A16z's latest investment aims to set a "price" for computing power

The story of AI has been told for a long time, with the protagonists changing one after another.

First came large models, then AI agents, followed by data centers, electricity, and so on.

With each wave of hot money flooding in, capital is always looking for that "yet-to-be-priced" corner.

Now, it's computing power's turn.

On June 24, Ornn, a company founded just a year ago, announced the completion of a $33 million seed round, led by a16z Crypto, with participation from Galaxy Ventures, Nordstar, SV Angel, and follow-on investments from existing investors Vine Ventures, Crucible Capital, Link Ventures, and Box Group.

A $33 million seed round—the number alone isn't particularly eye-catching. What's more worth pondering is another thing: why did a16z invest?

The answer lies in what Ornn intends to do. It doesn't plan to sell computing power; it wants to set a "price" for computing power.

The ultimate end of computing power is, of course, electricity. But before reaching that end, it might first need an "oil price" of its own.

Computing power doesn't yet have its own "oil price"

To understand Ornn's positioning, we first need to look at what the computing power market looks like today.

The answer is straightforward: this is a market with a massive spot market. Buying and selling GPU computing power today still largely operates on a "one deal, one price" basis. Buyers and sellers negotiate a price privately, and the next deal with a different counterparty will have a different price. The entire market lacks a recognized benchmark for "what the price is right now," and forward curves and hedging tools are relatively distant.

What does this mean?

It's like the world burning oil, yet there is no WTI or Brent. The price of each barrel of oil depends entirely on on-the-spot haggling.

Putting the key information of this funding round together reveals: this is not just another AI project. The two MIT graduates are not building a GPU cloud but rather the "foundation" of a financial market.

This product framework consists of three parts: a price index (OCPI), a derivatives entry point, and a platform called Ornn Compute that revitalizes physical GPU capacity. Ornn's three products precisely target these gaps.

First, OCPI, the "oil price" for computing power. It is a computing power index pieced together from actual transactions, not listed quotes, covering mainstream models such as H100, A100, H200, B200, and RTX 5090. It is reportedly already running for half a year, listed on the Bloomberg Terminal, and used by over 400 operators, investors, and AI companies. ICE's announcement further validates Ornn's direction. ICE stated that it will launch a series of GPU compute futures based on OCPI with Ornn, denominated in USD and cash-settled.

Second, futures and options, equipping computing power with hedging tools. As early as May, Ornn announced a partnership with the Intercontinental Exchange (ICE) to launch USD-denominated, cash-settled GPU computing power futures using OCPI as the reference rate. AI companies can go long to lock in future training costs; data centers can go short to lock in future revenue.

Third, Ornn Compute, turning idle capacity into transferable assets. It aggregates dedicated GPU capacity from multiple neoclouds onto a single platform, with unified access criteria, supporting secondary transfers and on-demand subleasing. Operators can connect to diverse tenants with a single offtake contract, while buyers can see the cluster's location, hardware, and terms. Capacity previously locked in long-term contracts with low liquidity is transformed into something akin to "transferable warehouse receipts."

This is more like the "computing power version" of ICE/Platts, aiming to earn a "toll fee" from the entire market.

The core pain point it targets is clear: current computing power contracts are almost entirely one-off negotiations between buyers and sellers, with highly dispersed prices. Operators have almost no visibility into utilization rates and resale value after 3–5 year long-term contracts expire, yet they bear the duration risk alone. This is a market with a massive spot size but a nearly blank financial layer.

The chain logic provided by Ornn is also the logic reiterated by a16z:

Reliable pricing data → Price discovery → Risk transfer → More efficient capacity allocation.

A16z's Bet

Looking back at a16z's choice, the logic becomes much clearer.

A16z's previous investment thesis stated it plainly: we are in the "largest capital formation event of our lifetime." Over the next decade, trillions of dollars will flow into GPUs, data centers, and electricity. This buildout will rank among the largest industrial mobilizations in human history.

The money is already on its way, but the market structure to accommodate these funds has not yet been built.

Mature commodity markets often develop in a relatively fixed order: first comes reliable price data, then price discovery; with price discovery, risk can be transferred; with risk transfer, efficient capacity allocation becomes possible. Ornn's products are each wedged into a gap in this chain.

If we break it down by financial market structure, Ornn's target can probably be divided into four layers:

Ornn aims to become the benchmark price layer, risk transfer layer, and capacity trading layer for compute.

This is the mismatch a16z saw, and the fundamental reason for its early bet.

First, the spot market is already massive, but the financial layer is almost blank. a16z is not investing in a cloud service; it is installing the underlying pipeline of "price discovery + risk transfer" for this emerging commodity. This is the classic "seller of shovels among sellers of shovels": while others are scrambling to sell GPUs, Ornn is selling the rules that allow GPUs to be priced and traded.

Second, the collaboration with ICE is a key signal of "verified first, invested later." The ICE futures partnership was announced in May, before this seed round was publicized. This means that by the time a16z placed its bet, Ornn had likely already secured the endorsement and clearing channel of a top global commodity exchange. In other words, the "settleability" of the product had already been accepted by the traditional financial system, eliminating a large part of the risk before the bet. a16z was likely waiting for this moment when the "traditional financial interface" was proven.

Third, seizing the position of "standard setter." There is an iron law in commodity markets: benchmark prices tend to be winner-takes-all. Crude oil looks to WTI and Brent, interest rates to LIBOR and SOFR. A dominant benchmark usually has only one. Whose index is adopted by exchanges as a clearing reference locks in the toll booth for the entire track. a16z does not want just any player; it wants the "benchmark price monopolist" of the computing power era. Such a window for positioning often exists only in the earliest stage of a track.

Where is the Real Intent?

AI computing power is already ripe for commoditization: huge demand, fragmented supply, opaque prices, and unhedgeable balance sheet risks. Ornn's ultimate goal is to turn GPU-hours into a financial underlying asset similar to crude oil, electricity, interest rates, or mortgage assets.

This is a16z's bet on "computing power transitioning from a cloud service resource to a pricable, tradable, and financeable commodity." This calculation is layered, with each level being a larger strategy.

First, capturing the pricing power of computing power. If OCPI truly becomes the WTI of the computing power world, then all future contracts, financing, and derivatives referencing it will have to go through Ornn. This business is higher-level than operating a GPU cloud. Clouds earn capacity margins; indices and settlement benchmarks earn the "infrastructure tax" on the entire market.

Second, turning AI computing power into a brand new asset class. With a benchmark price and derivatives, dedicated GPU capacity transforms from a "heavy asset burden" into a "liquid asset": idle capacity can be revitalized, and long-duration risks can be hedged. Around it, an entire ecosystem of lending, structuring, market-making, etc., may emerge. a16z aims to be the "founding shareholder" of this new asset class.

Third, raising a flag for the fund itself. Using a crypto fund to invest in a computing power market running on traditional exchanges like ICE is in itself a declaration: the financial engineering capabilities distilled from Web3 are spilling over into the much larger physical market of AI.

Ornn is cutting into the layer below: the financial market infrastructure: pricing, clearing, hedging, secondary liquidity of capacity, and collateral valuation. The financial primitives Web3 is best at—pricing, liquidity, risk transfer—are being moved from digital assets to the physical computing power of AI.

Following this line of thought, new forms become imaginable: on-chain perpetuals for the computing power index, RWAization of GPU capacity, stablecoin settlement for neocloud cash flows, and DePIN networks directly referencing OCPI quotes.

In other words, the most critical means of production for AI is beginning to be recoded by capital markets.

How to Find an Anchor?

If we look for an analogy in history as an anchor, this investment most resembles not some AI infrastructure financing, but a16z's Series A lead in Uniswap in 2020.

In August 2020, a16z led an $11 million investment. At the time, the market was still debating whether AMMs and DEXs were just niche toys, but capital had already sniffed the signal: on-chain trading would grow into an independent financial market structure. The rest, as they say, is history. Uniswap became a major part of the DeFi trading layer.

One could also reference a16z's investment in OpenSea ($23 million) in 2021. OpenSea anchored itself as the entry point for the NFT trading market cycle. What they truly did was redefine "how the market itself forms."

The commonality lies here: betting on the person who "defines the rules" when a track is just emerging and the infrastructure is still blank.

A16z's lead is not a bet on the GPU cloud business, but on a more fundamental trend: AI infrastructure will eventually develop its own price benchmarks, derivatives markets, and capital allocation systems, just like energy, interest rates, mortgages, and crypto assets.