The next great commodity?

Enron, remembered, of course, for turning electricity into a tradeable commodity, had designs on making lots of other things tradeable, too: weather forecasts, water rights, carbon emissions… even advertising minutes.

But their biggest target was the thing that seemed in shortest supply circa 1999: internet bandwidth.

The CEO of UUNET, then the largest provider of bandwidth at the time, predicted that internet traffic would continue to rise exponentially. “Three years from now,” he said in 1997, “we expect our network to be 1,000 times the size it is today.”

It fell well short of that, unfortunately. But the forecast was directionally correct: demand for bandwidth  — the capacity to move data across the internet — doubled approximately every year from about 1998 to 2002.

Either way, bandwidth looked destined to become one of the world's most valuable resources — and Enron saw an unprecedented opportunity to make it tradeable. 

CEO Jeffery Skilling believed that turning the informal community of bandwidth buyers and sellers into a commoditized market would transform Enron from “The World’s Leading Energy Company” into “The World’s Leading Company.”

(Which really would have been something.)

As Andrew Schwartz wrote in a study of Enron’s plans, the company first tried to make the data packets that move information over the internet tradeable.

That proved infeasible as packets resist nearly every pre-requisite for becoming a commodity: there was no practical way to standardize them, deliver them, or verify their quality.

So Enron chose instead to create a market for bandwidth — the capacity to send those packets over the internet.

In practice, that meant standardizing contracts for transmission: a specified number of bits per second between two cities, say, delivered for a fixed period of time, at a guaranteed level of service.

This is how bandwidth — traded in the form of a futures contract — would ultimately be delivered from seller to buyer: by routing the buyer's data over capacity reserved on the seller's network.

Turning this into a market, however, meant finding a way to deliver that capacity — the bandwidth equivalent of a central clearinghouse to intermediate between buyers and sellers.

Enron's answer was to build a network of switching hubs where capacity purchased from carriers could be routed to the holders of an expiring futures contract.

Routing traffic through these hubs would also allow an auditor to verify that the promised bandwidth was correctly delivered.

(Enron had PwC lined up to do the verifying — perhaps because Arthur Anderson had enough on its plate.) 

Most trades were expected to settle this way — with “physical” delivery — because the buyers wanted the actual bandwidth, not just a cash profit (if the price of bandwidth happened to go up).

Lots of people predicted it wouldn’t work. 

Bandwidth, they said, could not be standardized. The reliability, quality, and service of moving packets over the internet varied too widely across providers.

The biggest sellers wouldn’t participate. Incumbents benefit from opaque pricing, so why expect them to submit to transparent exchange pricing?

The market was constantly in flux. Markets need established players and established rules, and the market for bandwidth had neither.

Physical delivery is too difficult. No one really wanted to incur the expense and bother of connecting to Enron’s new hubs.

Enron's management responded that every new commodity market faces these same objections. And that they knew how to overcome them, just as they had in creating markets for natural gas and electricity.

Now, people are raising all the same objections about compute.

It’s different this time.

If Enron were still with us, they would be all-in on creating a tradeable market for compute — the capacity to process AI computations.

Kalshi founder Tarek Mansour expects that a market in compute derivatives could process as much as $200 trillion of annual trading volume by 2030.

People are again skeptical, for familiar reasons. 

The benchmark people would want to trade keeps changing as new chips constantly replace old ones; sellers prefer opaque contracts to transparent exchange pricing; "delivery" of compute — from the right chip in the right location — cannot be standardized.

But the potential market is too large not to try.

Also, there are reasons to think it could work. In a note for Blockworks, Nick Carpinito explains why compute might successfully be commoditized.

Standardization should emerge over time, he argues: “canonical benchmarks get made through trading.” 

And commodities don’t have to be perfectly fungible: “cross-hedges on far looser relationships already anchor multi-billion-dollar US markets.”

As for the all-important issue of delivery, Carpinito cites ComputeConnect, an “exchange-for-physical (EFP) network”, that aims to settle futures positions by matching an expiring contract with a provider that can deliver the specified compute.

Will this work better than Enron’s interconnection hubs?

It’s true that bandwidth futures never caught on. But Andrew Schwartz believes Enron should get much of the blame for that — primarily because it made itself extremely unpopular.

“The enmity between Enron and other networks supposedly grew to such a point that a conference planned in Europe attracted only Enron,” Schwartz writes. “The other networks canceled when they learned of Enron’s attendance.”

Ultimately, though, bandwidth futures may have failed for a much simpler reason: bandwidth itself became abundant. When the dotcom bubble burst, prices collapsed in an avalanche of oversupply.

Compute has so far followed the opposite trajectory, remaining scarce despite all the datacenters being built.

Prices have even been rising as of late, which makes a market to trade them much more likely to happen, one way or another.

Also, no one will have to trade with Enron.

— Byron Gilliam