Artificial Intelligence

Commodity traders bet on big data and AI

The push to harness the latest technological tools is partly a reaction to competition from hedge funds and other data-led trading teams, which move fewer physical commodities but have built lucrative businesses trading commodity-linked securities and other financial products.

The most advanced data-led trading operation in the sector is arguably found at the Miami-based hedge fund Citadel, which hired commodity trader Sebastian Barrack from Macquarie in 2017 to lead a bigger push into energy and raw materials.

In the hunt for an informational advantage, one of Mr Barrack’s first moves was to hire a 20-strong team of weather forecasters. The broader commodity trading team has since swelled to more than 300 people including analysts and engineers.

Oil and refined products is one area where the available data on supply levels, demand patterns and logistical variables has ballooned in recent years, Mr Barrack told the Financial Times. “The prolific growth of the data that’s becoming available to us is literally making us better-informed investors.”

Electric advantage

For data-led trading strategies the energy transition promises to be a boon because it will increase complexity and require more sophisticated tools to model markets, particularly when there is a lack of historical information in a new area, he added.

“The more detail, the more complexity, the more there is a lack of backward-looking data, the better for us.”

Citadel made a record $US16 billion in 2022 to displace Bridgewater as the most successful hedge fund of all time, according to research by LCH Investments. About half of that came from commodities as the firm, like other traders, profited from the extreme volatility in energy markets following Russia’s invasion of Ukraine, the FT has reported.

Citadel declined to disclose its 2023 performance.

The ability to process large volumes of data is particularly important in the fast-growing area of power trading, where the regulated nature of electricity markets produces large amounts of information.

Consultants McKinsey estimate that data-driven trading firms captured a quarter of gas and power trading profits globally in 2022, up from less than 5 per cent in 2021. That competition has forced traditional commodity traders such as Trafigura, which made a record $US7.4 billion ($11.5 billion) in 2023, to invest to keep up. It set up a power trading division three years ago.

Richard Holtum, head of gas, power and renewables at Trafigura, said his team “uploads several billion discrete bits of data into the cloud” every day. “The challenge there is using AI to interrogate that data in a better, more efficient way to improve the trading decisions that we are then making,” he said. “I think right now we are at the very tip of the iceberg on what AI can do.”

Switzerland-based Mercuria, founded in 2004 by Marco Dunand and Daniel Jaeggi, was predominantly an oil trader but bolstered its power trading operations in 2014 through the acquisition of part of JPMorgan’s physical commodities business.

Mr Dunand told the FT that the informational advantage accumulated by data-led players such as Citadel was helping them take larger positions in the market, but that AI could help Mercuria bridge that gap.

“If you wanted to mirror Citadel as an example for collecting data, I think it would take a lot of time, a lot of money … so we are spending a lot of time and effort trying to develop our own AI machines in order to kind of close those gaps,” he said. Mercuria made about $US2.7 billion in 2023, he added, slightly down from the record $US3 billion set the previous year.

However, the physical traders are not about to stop getting their hands dirty.

“You can actually be a player in the market without trading physical, but that’s not for us,” Mr Dunand said. “I think ultimately, the world needs energy, and we’re energy traders, so if you don’t move this stuff, you know the world doesn’t work.”

Financial Times


This website uses cookies. By continuing to use this site, you accept our use of cookies.