The Final Hurdle in Turning Semiconductors into Oil... What Are "GPU Futures"? [Tech Talk]

Nvidia GPU-Based Futures Product Under Review
Derivatives Market for Semiconductors, a Core Part of AI Infrastructure
Semiconductors Poised to Become Strategic Assets Less Sensitive to Cycles

Editor's NoteWe make the unfamiliar technology that is essential to our lives easier to understand, covering everything from AI and semiconductors to telecommunications and biotechnology.

U.S.-based CME Group, the world's largest futures exchange, is working with Silicon Data, a market data company specializing in artificial intelligence (AI), to launch the world's first graphics processing unit (GPU) futures market. CME Group's GPU futures product is already under review by U.S. financial regulators, and the launch has been postponed until November because of the product's unprecedented complexity and the need for additional review.

Futures are derivatives that trade the right to buy or sell an asset at a specific price. Crude oil, gas, agricultural products and gold are among the assets commonly traded in futures markets. Why is CME Group trying to create futures for semiconductors? And what impact could a GPU futures market have on semiconductor manufacturers such as Samsung Electronics and SK hynix, as well as major tech companies?

Nvidia GPUs: The Embodiment of Cutting-Edge Semiconductors

Graphics processing units (GPUs), the core of NVIDIA's artificial intelligence (AI) semiconductors. NVIDIA

GPUs, particularly Nvidia's GPU cards, are a collection of some of the most advanced semiconductors in the world. The latest Vera Rubin, for example, combines GPUs for AI computing with advanced central processing units (CPUs) for auxiliary tasks, high-performance memory devices, optical communications devices and switches. Buying and selling GPUs is essentially equivalent to trading all the semiconductors needed to run AI.

CME Group and Silicon Data have also drawn on these characteristics of GPUs as they work together to establish a GPU futures market. CME Group already operates the world's largest futures exchanges for grains, commodities and fossil fuels. Silicon Data is a platform that tracks in real time the prices of GPUs traded or leased by data center builders around the world. The plan is to create a futures market using Silicon Data's GPU rental prices as the underlying asset.

Mature Futures Markets Curb Price Volatility

Why, then, is CME Group trying to build a GPU futures market? The answer lies in a longstanding dilemma between semiconductor companies and major tech firms. Semiconductors are a classic cyclical industry: prices surge during boom periods and slump when a downturn sets in. Crude oil and agricultural products also go through recurring booms and downturns, but their price swings are not as extreme as those in semiconductors.

CME Group is the world's largest derivatives exchange. Reuters-Yonhap News Agency.

The reason is that mature futures markets exist for those commodities. Suppose you bought the right to purchase corn at a set price several months from now. Even if an unexpected natural disaster strikes farms and causes spot corn prices to soar, you will still receive a stable supply of grain. In this way, buyers and sellers hedge their risks against each other, helping control price volatility.

If futures markets also become active in the semiconductor market, Samsung Electronics and SK hynix would not have to fear a production race, while major tech companies could ease their price burden.

Complex Pricing and Market Opacity Pose Challenges... "GPU Futures Are Inevitable"

However, what CME Group is seeking to turn into futures is not the physical GPU itself, but "GPU time." GPU time is a pricing model that cloud providers operating data centers established long ago. Cloud companies rent GPUs to customers by the hour. Contracts can range from a few hours to several years. Customers temporarily rent GPUs from cloud companies and log out when their rental period expires.

Unlike oil, corn or gold, which have value as physical resources in themselves, semiconductors are useless unless connected to electricity and operated. As a result, it is expected to be far more difficult to accurately measure the price of the underlying asset in a GPU futures market than in existing futures markets. This is why U.S. regulators have put the brakes on approving GPU futures.

GPU futures are highly complex because they trade the right to buy and sell usage time on GPUs already connected to data centers, rather than the GPUs themselves. The photo shows Microsoft's hyperscale data center. Microsoft

CME Group had already sought to launch a futures product based on Nvidia GPU time, but the U.S. Commodity Futures Trading Commission (CFTC) extended its review period by 45 days on September 22. According to The Information, a tech industry publication, the CFTC is concerned about the "unprecedented complexity" inherent in the structure of GPU futures and has determined that it needs more time to assess potential market risks.

The CFTC also pointed to the opacity unique to the semiconductor market as a concern. The market is dominated by a small number of manufacturers and a handful of major customers. Nvidia already has a dominant position in the GPU market. Markets controlled by a small number of companies face a high risk of price-fixing.

Despite the hurdles, tech entrepreneurs see the emergence of a semiconductor futures market as inevitable. Silicon Data CEO Kamen Lee told the publication, "Additional regulatory review is natural. No matter how long it takes, the fundamental principles of the market will not change." He added, "As more capital flows into AI infrastructure, the need for reliable pricing of computing resources and tools to manage the risk of price fluctuations is growing."

This content was produced with the assistance of AI translation services.