next ~6 months, I am preparing to purchase a basket of these assets at prices that I think are depressed as a result of current market dynamics. I am bullish on the fully decentralized tech stack, and think this theme will be important as this industry matures.
