Compute Pricing: Four Short Stories

1. A few years ago, a salesperson from one of the major cloud providers offered an “incredible discount” on cloud storage. Obviously there’s some sales tactics here, but I remember running the numbers and the NPV was 3x worse than just holding treasurys.

2. Then, last year I was at Central Computers and a salesperson recommended I buy RAM now because the price was going to go up. I wrote it off and assumed it was a standard sales tactic.

3. And then today, from a vendor: “Right now RAM prices and availability is very unstable and changing everyday. We currently only have 1 set of RAM in-stock right now that is readily available.”

4. And also today, another big vendor quoted me \$91k for 2TB of RAM (32 x 64GB RDIMM, 6400MT/s, Dual Rank memory). For comparison, last year I was quoted \$8k for 1TB (16 x 64GB), so that’s effectively a 5.7x increase in six months.

Is the market wild, inefficient, or something else? And most importantly:

If I have a lot of conviction that the price of RAM will go up/down, how do I monetize that?

I originally thought about this strictly for RAM, but the same logic applies to CPUs, GPUs, and raw compute hours. Here is what I’ve been thinking about regarding the creation of a financial derivatives market for compute. If you are working on this or thinking about this, I’d love to hear from you.

1. Taking Physical Delivery

The most obvious way to express a view on rising hardware prices is simply to buy the hardware now.

Constraints

• Taking delivery means dealing with storage, insurance, and holding costs.
• Hardware has a steep obsolescence curve. Initially, I thought this rapid depreciation was a dealbreaker for a tradable asset. But on second thought, maybe it’s a feature, not a bug. Perishable commodities (like agricultural products) and options contracts both have time decay built into their pricing models. A predictable obsolescence curve could actually create a highly tradeable, dynamic market.

2. Buying or Shorting the Stock (Proxy-Based Hedging)

If you can’t buy the RAM, why not buy the company that makes the RAM?

The Problem: Buying equity is a “dirty” hedge. If I buy Micron or Nvidia stock to bet on memory or GPU prices, my bet is contaminated by corporate variables. What if two major RAM manufacturers merge? What if the CEO resigns? What if there’s an accounting scandal?

The market for a company’s equity is driven by countless factors; the spot price of the hardware they produce is just one input. I don’t want to trade the hypothetical future cash flows, I want exposure to the commodity itself.

3. Asset Standardization and the Fungibility Hurdle

For a derivatives market to function, the underlying asset must be perfectly fungible.

The WTI Comparison: In energy markets, liquidity requires strict standardization (e.g., West Texas Intermediate for crude oil). What is the WTI of compute?

It cannot be a generic “GPU” or “RAM.” It requires a highly specified contract—for example, 10,000 units of 32GB DDR5-6400 ECC memory, or 1,000 hours of bare-metal H100 compute from an ISO-certified basket of suppliers.

One major hurdle is contract duration. A five-year futures contract for crude oil works perfectly; a five-year futures contract for DDR4 RAM would be fundamentally flawed because the demand for the underlying asset will have collapsed by expiration. But realistically, that is something the market should price in and decide.

Maybe traders would likely stick to 3- to 12-month durations.

4. Market Participants: Counterparties and Liquidity

For this market to exist, you need participants willing to take opposite sides of the trade:

• The Hedgers (Shorts): Hardware manufacturers (foundries and chipmakers) who want to lock in future revenue and protect against sudden inventory devaluation.
• The Hedgers (Longs): Hyperscalers (AWS, Azure, GCP), data centers, and AI development firms who need to secure future CapEx budgets and protect against component price spikes.
• The Speculators: Funds and institutional traders who provide market liquidity by taking on the risk the hedgers are trying to offload, betting purely on supply chain data and global demand.

5. Current Proxies and Market Inefficiencies

Has anyone tried this? Yes, but current solutions are highly fragmented and inefficient.

• DRAM Spot Markets (DRAMeXchange): This functions as a clearinghouse and pricing index, but it is largely restricted to massive institutional bulk-buying rather than offering standardized, tradable options or futures contracts for the broader market.
• Cloud Reserved Instances (RIs): AWS, Google Cloud, and Azure offer RIs, which function similarly to forward contracts (paying today to lock in compute prices for 1–3 years). However, these are highly illiquid. You cannot easily short a cloud RI, nor can you actively trade it on a secondary market if the spot price of compute increases.
  
• Hash-Rate Forwards: The cryptocurrency mining sector has actually developed “hashprice” Non-Deliverable Forwards (NDFs). This allows miners to hedge the future value of their compute power in fiat currency. This proves the concept is entirely viable within specific, standardized compute ecosystems.

Some Possible Answers

Your scale isn’t scale

Maybe scale prevents this.

• You’re either AWS/GCP/Azure and there’s someone in your org chart who’s job it is to know where the price of RAM will go, and you’ve already built these risks into your planning cycle,
• or you’re an odd lot and your order size is too small.

No short-term perishability

Futures makes sense for things that are perishable, like coffee beans or oranges, in 2 years from now. They don’t exist yet.

Maybe the way to hedge compute is to just buy it now or some combination of lease/upgrade cycles with OEMs.

Conclusion: curious if others have thought of this

As computing power cements itself as the foundational commodity of the modern economy, the financial instruments used to trade it must mature. The leap from physical stockpiling and vendor lock-in to liquid financial derivatives seems inevitable.

I wonder if this market would solve a real problem for infrastructure builders, or if it’s an exposure that traditional finance actually wants.

If this resonates, or if you know more about the intricacies of this industry, please reach out.