AI in crypto

AI Fund

A concentrated thesis on where artificial intelligence needs open networks.

A thematic portfolio in the digital assets that connect artificial intelligence with open networks: decentralized compute, training data, agent payments and model verification. It is an early and very volatile sector. That is why every position starts with a written thesis, carries a size cap and is sold when the thesis breaks. It is designed as a bounded complement to your portfolio, never its core.

Horizon: Multi-year, 3 years or more
Risk: 5 / 5

Talk about this fund See terms

Horizon · Multi-year, 3 years or more

01 / The opportunity

Why this fund exists.

A high-risk thematic satellite on top of a base already built in Bitcoin, Ethereum or the Innovation Fund. Its weight should be what you could lose without it changing your life.

Open compute against scarcity

Demand for GPU capacity to train and run models is growing faster than supply, and that supply is concentrated in a handful of cloud providers. Decentralized compute networks aggregate GPUs from data centers and independent operators, open them to anyone and settle payments on-chain. They do not replace the large clusters used to train frontier models; their market is inference, fine-tuning and access for those who cannot obtain capacity at a reasonable price today.

Agents that pay for themselves

An AI agent that buys data, API calls or minutes of compute needs to pay small amounts, instantly and without a person approving every transaction. Cards and bank accounts were not designed for that; stablecoins on public networks settle in seconds, around the clock and under programmable rules. In July 2026 the Linux Foundation launched an open body to govern a payments standard over HTTP, built on the 402 Payment Required status code, designed for AI agents and supporting stablecoins, backed by 40 organizations from payments, cloud and finance (Linux Foundation, July 2026). The networks where that machine-to-machine commerce settles may capture part of its value.

Verifying what a model does

As models make decisions with economic consequences, it matters to be able to prove which model ran, on which data, and that its output was not altered. Zero-knowledge proofs and verifiable inference make it possible to check a computation without repeating it or trusting whoever ran it. Along the same lines, recording the origin of content on-chain helps distinguish what people created from what is synthetic. The technology is early and costly today, but it addresses a real problem that centralized AI does not solve on its own.

Owned data and open markets

Models depend on data that today is almost always extracted without payment or traceability. Decentralized storage networks and data markets let those who contribute information get paid and keep a record of how it is used, while open model marketplaces coordinate who trains, evaluates and serves each model, with rewards per contribution. The thesis is that part of the AI value chain will be built on open rails where ownership and payment are programmable. Which designs will prevail is not yet known, which is why the fund spreads its exposure across subsectors.

02 / How we manage it

Written rules, not impulses.

01
A written thesis for every position
Before buying, every position has in writing which subsector it belongs to, which usage metric confirms it, what would invalidate it and when to exit. If a position depends only on the AI narrative rather than measurable usage, it does not qualify.
02
Real usage before the label
Only assets with verifiable usage qualify, such as compute sold, data stored or payments processed, with revenue paid by real users and not just incentives. We assess whether the token is needed for the network to operate or only serves speculation, public code and team activity, the unlock schedule, and the liquidity to exit in a stressed market. Tokens that only carry the AI label, memecoins and presales are excluded.
03
Limits per position and per subsector
Every position has a strict size cap, and each subsector, whether compute, data, agents and payments or verification, has its own concentration limit. More volatile positions carry less weight. The fund uses no leverage and no speculative derivatives.
04
Frequent review, exit on thesis
In a sector that moves this fast, every thesis is reviewed more often than in the funds built on established assets. We exit when the thesis breaks, not when price hurts. Custody is segregated under your contract, with periodic position reporting.

03 / Portfolio

What goes in and what doesn’t.

Includes

Liquid assets of AI compute, data and agent networks with verifiable usage
Infrastructure connecting AI models with on-chain payments and data
Operational USD/USDC for rebalancing and tactical entries

Excludes

Tokens that use "AI" only as a narrative, with no product or usage
Memecoins, unaudited presales and leverage
Positions without enough liquidity to exit in stressed markets

04 / Investor profile

Who it is for. And who it is not for.

A fit if

Investors with crypto experience who want targeted exposure to the convergence of artificial intelligence and open networks, and accept very high volatility.
Those who already hold a base in Bitcoin, Ethereum or the Innovation Fund and want to add a small fraction with a thematic thesis.
Those who follow artificial intelligence closely and prefer a documented process over picking tokens on their own amid the noise.

Not a fit if

First-time crypto investors, or anyone without a base in more established assets yet.
Anyone putting savings they will need in the coming years, borrowed money or a large share of their wealth into this fund.
Anyone seeking exposure to listed artificial intelligence companies: this fund invests in digital assets, not in shares.

05 / Terms

Clear from day one.

These are the fund’s reference terms. The final ones are set in your private contract, which we review with you point by point before signing.

Minimum investment: USD 25,000Per investor, under contract
Management fee: 2% per yearOn assets under management
Performance fee: 20%Only on net gains above the previous peak (high-water mark)
Minimum term: 12 monthsQuarterly redemptions after the term
Redemption notice: 90-day noticeBefore each quarterly window
Benchmark: CoinGecko AI category market capFor comparison, not a return target

Segregated custody under contract, with periodic position reporting. Daleki Capital is not a regulated financial institution; it operates through private contracts.

06 / Risks

What can go wrong.

Narrative cycles and sharp drawdowns. Assets in this sector tend to move more on market sentiment toward artificial intelligence than on their own usage. They can rise sharply within weeks and fall drastically just as fast. Individual positions can lose all of their value, and the fund as a whole can lose a very large part of its capital.
Competition from the AI giants. Large technology companies control the most advanced models, chips and data centers, with capital no decentralized network can match. They can offer cheaper or better solutions, including their own payment rails for agents, and leave open networks in a marginal role.
Technology and execution risk. Much of this technology, such as distributed training or verifiable inference, is experimental, costly and may not scale. Teams may fail to deliver their roadmap, and smart contracts can contain bugs or be exploited.
Token design, dilution and liquidity. Many tokens in the sector have a large share of supply still to unlock for teams and early investors, or issue new tokens to attract providers, which dilutes existing holders. A network can succeed without its token capturing that value. On top of that, liquidity in these assets can vanish exactly during a sharp decline.
Regulation, concentration and vehicle liquidity. Rules on both artificial intelligence and digital assets are still changing, and a single regulatory decision can hit an entire category. Because it is concentrated in one theme, the fund has no diversification to shield it from a sector-wide decline. Your interest can only be withdrawn in the windows and periods set by the contract.

07 / FAQ

What investors ask us.

Why a fund dedicated only to AI and crypto?

Because it is a theme with its own dynamics and higher risk than the other ecosystems. Separating it from the Innovation Fund lets you decide how much exposure you want to this thesis and measure it on its own.

What assets does it invest in?

Liquid digital assets in decentralized compute, data and storage, agent payments, verifiable inference, model marketplaces and content provenance, plus operating cash. Specific positions are detailed in investor reports, not in public materials.

Does it invest in shares of AI companies?

No. The fund invests only in digital assets. Exposure to listed artificial intelligence companies belongs in a different kind of vehicle.

How do you avoid tokens that only carry the AI label?

Through filters on measurable usage, revenue from real users, token utility and code activity. The filters reduce that risk but do not remove it: there will be mistakes, which is why every position has a size cap.

When can I withdraw my capital?

According to the holding period and redemption windows shown in this profile’s terms table and in your contract. The illiquidity of some assets can affect the value at the time of redemption.

08 / Next step

Let’s talk about the AI Fund.

A manager reviews your goal, your horizon and whether this fund fits. No commitment and no pressure.

Other funds

Bitcoin FundDigital store of value Ethereum FundOpen financial infrastructure Innovation FundActive, thesis-driven management

Private access

AI Fund

Open compute against scarcity

Agents that pay for themselves

Verifying what a model does

Owned data and open markets

A written thesis for every position

Real usage before the label

Limits per position and per subsector

Frequent review, exit on thesis

What investors ask us.

Let’s talk about the AI Fund.

Book a conversation