Founder of Relayer Capital Discusses AI-Driven Crypto Projects: VVV, GRASS, and NEAR

Organized & Compiled by Deep潮 TechFlow

Guest: Austin Barack, Founder of Relayer Capital (a digital assets investment fund focused on AI)

Host: Andy

Podcast source: The Rollup

Austin Barack: My AI Bull Thesis (...And What I'm Holding)

Broadcast date: May 23, 2026

This episode of the AI Supercycle features Austin Barack, founder of Relayer Capital, in a discussion centered on Venice, Grass, NEAR, Akash, and the broader Crypto x AI asset framework. Austin’s core thesis is that AI is scaling user data volumes to levels previously unimaginable for internet products, making privacy-focused AI, data supply, inference compute, decentralized training, and agent infrastructure critical sectors. He believes there is a significant misalignment between the revenue growth, user growth, and valuations of Venice and Grass, and that NEAR’s positioning in cross-chain intents and agent infrastructure is undervalued. Regarding the broader crypto market, Austin emphasizes that investors should evaluate token value creation through the lens of “net token value flow,” rather than mechanically focusing on buyback and burn mechanisms, to truly determine whether token holders are capturing the value generated by the underlying business.

Summary of Key Insights

The Real Value of Venice and Privacy AI

• In AI, privacy is more important than in any other context, because you're sharing health data, financial data, and connecting all your files, revealing your entire life in ways never seen before.
• This is not 10 times more data than social media, but 100 times more.
• What’s truly great about Venice is that it doesn’t just let you use AI in a private environment—it does so without compromising the user experience, and even enhances it.
• Tokens can become a vital part, significantly enhancing the experience, but most users don’t need to understand tokens to find the product useful.

The economic models of VVV, DM, and Venice

• The role of DM is: for every 1 DM Token you hold, you receive $1 worth of free inference computing credits daily on the Venice platform. Think of it as a perpetual benefit, equating to $365 in computing credits per year.
• Its limit expires if unused and does not accumulate over time. If you only use 50 cents one day, it won’t become $1.50 the next day—you’ll start fresh with $1 again.
• If all DMs are locked and used for inference computation, Venice’s maximum cost is $38,000 per day, with an annualized cost of approximately $10 million, and this cost will not exceed this amount.
• I believe DM should be valued more like corporate bonds rather than being discounted at an excessively high rate.

Grass and AI Data Demand

• Grass collects datasets and sells them to cutting-edge AI labs that need data to train new models.
• This isn't random web scraping; it must be highly professional, involving very specific datasets with high quality.
• The scale of investment in models is very large, making Grass a beneficiary of this trend. The more investment in models, the greater the demand for data.
• According to recently disclosed data, the project’s ARR is approximately $50 million. Currently, its valuation is around $400 million. In my view, valuing a project growing this rapidly at just 5 times revenue is completely unreasonable.

NEAR, Akash, and the AI Stack

• EAR Intents is highly practical and may be one of the best cross-chain swap experiences available today. It also plays a crucial role in the agent space.
• I think NEAR is doing an excellent job on the intents side. They’re also working on many other things, such as privacy intents and other elements around AI usage, making it one of the few L1 projects that has truly found its unique positioning.
• Akash. They initially started in the decentralized CPU market and later shifted to the GPU market.
• My primary areas of focus include decentralized training, inference, and compute markets; agent infrastructure; data; and consumer-facing model usage applications.

Token Value Capture and Market Differentiation

• Hyperliquid is first and foremost a highly successful business model, which is why people are drawn to its token; buybacks are simply one way it delivers value to token holders. If the underlying business weren’t operating well, the token price wouldn’t naturally rise even with a buyback mechanism in place.
• The core issue isn't what the mechanism is called, but whether token holders can capture the maximum value generated by what you've built.
• Each project and mechanism requires specific analysis. But the core question is: Can token holders benefit from the value being generated by the system?
• Investors can choose from a smaller pool of high-quality projects. Currently, capital is flowing concentratedly toward projects like Venice, HYPE, Grass, AERO, NEAR, and Zcash.
• For investors seeking 5x to 10x, or even 3x returns, this is now a more favorable time to succeed than ever before. While you might still achieve 100x returns, I believe there’s a group of projects currently doing something truly compelling—these are the assets I’m paying attention to and investing in.

Host Andy: Not long ago, I used Venice for the first time. I typed into Venice: "Is this really private?" It replied: "Yes, the reasoning process is private," and then went on to explain a lot. I responded with: "That’s awesome." It immediately followed up: "Yes, it really is awesome, isn’t it? With Venice, you can…"

So, the first time you use Venice, there’s a striking moment: you suddenly realize that all the chat content you’ve ever entered with typical AI providers—though not necessarily public—has been sent to large vendors. Your most private journals, business secrets, plans, and more have all been handed over to them.

From a high-level perspective, how do you view Private AI and Venice in terms of market structure, investment thesis, and founding teams?

Austin:

Venice is interesting because it has gone through many different stages of iteration. I first came across this project in January last year. At the time, I was closely following Virtuals and aixbt, and a significant portion of Venice’s early airdrop was allocated to holders of tokens within those ecosystems, which is how I first encountered it.

At the time, it was already an intriguing product. It’s crazy that, although only about 16 months have passed since then, AI back then was nowhere near as ubiquitous as it is today, nor had it become an indispensable part of everyone’s daily life. During this period, AI—whether Claude, ChatGPT, or other services—initially seemed to be replacing Google searches. People would say, “I no longer use Google to look up a question; I just ask an LLM on an AI platform.” But now it has evolved into enabling content creation, task resolution, and even the ability to have an entire team of agents working alongside you.

Austin:

I believe people are gradually realizing that privacy is more important in AI than in any other context, because you're sharing health data, financial data, and connecting all your files—revealing your entire life in ways never seen before.

In the past, discussions about privacy were mostly within the context of social media—such as whether my account was public or private, or whether Facebook held too much information about me. But AI doesn’t just have 10 times more data; it has 100 times more.

What’s truly great about Venice is that it doesn’t just let you use AI in a private environment—it does so without compromising your user experience, and even enhances it. That’s because you’re not locked into a single model. For example, if you use ChatGPT, you’re tied to OpenAI’s model updates; if you use Anthropic, you follow Anthropic’s model evolution; similarly, with Gemini or open-source models, each comes with its own limitations.

In Venice, you can select the most suitable model for each task or choose which models you want to use yourself, making it highly customizable. They first built an excellent consumer product, and most users don’t even know what a token is.

Tokens add an interesting element on top. I’m very optimistic about what they’re building. The key point is that I believe crypto consumer products will evolve into a form where tokens can become a vital part, significantly enhancing the experience—yet most users won’t need to understand tokens to find the product useful.

Host Andy: This does resemble a form of consumer product breakthrough—underpinned by Crypto, but users don’t need to understand it first. However, it also introduces an interesting token structure. Some have compared it to Luna: staking VVV yields DM Tokens, which then form a kind of debt structure through credit limits.

Host Andy: So how should we understand the VVV Token and DM Token within the current Venice flywheel? Also, could you explain Venice’s revenue side, since they are indeed conducting some buybacks, but the scale isn’t particularly large? How exactly do these two tokens function, and why is it not like Luna?

Austin:

They just announced 3 million users, with very rapid growth. Approximately 1 million new users were added in the last 3 months, whereas it took around 7 months to reach the previous 1 million. Growth has been accelerating.

Austin:

They have two tokens. The first is VVV, and protocol revenue is used to burn VVV. Users can also stake VVV to receive free membership. But the most interesting feature is that users can stake and lock VVV to mint a token called DM. You can also buy DM on the open market, but the core mechanism is staking VVV to mint DM.

The role of DM is: for every 1 DM Token you hold, you receive $1 worth of free inference computing credits daily on the Venice platform. Think of it as a perpetual benefit, equating to $365 in computing credits per year.

However, if you don’t use your allowance, it expires—it doesn’t roll over or accumulate over time. If you only use 50 cents one day, it won’t become $1.50 the next day; instead, it resets back to $1. I think this creates a very interesting mechanism, somewhat akin to a loss-leading customer acquisition tool. This differs from Luna, which reached an extreme state by issuing an enormous number of tokens, causing the stablecoin’s scale to reach billions, even tens of billions, of dollars. Venice, on the other hand, is very clear about this: they’ve capped the potential cost within a well-defined range.

Currently, the number of DMs that can be minted per Venice Token gradually decreases as the number of DMs in circulation increases, effectively setting a hard cap of approximately 38,000 DMs. Under current conditions, if all DMs were locked and used for inference computations, Venice’s maximum cost would be $38,000 per day, with an annualized cost of approximately $10 million—and this cost will not exceed this amount.

Currently, approximately 10,000 DMs are used daily for inference computations, corresponding to an annualized cost of about $3.5 million. This cost is offset through their business revenue. They offer Pro and Premium subscription plans, priced between $18 and $68 per month, and even higher. Additionally, users purchase Tokens or additional credits to access models while using the platform.

Notably, their daily token usage has grown from billions initially to approximately 70 billion recently, an increase of about 15 times over the past few months. I believe the key difference here from Luna is that the company has a maximum potential cost, and DM users, while using DM, also utilize subscription services. If their daily needs exceed $1 per token, they purchase additional credits. This cost is easily covered by business revenue, which has already significantly surpassed it.

Austin:

On the other hand, what makes DM so compelling is its ability to guarantee you access to computing resources in the future. The market is currently valuing it at approximately a 20% discount rate, with the price hovering around $1,800.

I believe this asset should be priced more like a corporate bond, using a discount rate of 8% to 12%. At a 10% discount rate, its price would be approximately $3,650. For context, when I first started following it, the price was in the $200 range.

Host Andy: I was also thinking, how could an asset that generates $365 in equity per year possibly be worth only $200? Unless the market believes Venice simply cannot sustain this mechanism.

Austin:

That's right. At that price point, it was almost a no-brainer investment for me. Even now, I still believe there's room for further upside.

However, if you look beyond DM and examine Venice’s overall economic situation, you’ll find numbers that are astonishing. Moreover, its growth pattern is completely different from that of most projects we see in the cryptocurrency industry—it resembles growth rates only possible in the AI field, which is what makes it so compelling.

Host Andy: So you firmly believe that Venice’s VVV asset price is approaching $20. Do you still think the $1.5 billion to $2 billion valuation range is significantly undervalued for VVV?

Austin:

Yes, when I bought it for the first time in January, it was around $2.50. At that time, they were processing only a few billion tokens per day. Now, it’s about 15 times that amount.

Initially, they processed only a few billion dollars in token trading volume per day, but this has now grown to 15 times that amount. Their user base has increased from 1.5 million to 3 million today. Based on my estimates, their revenue is at least three times what it was back then.

Currently, Venice is valued at approximately 20 to 30 times its annual revenue, and it’s a company still growing at a rate of 20% per month. From this perspective, I believe its valuation remains very low. You could even compare it to OpenRouter—while OpenRouter’s valuation is similar to Venice’s, its revenue may be slightly lower and its growth rate may not be as fast as Venice’s.

The key difference is that Venice has direct access to customer resources. It is not merely an infrastructure providing backend services, but a platform that users actively engage with daily. Personally, my only way of using AI right now is through Venice.

So, I believe its potential is still significant. Of course, this is merely my personal opinion and does not constitute any investment advice.

Host Andy: I’m not very familiar with Grass. You’ve mentioned this project several times before, and it seems to be on the verge of rapid growth. Of course, its price may have pulled back today. I heard its annualized revenue has already exceeded $50 million and is accelerating at a triple-digit growth rate. Could you briefly explain Grass’s core revenue model? How does it make money, and why is it so appealing?

Austin:

Grass collects datasets and sells them to cutting-edge AI labs that need high-quality, specialized data to train new models. These labs are generating new models at a rapid pace, but doing so requires vast amounts of data—not randomly scraped from the internet, but highly specific, high-quality datasets.

This is the role Grass plays, because the scale of investment required to build these models is enormous, making Grass a beneficiary of this trend. The more investment in models, the greater the demand for data.

Austin:

The Grass team has been building for many years. I remember that last quarter, they generated approximately $3 million in revenue. By the end of the year, they were earning $12 million to nearly $13 million in a single quarter. Based on my estimates, they are now growing even faster. In the next one to one and a half months, they will hold a token holder call, where we’ll get more information.

But this is a project experiencing triple-digit growth. According to recently disclosed data, the project’s ARR is approximately $50 million. However, I expect it may now be approaching $80 million. Currently, its valuation is around $400 million. Therefore, valuing such a rapidly growing project at just 5 times revenue seems entirely unreasonable to me—it’s a strong candidate for a significant repricing.

Host Andy: Is there any working relationship between Grass and Venice?

Austin:

Not at the moment. Venice typically doesn’t build its own models, so it doesn’t matter right now. Who knows what the future holds? But I see them as two different sides of the same equation. One question is: How do you use AI, and how do you use AI privately? The other question is: How are models originally built? Grass and Venice are each addressing one of these two aspects.

Host Andy: So Grass is trading at about 5 times revenue. In the crypto industry, some things trade at 20, 30, 40, even 50 times revenue. Do you think the $400 million range is a bit of a no-brainer?

Austin:

Yes. I think an important point is that the crypto industry also has other assets trading at relatively low multiples, but they haven’t grown. People come to crypto because they want to invest in growth.

So I think many low-leverage cases aren’t necessarily valid because there’s no cash flow. But something like Grass is one of the best examples of rapid growth. I believe that alone makes it worth paying attention to, not to mention that, in my view, it’s also quite cheap.

Host Andy: Do you have an investment thesis for NEAR? Are you following NEAR?

Austin:

I’ve been following NEAR closely. Even without considering the AI component, NEAR is a fascinating project because it serves as the underlying infrastructure for a large volume of cross-chain swaps. In October and November last year, NEAR garnered significant attention due to its role in Zcash transfers.

NEAR Intents are highly practical and may be among the best cross-chain swap experiences available today. They also play a crucial role in the agent space. In my view, NEAR is one of the most suitable infrastructures for enabling cross-chain swaps, as it avoids many of the dependency issues faced by other projects.

They are growing rapidly in this area. Now, if you're a Layer 1, I think you need to excel in one of several directions: either you offer a vertically integrated app experience, or you're 10 times better at something specific, or you're extremely strong in a particular category of applications.

I think NEAR is doing an excellent job on the intents side. They’re also working on many other initiatives, such as privacy intents and other elements around AI usage, making it one of the few L1 projects that has truly found its unique positioning.

This reminds me of how NBA players are categorized. Today, there are many new L1 and L2 projects on the market, much like promising rookies. Over time, some will grow into superstars, while others will fade away. But there’s also a category of players known as “role players”—those who excel precisely within their defined roles, such as Lu Dort or Alex Caruso of the OKC Thunder.

NEAR feels to me like this kind of player. It’s not LeBron James, but it’s incredibly important because it excels at what it does.

Host Andy: Another project that has always been underestimated, and that Robbie always emphasized to me, is Akash. Unfortunately, he’s not here today. Akash entered the fields of distributed inference, distributed models, and decentralized training very early on, right?

This sounds like the first narrative cycle of Crypto AI. After that, we went through a wave of fake agent projects tied to meme tokens. Now, it seems we’re entering the next phase of decentralized inference and model training—but this time, the products are far more powerful. Have you seen what Akash is doing? What’s your investment perspective on this project?

Austin:

I did follow Akash—they originally started in the decentralized CPU market and later shifted to the GPU market. Now, you can actually check how much data is flowing through OpenRouter, and a significant portion of that data passes through Akash, specifically Akash ML, which is pretty cool. And this data is public, so anyone can view it.

However, I must admit that Akash hasn’t been one of the projects I’ve been most closely tracking. But it’s exciting to see that, after so many years and continuous iteration, this team has finally achieved genuine product-market fit—and it appears to be accelerating.

Host Andy: There’s a project called GitLab that has a very small market cap on Base, but its daily token production is strong. Currently, a wave of highly speculative AI tokens has emerged on Base, and there are many niche sub-sectors within this puzzle that require understanding.

I’d like to ask more broadly: within this AI stack, are there certain components that are best suited for integration with blockchain to achieve massive growth? We’ve already seen Venice offering private, uncensorable ChatGPT; NEAR serving as infrastructure for agent markets; Akash providing Akash ML; and Grass focusing on datasets.

In the AI stack, which key sectors or components are most likely to be replaced by blockchain technology, or are best suited for on-chain implementation?

Austin:

I believe the primary focus is on privacy contexts, including private use of large language models (LLMs) and uncensorable use. Then there is the data collection required to train models, which is what Grass is doing.

Next is the inference computing and computing power market. You mentioned Akash earlier. We’re also seeing other inference markets emerging. There’s another project built around DM that also offers additional services, allowing users to sell their idle computing power—it’s called AnC. This is an interesting project I’ve been following. Although it hasn’t launched a token yet, I think they’re doing some very cool things, especially in integration with Venice and DM.

I believe another important direction is decentralized model training. The challenge lies in preserving ownership and monetization capabilities of models while building open-source models through private weights. Several teams are currently exploring this space. For instance, I find Pluralis one of the most interesting projects in this area. Nous Research is also doing some very compelling work around Hermes. Additionally, Prime Intellect and several other teams are actively involved in this field.

Therefore, my primary areas of focus include: decentralized training, inference, and compute markets; agent infrastructure; data; and consumer-facing model usage applications.

Host Andy: Recently, you’ve been emphasizing another perspective: we need to understand token models and economics in new ways. You’ve consistently been a strong supporter of projects like Aerodrome and Hyperliquid.

Before wrapping up, I’d like to set aside the AI context and ask a broader question: How do you view net token value flow? In other words, using the framework of credits (inflows) and debits (outflows), and analyzing a crypto asset through a plus-minus ledger, how do you think the industry’s mindset around token economics is evolving? What is your current framework? Do you agree that investors should understand an asset’s net token value flow as if viewing a balance sheet of inflows and outflows?

Austin:

I believe there are several different ways to look at this issue, and it’s not a one-size-fits-all situation.

We can start by discussing the high-level mechanism of buybacks and burns. Hyperliquid made this mechanism very popular—people would say, “Look at how well Hyperliquid is doing; it has buybacks and burns.” But for every Hyperliquid, there are nine other tokens trying to adopt the same buyback and burn mechanism, yet their price performance is terrible.

What’s the lesson here? The lesson is that Hyperliquid is first and foremost a highly successful business model, which is why people value its token—and buybacks are simply one way it delivers value to token holders. If the underlying business weren’t functioning well, even a buyback mechanism wouldn’t naturally drive the token price upward.

This is the first issue I believe people often confuse.

The second question is whether you are truly creating value for token holders. Whether you use buybacks and burns, buybacks and distributions, reinvest funds back into the business, or deposit funds into a bank account to enhance balance sheet flexibility, the core issue is whether token holders can capture the maximum possible value generated by what you’ve built.

For example, Hyperliquid operates this way, and so does Aerodrome. As for Grass, many hope it will conduct more buybacks, but it’s clear that all its contracts are with the foundation, and all revenue flows into the foundation’s bank account—assets controlled by token holders.

So, I think there are many different ways to understand this.

Austin:

Next is the issue of token liquidity. Taking Hyperliquid as an example, theoretically, it has a maximum unlock amount each month, but in practice, only around two or three hundred thousand tokens may be unlocked. Meanwhile, buying volume from ETFs, DAT, and the assistance fund is significantly higher. As a result, there is naturally a situation where buyers outnumber sellers.

Now let’s look at Aerodrome. If you lock AERO to receive veAERO, after their expansion to the Ethereum mainnet in July, veAERO will be renamed sAERO. Holders can not only earn all of the platform’s revenue but also direct token emissions toward liquidity pools that need liquidity the most and generate the highest revenue.

Some might say that if the value of token emissions during a cycle exceeds the value of income, then that cycle is net negative. But I believe this perspective is entirely incorrect.

The correct way to analyze this is: How much revenue did the system generate during this period? How many tokens saw an increase in circulating supply without actually being sold? For example, Aerodrome recently renamed one of its mechanisms to the Momentum Fund, which essentially functions like a continuous buyback program by a foundation. Additionally, many individuals who earn AERO choose to lock and stake it as veAERO to earn additional income. Furthermore, some holders simply have strong confidence in the token’s future and never intended to sell in the first place.

From this perspective, the number of tokens actually flowing into the public market each cycle—i.e., each week—is far smaller than the platform’s revenue generated during the same period.

Combined with recent launches such as Atlas, Aura, and other projects, Aerodrome's revenue has significantly increased. By revenue, I mean the earnings generated for token holders from the platform, which have now clearly surpassed the value of actual emissions outflows.

Therefore, each project and mechanism requires specific analysis. But the core question is: Can token holders benefit from the value being generated by the system? This is the key point of analysis. Based on this, you can delve deeper from this perspective.

Host Andy: I feel the entire industry is shifting toward similar mental models, albeit very nuanced ones. Right now, two types of projects seem to be emerging: one category consists of companies with revenue and solid fundamentals; the other focuses more on narratives, is more specialized, but leverages highly useful technology—such as assets related to AI and privacy like Zcash, Venice, and NEAR. Additionally, there are projects built purely on-chain, while the middle ground currently appears to have little activity.

Austin:

I agree with your perspective. One interesting aspect of this market is that the set of tokens truly worth paying attention to has become smaller. Now, people have a clearer understanding of which projects genuinely have market appeal and which are real, rather than just hype—so perhaps only 10 to 20 tokens currently possess very strong fundamentals.

Therefore, we see these tokens significantly outperforming the market. This is the first time in a long while that investors have been able to choose from a smaller pool of high-quality projects. Currently, capital is flowing concentratedly toward projects like Venice, HYPE, Grass, AERO, NEAR, and Zcash.

Zcash is another project focused on privacy. Some people are now concerned that Bitcoin may increasingly be influenced by Michael Saylor (a separate topic), while Zcash embodies Bitcoin’s original spirit and has a structure very similar to Bitcoin.

Although Zcash currently generates no revenue, it remains an interesting asset—because the higher its price, the greater its real utility. The higher the price, the more likely it is to become entrenched, fostering stronger consensus and community value around it.

So, I believe we’re now in a very interesting phase: it’s becoming easier to choose the right token—just by focusing more carefully on research and distinguishing between legitimate projects and mere hype.

For investors seeking 5x to 10x, or even 3x returns, this is now a more favorable time to succeed than ever before. While you might still end up with 100x returns, I believe there’s currently a group of projects doing something truly compelling—these are the assets I’m watching and investing in.