Much of an AI venture round is spent on compute. What is compute-for-equity?Much of an AI venture round is spent on compute. What is compute-for-equity?

Key answer. Compute-for-equity is a financing model in which an investor provides computing capacity instead of part of the cash in a round, in exchange for equity. It exists because, in AI, much of the cash raised passes straight through to cloud providers. Where capacity is acquired below the rate the company would otherwise pay, and the company draws it down, the margin on that compute stays with the investor instead of the provider. That is a question about position in the supply chain, not a claim about returns.

Most AI investment commentary is about which model wins. Less of it is about what each round is spent on. With McKinsey estimating $5.2 trillion of data centre investment needed for AI alone by 2030, and with venture funding concentrated in a handful of companies, the question for private investors is not only which company to back but which side of the compute bill they stand on.

How much capital does AI infrastructure need?

About $5.2 trillion of data centre investment by 2030 for AI alone, according to McKinsey's The cost of compute: A $7 trillion race to scale data centers (April 2025). Across all workloads the figure is $6.7 trillion, and global demand for data centre capacity could almost triple by 2030.

The largest technology companies are already spending at that pace. Alphabet, Amazon, Microsoft and Meta plan more than $700 billion of capital expenditure in 2026, nearly double what they spent the year before, according to Sherwood News (April 2026), most of it on AI infrastructure.

The reason is how AI is used. Traditional software was expensive to build and almost free to run. An AI system needs compute once to train and again every time someone uses it, so compute is a running cost of the product, not a one-off cost of building it.

Why is AI venture funding so concentrated?

Because a few companies raise most of the money. AI took 80% of global venture funding in Q1 2026, and four companies, OpenAI, Anthropic, xAI and Waymo, raised $188 billion between them, nearly 65% of all global venture investment in the quarter, according to Crunchbase (April 2026). In the US, 82.7% of 2026 venture deal value to the end of September went to AI, per the Q3 2026 PitchBook-NVCA Venture Monitor.

The far larger group of early-stage companies building on AI competes for what remains, and each of them pays the same compute bill out of a smaller round.

Time is the second pressure. Of 909 active US unicorns, 38.3% had not raised funding in three years by 2026, against 2.6% in 2016, according to PitchBook data reported by Axios (March 2026). Venture funds are typically built around a ten-year life, and a growing share of their holdings is neither raising nor exiting. The same pressure on exit timing in UK private equity is covered in UK private equity exits and portfolio duration, and why the listed market is a shrinking share of the whole in private markets vs public markets.

What is compute-for-equity?

Compute-for-equity is a financing model in which an investor provides computing capacity, usually GPU hours, instead of some or all of the cash in a round, in exchange for equity. The company draws the capacity down directly, so it raises less cash and gives away less of the company for the compute it needs.

The model is in use. Since 2024, venture investors and credit funds have bought or contracted GPU capacity and made it available to the companies they back, in exchange for equity or for terms tied to it. Implementations differ in three working parts:

1. How the compute is priced. Capacity is valued at the investor's acquisition cost, at a market rate, or somewhere between. That price sets how much equity the compute buys.
2. Who supplies it. The investor owns or leases the hardware, or contracts it from a specialist cloud provider.
3. How the position is held. Directly as shares, through a fund, or in some newer vehicles as certificates that can be traded on a secondary platform.

Why does compute ownership matter for investors?

Because in a cash-only AI round, part of the investor's money becomes a cloud provider's revenue. The startup raises cash, pays it out for compute at market rates, and the margin on that compute stays with the provider. The more than $700 billion the largest technology companies are spending in 2026 is, in part, built to earn that margin.

Compute-for-equity changes which side of that bill the investor stands on. Three things follow:

• The entry price changes. If capacity is acquired below the rate the startup would otherwise pay, the same equity costs the investor less in real resources than it would in cash.
• Founders dilute less. A company that needs less cash for compute raises a smaller round and gives away a smaller share for the same capacity.
• The investor holds two exposures. One to the company's equity, and one to the value of compute itself, which can rise with demand or fall as newer hardware arrives.

This is the part of the AI boom that rarely appears on a pitch deck. It does not make any single company a better investment. It changes what an investor is paying with, and that is a question worth asking of any AI round.

What are the risks of compute-for-equity?

The model moves risk rather than removing it. The ones investors weigh:

• Valuation risk. The discount on capacity is only as good as the market price it is measured against, and GPU rental prices move.
• Obsolescence risk. Chips and data centre capacity lose value as newer hardware arrives, so compute held for later use can be worth less when drawn down.
• Utilisation risk. A company that uses less capacity than planned, or needs a different kind, may not get the saving the terms assumed.
• Provider risk. The model depends on the infrastructure provider delivering capacity as agreed.
• Liquidity risk. Secondary trading depends on buyers being there. A certificate can be tradeable in principle and still hard to sell.
• Early-stage risk. The underlying companies are early-stage, and many will fail.

Capital is at risk and investors may not recover the full amount invested. Vehicles of this kind are generally open to professional and qualified investors only. Reading the documents and taking independent legal and financial advice before committing is the investor's responsibility. HBRD Investments is an introducer, not an adviser, and is not authorised or regulated by the FCA. How we introduce opportunities is set out on How We Operate; for how venture and growth exposure fits alongside other private market routes, see Private Equity and Pre-IPO.