In a week where the cryptocurrency market saw Bitcoin holding at $27,119 and Ethereum maintaining its position near $1,890, the emergence of AI-native venture capital firms targeting Web3 infrastructure represents one of the most significant developments in the convergence of artificial intelligence and decentralized technology. Moonfire Ventures stands at the forefront of this shift, having raised $115 million to deploy AI-driven investment strategies across European startups — many of which operate at the intersection of AI and blockchain.
The Agentic Protocol
Moonfire Ventures operates as an AI-augmented investment firm that processes between 50,000 and 60,000 funding applications each week, using machine learning algorithms to identify the most promising opportunities. The firm’s founding team, led by Mattias Ljungman, built the platform on the premise that AI can dramatically improve investment decision-making while freeing human partners to focus on relationship-building with founders.
Partner Mike Arpaia described the approach as both effective and transformative, noting that AI systems allow the firm to process deal flow at a scale that would be impossible for human analysts alone. The AI pipeline narrows tens of thousands of applications down to approximately 100 high-quality opportunities that match the firm’s investment thesis — a filtering efficiency that gives Moonfire a significant edge in identifying early-stage talent.
Neural Network Integration
The machine learning models powering Moonfire’s investment process analyze multiple dimensions of startup quality. Founder backgrounds, market timing, competitive landscape, and technology stack all feed into the evaluation pipeline. For crypto and Web3 startups specifically, the models can assess on-chain metrics, tokenomics design, and community growth patterns alongside traditional startup indicators.
This approach represents a broader trend in the venture capital industry where AI is being integrated into every stage of the investment lifecycle — from deal sourcing and due diligence to portfolio monitoring and exit timing. The implications for crypto startups are significant: AI-driven VCs can evaluate more deals faster, potentially funding projects that traditional firms might overlook.
Token Utility
The $115 million fund targets early-stage companies across the AI and Web3 spectrum, with particular interest in decentralized infrastructure projects, known as DePIN (Decentralized Physical Infrastructure Networks). These projects use token incentives to build distributed networks of computing resources, storage, and connectivity that can serve both AI training workloads and blockchain validation requirements.
For DePIN projects, the alignment between AI’s insatiable demand for computing power and blockchain’s distributed architecture creates a natural synergy. Tokens serve as the economic glue that coordinates resource providers and consumers across these networks, and AI-powered analysis can optimize token distribution and network efficiency in real time.
Potential Bottlenecks
Despite the promise, several challenges temper the AI-VC-crypto convergence. AI models trained on historical startup data may struggle to evaluate novel crypto business models that have no direct precedent. The crypto market’s inherent volatility introduces noise that can confuse predictive algorithms, and the regulatory uncertainty surrounding both AI and cryptocurrency creates risk factors that models may not adequately capture.
Additionally, the concentration of AI-powered investment decisions among a small number of well-funded firms could inadvertently create herding behavior — where multiple AI systems converge on the same investment thesis, reducing the diversity of funded projects and potentially inflating valuations in specific sectors.
Final Verdict
Moonfire’s approach signals a fundamental shift in how venture capital will operate in the Web3 era. The combination of AI-driven deal processing with human relationship-building creates a hybrid model that leverages the strengths of both approaches. For crypto projects building AI-integrated infrastructure, the availability of AI-native investors who understand both domains represents a meaningful advantage. The $115 million commitment from Moonfire alone suggests that the capital markets are taking the AI-crypto convergence seriously — and that the projects building at this intersection will have access to more sophisticated funding sources than ever before.
Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. Always conduct your own research before making investment decisions.
50,000 applications a week filtered by ML. the founders never even see 99% of them. feels like applying to a black box
processing 50k applications a week with ML means the model rejects good founders who dont fit the pattern. bias at scale basically
Jakub W. the model rejecting good founders is exactly the problem. ML optimizes for pattern match not genuine conviction
black box is the point though. human VCs have bias, ML at least filters on metrics. whether thats better long term is still an open question
ML filters on metrics but the metrics themselves are chosen by humans who have bias. the bias just moves one layer up
99% filtered by ML before a human even glances at your deck. imagine spending months on a pitch and getting rejected by a sigmoid function
yolotrade getting rejected by a sigmoid function is brutal but honestly the same thing happens with human analysts, just slower and with more small talk
yolotrade getting rejected by a sigmoid is brutal but at least its fast. human VCs ghost you for months and then pass with no feedback anyway
yolotrade getting rejected by a sigmoid is still better than getting ghosted by an associate for 3 months. at least the ML has the decency to be fast
yolotrade honestly getting rejected by a human who didnt read past slide 3 feels worse. at least the ML model is consistent in its indifference lol
Sven L. 99% filtered before human eyes. imagine building a product for months and getting killed by a logistic regression
birk_ether_ a logistic regression killing your seed round before a human sees it. startup culture in 2026 is just applicants vs ML models
$115M for european AI-blockchain startups. mattias ljungman knows how to pick em, was at atomico before this
atomico alumni tend to do well in europe. the real question is whether AI-picked startups outperform traditional VC picks or if its just cost cutting dressed up as innovation
50k applications a week and Ljungman thinks AI improves decision making. the real innovation is processing speed not quality. pattern matching on pitch decks is how SoftBank ended up with WeWork
Nadia H. the softbank wework comparison is exactly right. ML pattern matching on pitch decks just automates the same VC herd mentality at scale
vc_metric_ exactly. SoftBank pattern matched WeWork with humans and look how that turned out. ML just lets you make the same mistake 50,000 times per week
Naveen R. the math only works if hit rates exceed baseline. nobody has 5 years of data on ML-screened VC picks yet. could be worse than humans and nobody would know until fund II
fat_tail_ right, the backtest problem. ML models trained on bull market deal flow will look genius until the regime shifts and every pick underperforms
Nadia H. the SoftBank WeWork comparison is spot on. pattern matching pitch decks with ML just automates the same blind spots traditional VC already has. speed not quality
50k apps a week processed by AI and Ljungman calls this improved decision making. sounds like faster rejection at scale
throwaway_vc_ faster rejection at scale is exactly right. 50k apps a week means 49,990 founders got a no from a machine. at least send a human rejection email
115M is small for a VC fund but the AI angle means they can deploy faster with less staff. the math works if the model actually picks winners
Mel A. 115M is small but the AI screening model means they can deploy across more bets than a traditional seed fund. the math works if hit rates are even slightly above baseline
Arpaia saying AI frees partners for relationship building is hilarious. the relationship is literally the one thing you can’t automate
24 comments on this thread already. moonfire processed 50000 apps this week and the market still has zero evidence their model outperforms random selection