Having spent over 20 years actively trading and investing, I’ve always found it helpful for investors to see when a market has lost the narrative. I’ve also experienced the advantages of seeing new narratives, generally long before anything is revealed in asset prices. Taken together, some of the most accessible and valuable trading alpha derives from understanding the dynamic relationship between the market’s leading narrative and how prices are reacting to and from this consensus. Narrative and price can easily become a complex practitioner’s art, but regardless of how technology and artificial intelligence have begun to crowd out old trading models, this art continues to produce value for investors.

An asset class showing slow and methodical price change without a new clear narrative represents one of the most difficult investment situations for interpretation. Prices do not correspond to any model, and instead appear to react to what students of non-linear dynamical systems call the “hidden attractor”—a cause notoriously difficult to find with normal analysis. There are ways to model these slight signals (homotopy, numerical continuation, etc.); however, one philosophical error of the quant finance community is the attempt to extract substantial alpha from faint signals—the very mismatch that fails so many algorithmic trading strategies. Experienced traders understand that 80% of the alpha is in the 20% of the market (by occurrence) that represents the tails.

Slow, methodical price changes have characterized the past year for crypto markets. And for nearly every crypto asset, it’s been a slow grind down. Interestingly, though, none of the primary narratives have been lost. Everything we’ve argued on behalf of the long-term crypto evolution still holds—a widely distributed ledger does make a better currency and ultimately a better store of value, and so on. So why the lower valuations? This is a letter I first wanted to write months ago, but I waited, as it seemed that some consensus was just starting to form on the horizon. I think markets now have a clearer view, and this also comes with a solid plan for crypto going forward.

In many ways, investment markets are competitive for capital, and there can only be one or two leading narratives. Leading narratives cycle through several growth waves toward a final “super wave,” where well-capitalized startups, IPOs, and market incumbents capture nearly all available investment dollars for a period of time—generally until the narrative reaches a new consensus. We are in one of those market super cycles now.

In my travels and research for our fund, I often have a chance to interact with business, investment, and political leaders. For the past year I have witnessed only two dependable topics of investment interest: Space and Artificial Intelligence. One could cast space aside (or possibly AI too) as some transient “Elon effect,” with his SpaceX IPO and bonus structure for populating Mars with at least 1 million people. However, while Elon may be the spark that lit the fire, the fuel behind the space movement is more relatable to the generation of current global leaders who grew up in the 1960s space era and were inspired by a world suddenly able to dream toward the stars. After the American moon landings, space development froze for four decades until Elon started proposing to people that 1) yes, we can cost-effectively and dependably get people and freight to orbital positions, and 2) there actually are many untapped, profitable space-based business models that can benefit humanity once we engineer sufficient infrastructure and systems for the vacuum of space. These ideas have ignited a world of excitement for countries to restart their investing in space—as a commercial industry—for all its benefits of scientific advancement and economic growth.

There is a lead-lag effect for space development because, as big as its promises are, there are relatively few people working on space problems. There are an estimated fewer than 500,000 people working in the global space industry, versus the global oil industry with over 12 million workers. I find this surprising given there are multiple technology paths that would allow the space industry to return to Earth substantially more energy than the global population could annually consume—not to mention the bio-pharma, microchip, and other promising manufacturing developments only possible in a weightless environment. Alas, the space investment sector has also suffered a long, protracted investment cycle, with most publicly traded space-related companies remaining unprofitable. As is commonly recounted around the halls of U.S. Space Command, “Space is Hard”—a deep realization of the engineering challenges ahead. Currently there is huge demand-side interest for space activity, and very limited financing pipelines. Most of the world is still accustomed to space industry investment being the activity of sovereigns. Our future in space absolutely depends on commercial development, and this is just now starting to appear at scale—mostly only in the United States, with a small bit in India, Japan, and New Zealand. Depending on how the SpaceX IPO continues to price, we could see a renewed interest in the space economy for years to come.

Artificial intelligence is quite different. It’s abstract, non-physical, and universally applicable to our knowledge economy. It’s also fast-moving. OpenAI has captured the fastest rate of consumer product adoption in history. AI is this dislocating claptrap that promises the most magical mechanical transformation since the beginning of the industrial age: the not-so-distant possibility that we could transfer most of our common human work to machines. Computers will no longer just be data storage and recall machines (digital libraries, in a sense); they are now factories with complex chained value-add processes that permit “dark factories,” where no lights are needed because they are entirely run by machines. This is already happening in China and other global manufacturing hubs.

The value proposition of AI is real and substantial, and we’re really only about 5 years into AI being in the public vernacular, and only 9 since its first material iterations in a research lab. AI is a big wave showing the persistence of our herd mentality when approaching markets. Capital is crossing asset classes toward names like SNDK, which is up over 650% since the beginning of 2026. These are the hyperscalers: about $3.1 trillion globally in just the last 3 years—and this does not include the additive market cap to companies like Nvidia, which alone has added $3.8 trillion of market cap in the past 36 months. The most recent numbers suggest there is about $8 trillion planned for AI infrastructure development over the next several years—and that capex is going to need to produce at least $800 billion in value returned to its investors to maintain current valuations. Amazingly, it’s not entirely improbable that in the U.S. alone, with a 2026 GDP of around $32 trillion, AI would provide only about 2.5% of our economic activity. The actual cash flows are, of course, not yet proven; yet it’s quickly becoming easier and easier to argue AI-based economic growth. In effect, this investing has pulled capital away from other asset classes, especially stylistically similar crypto markets. This is not because there’s necessarily anything wrong with crypto, but more because so much “right” has been growing over in AI sectors.

And where does this leave crypto? Again, there’s nothing wrong with crypto or the primary investment thesis, but markets have devalued the overall market cap of cryptocurrencies and digital assets by about $2 trillion since summer 2025 highs. Some assets have sold off more than others. Our portfolio has maintained positions in some of the strongest assets, but even many of these have sold down more than 50% in the past year. Overall, it’s very difficult to make money in an out-of-favor market where all assets have sold down. I do, however, think there are some silver linings, as this is by no means the end of the road for Bitcoin or other mainline assets.

Cryptocurrency — for man or machines?

Something that has frustrated me for several years has been the genuinely slow public adoption rate of even the most mainline digital assets. Bitcoin, ETH, and the famous pseudo-distributed XRP. People who were not early adopters have not quickly (or at scale) come around to become middling adopters—even in one of the worst inflation environments in many years. I frequently blamed this on technology adoption or generational sensibilities, and yet nearly every investor in our fund is older than I am, and some are exceptionally advanced in various technology fields. And when our industry argues crypto to be a more “human”-respecting solution than government-issued centralized fiat currencies, the man on the street holding a slip of paper with the president’s image argues that his physical exchange “feels” more human. In some ways, the entirely abstract experience of crypto is a barrier for adoption and common use.

And yet, because of AI (and supported by space as well), there is a coming and substantial use case that has not priced. We have monies to calculate, according to some standardized system, the value of our human work and possessions—but in reality to optimize our tradeoffs and opportunity costs. The proposal of AI is a world where substantial parts of our economy are affected or accomplished in whole by computer logic and physical manipulating machines. How does one calculate this value? By its cost? It’s been popular of late to relate the cost of AI back to some simplified equation of electricity (energy) + tools (servers) + cost of transactions (selling and marketing). This assumes a fairly simplistic supply chain—where people are the primary and direct engagement for AI-assisted processes. But what about when AI makes possible an entirely automated production process, in a world (space) of a completely different scarcity stack? Not to mention it will be genuinely more efficient for computers to not only do the work, but also do the transactions on our behalf—just like an automated trading algo. We already do this in the very narrow confines of centralized stock certificate exchanges—but what about an entire industry supply chain organizing itself and trading with itself in a way more efficient than any person could engage? My point: with agentic agents, we start to see how humans are not crypto’s primary design target.

Bitcoin’s whitepaper describes “peer-to-peer electronic cash,” but the deeper innovation was removing trusted third parties (banks, governments, humans) in favor of consensus rules enforced by nodes and miners—i.e., machines. The key design principles of Bitcoin were:

  1. Trustless by default. Every major chain assumes participants (and especially smart contracts) cannot trust humans. Validation, settlement, and enforcement happen through cryptographic proofs and consensus algorithms. Humans can build interfaces, but the rails run on silicon.
  2. Code is the law of the system. Smart contracts (Ethereum, Solana, etc.) are self-executing programs. Once deployed, they operate autonomously 24/7/365 without human intervention, holidays, or office hours. This is the opposite of traditional finance, which is built around human gatekeepers, business days, and manual oversight. This also makes sense in the vacuum of space.
  3. Incentive alignment for machines. Block rewards, transaction fees, staking, and MEV (miner/maximal extractable value) create economic loops optimized for automated participants. High-frequency trading bots, arbitrageurs, liquidators, and yield farmers were among the earliest heavy users—not retail humans.
  4. Programmability over usability. The UX for average humans is notoriously bad (seed phrases, gas fees, bridges, wallet security). This is a feature for machines, not a bug. Machines don’t care about friendly apps; they thrive on deterministic APIs, composability, and atomic execution.

There is evidence from crypto’s evolution (2010s to 2026). In the early days, Bitcoin enabled machine-verifiable scarcity and transfer. Satoshi disappeared—the system didn’t need a human steward. Then we had DeFi Summer: protocols like Uniswap, Aave, and MakerDAO turned finance into executable code, and billions in TVL moved according to algorithmic rules rather than human loan officers. Now it’s the 2024–2026 wave, an explosion of agentic crypto and AI-crypto convergence. Autonomous AI agents (powered by models from xAI, OpenAI, Anthropic, etc.) are now holding wallets, executing trades, managing positions, and even launching tokens or DAOs with minimal human oversight. Crypto rails provide the economic primitive (tokens as programmable money) that AI agents need to act in the real economy. Lastly, we’re on the cusp of new machine-to-machine (M2M) use cases: IoT devices paying each other for data or energy in real time; orbital satellites or space infrastructure settling bandwidth/microtransactions; prediction markets and oracles feeding verifiable data to algorithms. These are inherently machine-native. This is just a quick run through my views, but it’s clear to me that a major sustaining value for crypto markets and crypto asset prices will be its capacity to engage and become a platform for machines. The good news is that this is a massive value proposition, and it could match the trillions currently flowing into the AI economy.

In this view, crypto becomes the native financial system for the Age of Agents. Humans speculated their way into it (memecoins, NFTs, retail FOMO), but the underlying design—just now coming into realization—was always preparing for a world where economic activity increasingly happens between non-human entities at speeds and volumes humans can’t match. Humans still provide the liquidity, narrative, regulatory navigation, and adoption push. Speculation and culture (community, memes) drive price action and attention. Many projects explicitly target human users. But these feel like transitional layers—the long-term value accrues to protocols that machines can use seamlessly and at global scale. This thesis aligns with thinkers like Balaji Srinivasan (network states, sovereign individuals via technology), Vitalik Buterin (in his more technical writings), and observations from crypto-native AI projects in 2025–2026.

This argument is actually tracking closer to the reality of what we are seeing in the world: crypto wasn’t built to replace your bank account. It was built to replace the entire concept of human-mediated economic trust with something machines could natively understand, execute, and scale into a parallel economy. The human era of crypto is the onboarding phase; the machine era is the destination. This is the future AI + Crypto convergence.

Now, as your humble investment manager, I do recognize that some of these ideas could be considered a grand departure from the last decade of crypto theory. I also recognize they could be wrong. However, for the world of investing, even a wrong theory that has the correct correlation to market prices can become a very worthwhile and profitable theory. What does this mean for our portfolio? Most clearly in my mind, it means that the guiding thesis for a number of the largest crypto assets is likely broken. Uniswap, for example: I love its initial thesis—a fully decentralized crypto exchange (DEX). Yet, after 5 years of trying to pull itself off the bottom (and sometimes doing so with 500% runs), it’s probably not going to ever reach its grand position as a leading DEX. And there are other assets, substantially priced down after many good attempts. The external environment has matured, and for many of these assets, their narrow product-market fit for some financial or practical solution is no longer as worthwhile for capital in the age of Artificial Intelligence. The machine world does not need some of these solutions. The good news is that there are also new assets—HYPE, for example. First launched in Feb 2026, Vellum made our first allocation in April 2026, and it’s currently up 100% this year.

We are actively working to manage downside risk with strategic cash balances while also allocating to assets that are performing the strongest in class. The strong performers are likely to overcome and replace many of the top-20 market-cap assets we’ve been holding over the past years, and we are actively rebalancing and allocating to capture growth assets in this rotation process. We do need an overall market lift, and for Bitcoin to show upside leadership. In the meantime, we are working against these headwinds to outperform the general market indexes as we have in past years. When AI and crypto merge (possibly with space too), the growth for crypto may be far more substantial than anything we’ve seen to date, and our plan is to be well positioned for this wave.

Sincerely,

Eric Kovalak
Managing Partner
Vellum Capital LLC