For investors
Contracted payments, secured by hardware.
GPU Capital owns NVIDIA GPU servers and leases them to AI cloud providers under multi-year contracts, or lends against them with a first or second lien. This page explains the asset, how cash moves and what can go wrong.
Investor login is for existing investors and opens Carta, a third-party platform whose terms and privacy policy apply. We will never ask for your Carta password by email or phone.
In brief
- What is held
- Title to leased servers, or a first or second lien on financed ones
- Deployment size
- $1M to $5M
- Server platforms
- NVIDIA HGX B300, HGX B200, HGX H200 and 8-GPU RTX PRO 6000 Server Edition systems
- Underwriting life
- Six years, with lower rental rates assumed at re-lease. Actual life may be shorter.
A familiar asset class, a new asset
Aircraft, railcars and shipping containers are financed the same way. An owner buys a productive asset, an operator runs it, and contracted payments repay the capital while the asset still has resale value. GPU servers fit that pattern.
- Aircraft. Leased to airlines by specialist lessors for decades, so an airline can fly a plane it does not own.
- Railcars. Financed with equipment trust certificates for more than a century, with the railcars as collateral.
- Shipping containers. Leased by the box to shipping lines, so carriers can grow a fleet without buying every container.
- GPU servers. The same structure for a new asset, with a much shorter economic life and faster price declines, so terms are shorter too.
Aircraft. Leased to airlines by specialist lessors for decades, so an airline can fly a plane it does not own.
Railcars. Financed with equipment trust certificates for more than a century, with the railcars as collateral.
Shipping containers. Leased by the box to shipping lines, so carriers can grow a fleet without buying every container.
GPU servers. The same structure for a new asset, with a much shorter economic life and faster price declines, so terms are shorter too.
The difference is time
Large cloud companies depreciate their servers over five to six years1, where an aircraft or a railcar earns for decades. Rental rates also fall each time a new GPU generation ships.
We underwrite a six-year useful life, with lower rental rates assumed at re-lease. Actual life may be shorter.
What carries over
- Title or a lien on an asset that can be repossessed, moved and re-leased.
- Payments contracted for a fixed term with a named lessee or borrower.
- An appraisal of each asset before capital is committed.
- A plan to re-lease or sell the asset if a lessee fails.
What is different
- A useful life measured in years, not decades.
- Rental rates and resale values that can fall quickly when a new generation ships.
- A young market, with only a few years of rental-rate history.
- Lessees that are often young companies with few customers.
Three ways we hold the asset
In each structure a customer contract pays and the servers secure it. What changes is who owns the servers, who is paid first, and who keeps the value at the end.
| Structure | What is held | How cash comes back | At the end of the contract | Main risk to the capital |
|---|---|---|---|---|
| Full-stack ownership | Title. The fund we manage buys the servers and owns them. | After data center costs, the end customer’s prepayment and monthly fees go to the fund until it has recovered what it paid for the servers. Profits after that are shared with the operator. | The servers are re-leased, sold or bought by the operator. | Utilization after the contract, re-leasing, and the servers’ resale value. |
| First-lien loan | A first lien on the servers, and an assignment of the customer contract. | Monthly interest and principal, scheduled to repay inside the customer contract. | The loan is repaid and the lien released. | Borrower default. Recovery depends on equipment values at the time. |
| Second-lien loan | A second lien, behind a senior lender. | Interest and principal from cash left after senior debt service. | The loan is repaid. | The senior lender is paid first. After a default, recovery may be small or nothing. |
After a default, we may repossess the servers and re-lease or sell them. That can be slow and costly, and it may recover much less than is owed: what we recover depends on GPU prices at the time.
Full-stack ownership
How cash moves
The end customer pays the operator under its contract. Data center and power costs come out first. The rest goes to the fund until it has recovered what it paid for the servers. Profits after that are shared with the operator. This is the order of payments, not a forecast of amounts or timing.
The customer pays
End customerSite costs come out
Operator platform- Paid to Data center and power
The fund recovers its cost
The fund’s costRecoveredThen profits are shared
Profit share- Shared with the Fund
- Shared with the Operator
At the end of the term
Servers- Re-leased
- Sold
- Operator buyoutSalvage option
Where it sits in a portfolio
GPU equipment finance is a form of asset-backed private credit. Its return comes from contracted payments and the value of the hardware, not from a market price, so it sits closer to equipment leasing than to growth equity. Investors give up daily liquidity in exchange.
| Item | GPU equipment leases and loans, held privately | AI public equities | Listed data center REITs |
|---|---|---|---|
| What you own | An interest in a private vehicle that owns servers or holds loans secured by them, with the cash flows of their contracts | A share of a company’s future profits | A share of a company that owns data center buildings and leases space in them |
| How return arrives | Monthly contracted payments, then the equipment’s resale or re-lease value at the end | Price changes, realized when you sell, plus any dividends | Dividends paid from rents, plus price changes |
| What drives value | Contracted rates, lessee credit, rates after the contract ends, equipment values | Earnings expectations, valuation multiples, interest rates | Occupancy, rents, interest rates, property values |
| Liquidity | Illiquid. Usually held for the life of the vehicle, with transfers restricted | Daily, on an exchange | Daily, on an exchange |
| Transparency | Periodic reports to investors. No market price, so values are estimates | Public filings and daily prices | Public filings and daily prices |
| Minimum investment | Set in each private vehicle’s offering documents, and usually large | One share | One share |
| Fees | Management fees and carried interest, usually higher than for listed securities | Brokerage costs, or fund fees if held through a fund | Brokerage costs, or fund fees if held through a fund |
| Correlation with equity markets | Not measured. Lessee credit, equipment values and AI demand can be affected by the same conditions that move equity markets. | High. AI companies make up a large share of major stock indices | Listed REITs trade with equity and interest-rate markets |
| Risk of loss | Lessee or borrower default, falling rental rates and equipment values. Investors can lose all of their investment | Prices can fall sharply. Investors can lose most or all of their investment | Property values, tenant credit and interest rates. Investors can lose part or all of their investment |
What can go wrong, and how the structure responds
GPU financing has real risks, and the largest is well documented: each new chip generation pushes older ones down the price curve. Here is how each risk can show up, and what the structure does about it. None of these responses removes the risk.
| Risk | What can go wrong | How the structure responds |
|---|---|---|
| Rental rates fall | Market rates drop as new GPU generations ship, and lessees may try to renegotiate. | The rate is fixed for the contract term. We underwrite a six-year useful life, with lower rental rates assumed at re-lease. Actual life may be shorter. |
| Obsolescence | The servers lose value faster than planned, or become hard to sell. | Contract terms sit inside the hardware’s useful life. Each server is appraised before capital is committed; appraisals by GPU Appraisals, our affiliate, are not independent (see conflicts of interest). |
| Customer or operator default | The end customer or the operator stops paying. | Prepayments and guarantees where available. Title or a lien lets us repossess and redeploy the servers, which can be slow, costly and incomplete. |
| Operator execution | Installation runs late, uptime is weak, or the servers sit idle. | We review the operator’s record and uptime before committing. In full-stack ownership the operator shares in profits only after the owner’s capital is recovered. |
| Power and data center | Power, space or cooling is not ready on time, or the site restricts access to the equipment. | Power and site are confirmed before capital is committed. |
| Recovery after default | Insolvency or a data center’s claims delay repossession. A second lien recovers only after the first. | Tenor inside the customer contract, and reserves where available. No structure guarantees a recovery. |
| Valuation | Assets have no market price, and estimates can differ from what a sale would bring. | Our affiliation with GPU Appraisals is disclosed, and we do not describe its valuations of fund assets as independent. Our offering documents describe how these conflicts are managed. |
| Illiquidity | Capital stays committed for the life of the vehicle. | No structural fix. Fund interests are illiquid; the term and any redemption rights are set out in the offering documents. |
Most of these risks do the most damage late in an asset’s life. That is why we look for deployments whose contracts are scheduled to repay their cost inside the servers’ useful life, and why each lease or loan ends no later than its customer contract. Repayment depends on lessees and borrowers paying as agreed, and it is not investor liquidity, which a fund’s terms govern.
The full summary of principal risks is in Important information.
Questions investors ask
What is a GPU lease?
A contract under which the owner of GPU servers lets an AI cloud provider use them for a set term in exchange for payments. The owner keeps title to the servers throughout.
In full-stack ownership, those payments come from what the operator’s own customer pays. After data center costs, that cash goes to the owner until it has recovered what it paid for the servers. Profits after that are shared with the operator.
What happens if an operator defaults?
The owner or lender can enforce its rights: repossess the servers, re-lease or sell them, and pursue any guarantees. This can be slow and costly, especially if the operator is insolvent or the servers sit in a third-party data center.
What is recovered depends on GPU prices at the time, which can fall quickly when a new generation ships. A second-lien lender is paid only after the first-lien lender.
How is the hardware valued, and by whom?
We appraise each server before we commit capital, using market comparables and an income approach, with a six-year useful life assumed. Our Managing Partner also leads GPU Appraisals, an appraisal practice of Value Prism Consulting, and that experience informs our underwriting.
Because GPU Appraisals is affiliated with GPU Capital, its valuations of equipment we own or finance are not independent. See conflicts of interest.
How correlated is this with AI stocks?
We have not measured it. Contract payments do not move with daily share prices, but lessee credit, equipment values and AI demand can be affected by the same conditions that move equity markets.
A deep AI downturn would likely pressure rental rates, lessee credit and hardware values at the same time as share prices fall. The asset class is also young, with only a few years of rental-rate history.
How long is capital committed to each investment?
Each lease or loan ends no later than the customer contract that pays for it, and inside the hardware’s useful life. We look for deployments whose contracts are scheduled to repay their cost within that time, which depends on lessees and borrowers paying as agreed.
That is not investor liquidity, which a fund’s terms govern. Interests in private funds are generally illiquid for the life of the fund.
Why is the appraisal practice affiliated, and how is that managed?
Marek Omilian leads both GPU Capital and GPU Appraisals. His appraisal work informs how we assess collateral. It is also a conflict of interest: GPU Appraisals’ valuations of equipment we own or finance are not independent, and he may benefit from fees it earns.
We do not describe those valuations as independent. Our offering documents describe how these conflicts are managed, and Important information summarizes them.
How to begin
Access follows a review, in five steps. Each step happens only after the one before it.
Step 1: Request access
Tell us about yourself using the form on this page, including whether you believe you meet the eligibility tests.
Step 2: Review
We read each request and decide whether we can proceed. Submitting a request does not make you eligible to invest.
Step 3: Eligibility questionnaire
If we can proceed, we send a short questionnaire privately, covering your eligibility and your investment experience.
Step 4: Introductory call
A call with our Managing Partner, Marek Omilian, about your background, experience and circumstances.
Step 5: Offering documents
Approved investors receive the offering documents. Any offer is made only through them.
Request access
The form asks who you are, how you would invest, and whether you believe you meet the eligibility tests.
Existing investors
Sign in to Carta for reports and documents.
Investor login (opens in a new tab)Questions about this form: marek@gpu-capital.com
GPU Capital manages private investment vehicles that are open only to eligible investors. We review every request ourselves. If we can proceed, we will send you a short eligibility questionnaire and arrange a call. We share investor materials only after that review.
This form is not an offer to sell, or a solicitation of an offer to buy, any security.
Thank you
We have received your request. If we can proceed, we will contact you about next steps. This does not mean you are eligible to invest, and nothing on this website is an offer of securities.
- Reference
- Received
Your email is ready to send
Your email app should have opened a message to marek@gpu-capital.com with your answers filled in. Send it from there to reach us.
Your message was shortened to fit an email link. Copy your answers to send the full text.
If no email app opened, copy your answers and send them to marek@gpu-capital.com.
Show your answers
Information on this page is as of October 30, 2026, unless another date is given.