This episode explores how AI's insatiable demand for compute is breathing new life into the DePIN sector, forcing a re-evaluation of tokenomics, user experience, and the very nature of value in a world increasingly shaped by intelligent agents.
Dmitriy Berenzon's Background and Focus
- Dmitriy Berenzon, a partner at Archetype, brings extensive experience in crypto investing since 2018, having worked with Coin Fund, Ballinger Investment Group, and 1KX.
- His current work at Archetype, since January of the previous year, involves investment due diligence, research, and portfolio support.
- Dmitriy's areas of expertise include DeFi (Decentralized Finance), payments, DePIN (Decentralized Physical Infrastructure Networks)—networks that use cryptocurrency tokens to incentivize the build-out of real-world physical infrastructure—interoperability, and scalability. He assists teams with go-to-market strategies, token design, and follow-on fundraising.
The State of DePIN: Successes and Hurdles
- Dmitriy Berenzon argues that while some DePIN projects like Helium (reporting 700,000 subscribers and 23 terabytes of data offloaded) and Geodnet (generating millions in revenue) have found success, it has been a gradual process.
- A key challenge has been achieving parity with centralized services on the demand side. This involves providing basic SLAs (Service Level Agreements)—formal commitments to a certain level of service quality—abstracting away blockchain complexities, and effectively communicating value beyond mere decentralization.
- Dmitriy notes, "nobody buys decentralization for the sake of decentralization. So they buy lower costs, they buy resilience."
- Another significant issue is the disconnect between private investor interest in the equity of DePIN companies and the value accrual reflected in their publicly traded tokens, a gap that liquid markets are keen to see bridged.
DePIN Tokenomics: The Quest for Value Accrual
- A critical factor for DePIN's evolution is refining tokenomics to ensure genuine value accrual to the token, which Dmitriy Berenzon highlights as often lacking in first-generation projects.
- He points to the complex interplay between equity structures (OpCos/DevCos receiving fiat) and protocol tokens, emphasizing the need for a "credible commitment to value accrual in the token." Regulatory uncertainty further complicates this.
- Dmitriy observes a shift where "ARR (Annual Recurring Revenue) is the new TVL (Total Value Locked)," indicating a focus on real revenue. While token incentives effectively bootstrap supply, the focus must now be on building strong demand-side products that solve "shark bites, not mosquito bites" for users.
- Strategic Implication: Investors should scrutinize DePIN projects for clear mechanisms that translate operational revenue into token value, avoiding structures where new equity raises dilute token holder exposure. A one-to-one flow-through between OpCo revenue and the token network is ideal.
Addressing DePIN Token Liquidity
- Dmitriy Berenzon acknowledges that DePIN tokens often suffer from poor liquidity in public markets, a concern for liquid funds.
- He suggests this isn't an intentional oversight by founders. Many DePIN founders come from non-crypto backgrounds, such as "the electrical engineering PhD that found issues in their own market structures," and may not initially prioritize liquid market dynamics.
- While solvable over time as networks mature and distribute more rewards, thereby increasing float, token depth currently remains an issue for institutional investors.
- Investor Insight: The technical, non-crypto-native background of many DePIN founders can explain initial shortcomings in token liquidity strategies, but a clear path to improving it is a crucial due diligence point.
DePIN Investment Strategy: The "Why Now?" Question
- Archetype's investment approach for DePIN, as explained by Dmitriy Berenzon, heavily emphasizes the "why now?" question, looking for industry inflection points that projects can leverage.
- This includes technological advancements (e.g., improved wireless capabilities) or favorable hardware economics (e.g., new solar panels being cheaper to install than maintaining old ones). The core idea is to identify what is naturally trending towards decentralization.
- Dmitriy draws a parallel to Webvan, a dot-com era grocery delivery service that was "20 years too early," highlighting the importance of timing. He states, "I think we've seen that with compute... where I do think the why now is AI demand."
- Actionable Insight: Researchers should analyze DePIN projects against current technological and economic trends to determine if the "why now" is compelling, particularly how AI's demand for resources like compute creates immediate viability.
The Crowded but Innovative Decentralized Compute Market
- Dmitriy Berenzon views the decentralized compute market as crowded, but considers this a positive sign fostering innovation.
- Current methods for orchestration (coordinating diverse resources) and routing (directing tasks) in these networks are described by Dmitriy as "quite elementary." A major challenge is efficiently coordinating heterogeneous resources (e.g., consumer-grade 4090 GPUs alongside high-end A100s) without performance being bottlenecked by the weakest machine.
- Emerging approaches include splitting AI models and routing different layers to suitable machines. Dmitriy mentions a senior executive from OpenAI starting a decentralized compute project focused on this, who sees "blockchains as the right tool for the job."
- The high cost of AI operations (e.g., $1.40 per inference for a deepfake AI video) underscores the need for cheaper decentralized alternatives.
- Strategic Consideration: The sophistication of a decentralized compute project's orchestration and resource management technology is a key differentiator in this competitive landscape.
Decentralized Compute: Competition and Verifiability Challenges
- Decentralized compute platforms compete with traditional providers primarily through cost discounts and the aggregation of diverse hardware supplies, from individual consumer GPUs (like 4090s) to small data centers with professional-grade A100s or H100s.
- Dmitriy Berenzon raises critical questions about supply: "do they actually own the supply and how sticky is that supply?" The ability to serve and host open-source AI models is another competitive vector.
- A significant hurdle is the verifiability of hardware. Dmitriy candidly states, "I do think it's trust me bro... I would be surprised if we had any trustless mechanisms to say your inference is coming from this specific machine type." Full on-chain verification would negate performance and cost benefits.
- Investor Alert: While some level of trust may be acceptable short-term, the lack of robust, trustless hardware verification in decentralized compute networks presents a risk that investors must weigh. Performance metrics might offer indirect clues.
The Evolving Landscape of Payments and Stablecoins
- Dmitriy Berenzon anticipates multiple winners in the stablecoin market due to its diverse "flavors." Beyond dominant 1:1 backed stablecoins like USDC and USDT, he sees growth for crypto-backed, yield-bearing (where yield is either a product feature or a go-to-market strategy), and non-USD stablecoins.
- He notes that non-USD stablecoins currently represent a small fraction of the market (around $600 million out of a $230 billion total, per Artemis data) but are emerging globally.
- The concept of closed-loop stablecoins is also highlighted, where a fintech might mint its own stablecoin for user deposits to earn treasury yield, similar to how "Venmo already does this today with your user deposits."
- Market Trend: The proliferation of stablecoin types, including those from traditional financial institutions and big tech, signals a broadening market but also increased competition.
The Potential and Puzzles of Non-USD Stablecoins
- Despite the potential for various non-USD stablecoins, Dmitriy Berenzon believes the US dollar will likely maintain its dominance due to its relative global strength.
- An ideal scenario for users in high-inflation countries is to "hold funds in digital dollars but pay in local fiat," with currency exchange (FX) occurring seamlessly at the off-ramp.
- Dmitriy sees an interesting angle for tokenized non-USD stablecoins in bringing FX markets on-chain. However, he questions the specific demand: "what is the demand for... an exchange to hold a fiat in a tokenized form rather than just on balance sheet."
- Local exchanges tokenizing local currencies could be a play for them to participate more effectively in B2B FX markets, leveraging crypto's efficiency.
- Research Focus: The key unsolved question for non-USD stablecoins is demonstrating a clear advantage for holding these assets in tokenized form over traditional banking for institutional FX operations.
Crypto's Role in "Payments 3.0"
- Dmitriy Berenzon frames the evolution of payments: Payments 1.0 (Swift-based correspondent banking) was slow and costly; Payments 2.0 (fintechs like Wise) improved speed via pre-funded accounts but remained opex-heavy.
- "Payments 3.0," he argues, utilizes blockchains as an "internet native settlement layer," enabling transactions in seconds for cents, 24/7, with full transparency.
- The "why now" for this shift includes mainstream stablecoin adoption, better on/off-ramps with deeper liquidity (as OTC desks and market makers hold digital dollars), embedded wallets like Privy enhancing UX, and pending clearer regulation.
- However, crypto-native firms face a huge challenge: distribution. Dmitriy points out, "Revolute alone has 60 million MAUs," contrasting with crypto neo-banks that have good products but struggle to acquire users at scale.
- Strategic Implication: While blockchain offers superior payment infrastructure, crypto-native payment solutions must crack the distribution code to compete with incumbent fintechs and financial institutions now entering the space.
The Ongoing Battle for User Experience (UX) in Crypto
- Dmitriy Berenzon agrees that crypto UX has historically been a significant barrier but sees promising improvements, particularly with a mobile-first approach.
- He highlights solutions like Privy, which can enable a "Gmail login, Privy wallet, Apple Pay on-ramp" experience, as crucial for mainstream adoption. Farcade, a mini-games platform on Farcaster, is cited as an example of smooth mobile UX.
- Dmitriy asserts, "I do think the next wave of at least on the consumer side is going to be mobile first." He believes good UX is achievable today, mentioning Slingshot's mobile trading app as a positive example.
- Key Development: The push for mobile-first applications with seamless onboarding (like Apple Pay integration) and App Store acceptance, rather than relying on PWAs (Progressive Web Apps)—websites that can function like native apps—is critical for broadening crypto's appeal.
Strategic Chain Selection for New Crypto Projects
- When advising projects on choosing a blockchain, Dmitriy Berenzon emphasizes the startup mantra: "your general goal is to not die." This means avoiding unnecessary risks by selecting established ecosystems.
- For consumer applications, Base or Solana are suggested as less risky choices, while Arbitrum is favored for DeFi due to its existing ecosystem.
- However, a crucial nuance exists: "if you are building something novel and you get distribution, you can choose whatever chain you want because the end users probably don't know." Projects like Pump.fun (with Pump Swap) and Pudgy Penguins (with Abstract) exemplify owning the end-user and commoditizing the underlying infrastructure.
- Founder Consideration: While established chains offer safety, projects achieving strong product-market fit and distribution can gain leverage to choose or even build their own tailored infrastructure.
Emerging Trends in Consumer Crypto: Blending Genres and Generative Content
- Dmitriy Berenzon observes a trend of "genres blending" in consumer crypto, where elements of SocialFi, trading, and gaming converge. Speculation remains a potent driver of demand.
- He highlights the "rise of generative content" as a significant area, where crypto can provide unique value in attribution and monetization for creators. This spans from simple memes to complex, AI-generated games.
- Dmitriy speculates, "I do believe in a few years we will likely have a new foundational model... specifically focusing on game generation." This could lead to an "over supply of content," making crypto-based mechanisms for attention and creator rewards (e.g., for remixed content) even more important.
- Future Watch: The intersection of generative AI and crypto for content creation, attribution, and monetization is a key area for investors and researchers to monitor, especially as AI tools democratize content production.
AI's Profound Impact on Value, Work, and Art
- The conversation touches upon the philosophical implications of AI-generated content. Dmitriy Berenzon predicts that "in five years only 5% of the content online will be human generated."
- While a "humanity premium" might exist for certain creations, for utilities like games, users will likely prioritize "the utility and the game mechanics and the fun," regardless of human or AI origin.
- This shift has deep implications for how society values various forms of work and creation, potentially redefining business models and educational approaches as AI democratizes skill acquisition and content production. Dmitriy notes learning about energy markets for hours via ChatGPT as an example.
- Societal Shift: The increasing prevalence of AI-generated content will force a re-evaluation of value, authenticity, and the nature of human contribution across many fields.
Integrating AI into Investment Processes
- Dmitriy Berenzon shares that while Archetype is considering AI for its investment process, they haven't fully implemented it yet, primarily due to concerns about the privacy of internal due diligence notes if processed via cloud-based AI.
- A potential solution being explored is using edge compute (like their portfolio company Exo, which allows for a "Mac mini data center" to run compute tasks locally) to run fine-tuned models (e.g., Llama) locally on their proprietary data, creating an "AI analyst."
- He acknowledges that other funds and DAOs are actively pursuing similar AI integrations for deal sourcing and diligence, suggesting it could become a competitive advantage.
- Investor Takeaway: Funds are exploring AI to enhance diligence and sourcing, but data security with sensitive information is a key consideration, potentially favoring on-premise or edge AI solutions.
AI Agents and the Future of On-Chain Commerce
- Dmitriy Berenzon argues that AI agents will predominantly use on-chain transactions because traditional banking systems are ill-suited for non-human entities, which lack the identity credentials required for account opening.
- Blockchains offer a natural fit, allowing agents to have wallets via public-private key pairs and transact using stablecoins for "microtransactions for specialized jobs." The low-cost structure of blockchains is essential for this.
- While these use cases are not widespread today, he anticipates development as more performant agents, better utility, and agent swarm (coordinated groups of AI agents) orchestration frameworks emerge. He mentions OpenAI's "Operator" product as a potential catalyst for giving agents real-world interaction capabilities.
- Dmitriy suggests a "three to six month time horizon" for seeing more significant developments in this space.
- Emerging Trend: The rise of AI agents is poised to create a new wave of on-chain activity, particularly for micropayments and automated services, making robust and low-cost blockchain infrastructure even more critical.
Reflective and Strategic Conclusion
- AI's relentless demand for compute is a powerful catalyst for DePIN, while AI agents promise a new era of on-chain commerce. Investors and researchers must focus on projects with sound tokenomics, viable distribution, and the technical prowess to harness these converging trends for real-world value creation.
