The AI build-out is powering global goods trade

Welcome to Trade Secrets. I’m Peter Foster, the FT’s world trade editor, filling in for Alan who is taking a well-deserved break. I was in Yaoundé, Cameroon for the World Trade Organization ministerial meeting (of which a few quick on-the-ground impressions later). But first I wanted to return to something that didn’t quite get the coverage it deserved before the debacle in Cameroon: the outsized role of the AI build-out in driving growth in goods trade.

This was a feature of the WTO’s 2026 Global Trade Outlook that was published ahead of the ministerial but got crowded out of the news agenda by the Iran war. In Charted Waters, where we look at the data behind world trade, we have a little Trade Secrets exclusive, with new econometric research from Nigel Driffield, Jun Du and co-authors at Aston University that quantifies the extent to which Brexit cost the UK inward investment. 

AI-related investment surge

For most of us, AI is an abstraction: a black box of predictive algorithms that we cannot fully understand. In so far as it impinges on the material world, we tend to focus on the threat to white-collar jobs and the grid-busting energy requirements of data centres.

But back in the real world, the AI build-out is feeding into the data at a speed that is catching trade economists unawares. That much was clear from the WTO’s revised estimates for growth in goods trade in the 2026 outlook. In October, the WTO was forecasting world merchandise trade would grow at 2.4 per cent in 2025. But in March, it revised that number up to 4.6 per cent, citing the AI boom. That’s a big fat undershoot, which speaks to the speed at which the build-out is occurring.

The WTO estimated that the surge of AI-related investment made up almost half of goods trade growth in 2025, with the US taking the lion’s share. That is unsurprising as North America accounts for more than half of the world’s data centre build-out, according to the McKinsey Global Institute.  

AI investment is also taking an outsized role in the US growth story, the WTO found, accounting for about 70 per cent of total investment growth in North America over the first three quarters of 2025. For context, during the housing boom of 2004-07, residential investment accounted for an average of 30 per cent of investment growth. Moreover, today’s AI boom is helping to mask the negative impact of President Donald Trump’s tariffs, both in global trade and the US economy.

Another key difference between the housing boom that led to the subprime crisis and the current build-out is that construction has a very low reliance on imports (less than 2 per cent), while research suggests AI investment has an “import intensity” of 70-90 per cent. In short, the future technology on which the US is betting the house is massively dependent on global trade, even as Trump throws handfuls of grit into the gears of the world trading system. 

As Jeongmin Seong at the McKinsey Global Institute puts it, “the AI value chain is a global one. No one country can own the whole end-to-end basis of a technology that is so critical to driving productivity.” The remarkable result of the boom, combined with White House tariff policies directed against Beijing, is that the US now imports more from Taiwan than it does from China. (The majority of China’s build-out, on the other hand, is fuelled by domestically produced kit as a result of chip export controls. John Miller at Trade Data Monitor delves further into China’s role in the AI boom here.)

Overall, McKinsey calculates that US trade of “AI-related” goods rose in 2025 by roughly 66 per cent, or an estimated $220bn. Seong adds that the seven big tech companies earmarked $743bn for research and development and capital expenditure — compared with £396bn in 2023 — and that South Korea’s current semiconductor capacity is now sold out for the next two to three years. That is to say, whatever the question marks over the long-term scalability of the boom, for now it is real enough. It can be measured in overflowing order books and stacks of 20ft containers.

A side-note here for trade nerds. “AI-related” is in quotation marks above because the global trade taxonomy hasn’t yet caught up with the AI revolution, so there is no agreed benchmark for what constitutes an “AI-related” or “AI-enabling” good. There is also a risk of some double-counting, since not all chips and semiconductors end up in data centres. 

The WTO uses a much broader definition, for example, than McKinsey does in its chart above, including everything from cooling pumps to specialist tracksuits used in data centres for their list of HS6 product identification codes. (You can find the list in its September report here.) McKinsey uses a much narrower chip-focused set of eight HS6 codes — with 847150, 847330, 851762 accounting for the majority of the total, according to MGI senior fellow Tiago Devesa. The US Federal Reserve and World Bank again use different subsets. 

Regardless, the varying measures show the same trend of strong growth. But the specific definition makes a big difference to any assessment of the size of the AI-related merchandise pie, and therefore the potential consequences of a slowdown. With its narrower measure, McKinsey reckons AI-related trade accounts for 5.6 per cent of total trade (but a much bigger slice of goods trade growth) while the WTO, with its broader measure, puts the share at nearly 17 per cent.

WTO struggles to stay credible

Much has been written about the WTO’s blowout at the Cameroon ministerial, which held up a mirror to a fracturing world. A more fragmentary system of plurilateral trade governance is now emerging. It’s not ideal but, as one glass-half-full middle power diplomat put it, “a patchwork of processes and rules that deliver a measure of certainty is better than standing on the sidelines”. Coalitions of the willing must do what they can.

Whether the WTO can be at the heart of this new plurilateral world is another question. Smaller coalitions are already gathering steam, such as the emerging EU-CPTPP grouping, that announced a plan for a digital deal of its own during the meeting. I defer, naturally, to those who’ve attended five or six ministerials, but as a newcomer it was hard to escape the feeling in Yaoundé that the WTO was struggling to stay credible. 

The scathing op-ed last week from US trade representative Jamieson Greer was a bit rich coming from a country that has done so much to undermine the WTO. But he did put his finger on the atmosphere in Cameroon. His withering description of delegates bopping along as the WTO was “playing a self-congratulatory song about progress on an incomplete agreement on fisheries subsidies” really wasn’t far from the truth. 

That was amplified by the “we just ran out of time” line from the director-general at the closing press conference and the final press release headlined: “MC14 concludes with adopted decisions, progress on key outstanding issues”. There’s a line between putting a brave face on things and utter complacency. For me, the director-general was on the wrong side of it. 

I think that’s ultimately why I struggled to keep believing when reading this excellent Global Trade Alert piece by Simon Evenett and co-authors that set out how the WTO could become a forum for middle powers avoiding a race to the bottom during the current energy crisis provoked by the Iran war.

They suggest co-ordinated systems for notifying and justifying export controls, not out of “charity” but collective self-interest. They cite a now expanded New Zealand-Singapore Covid-era declaration on medical supplies as a template of what’s possible during a crisis period.

Evenett isn’t naive about the state of the WTO. The piece, he tells me, is designed to encourage small and midsized players to seize the opportunity to salvage what they can. “Those states that say the WTO is vital need to tell us what strategy they have to protect it — rather than lamenting every retrograde step by the big guys,” he said. 

Never say never and all that, but it was just hard to be present in Cameroon and still believe Geneva was the place where this could still happen. 

Charted waters

The UK government has recently taken to citing the upper estimates of Brexit damage when making the case for a closer relationship with Europe. Chancellor Rachel Reeves recently cited a firm-level NBER study that puts the hit at 8 per cent of GDP. It’s an interesting move, since it throws quite the spotlight on the paucity of Labour’s “reset” with Brussels, which is worth only 0.24 per cent of GDP upside, according to the Office for Budget Responsibility.

Yet further ammunition for those advocating a bolder approach to repairing the damage done by Brexit is provided by a new joint study from the Productivity Institute and Aston University, which finds that leaving the EU redirected approximately £22bn of UK greenfield investment and more than 100,000 jobs to the bloc between 2016 and 2019. And former prime minister Boris Johnson’s bare-bones Trade & Cooperation Agreement (TCA), when it came into force, only partially arrested the slide.

The research points to divergent regulation as the key driver of decisions to forgo investment in Britain. As Driffield observes, the UK may be the sixth-largest economy in the world, “but capital intensive production cannot reach an efficient scale just selling in the UK”. Or as Du put it: “The TCA addressed tariffs, but tariffs weren’t the problem. Regulatory friction was, and still is. The investment hasn’t come back.” 

Trade links

• A year on from “liberation day”, libertarian think-tank the Cato Institute has run the rule over the impacts of Trump’s trade policies. It’s all worth reading but the short version runs: “Exemptions proliferated; lobbying skyrocketed; and most Americans were worse off.” Bravo.

• The IMF has published a techy paper on global imbalances that warns of disorderly unwinding if steps are not taken to address them. It lays out a plan to raise Chinese consumption and cut US fiscal deficits. That was easy. 

• Former chief economist at the Bank of England Andy Haldane argues in the FT that the integrated nature of global supply chains means that the ongoing rupture of the world order is to the “Achilles, not the aorta — painful, not fatal”.

• The OECD reports on a historic decline in overseas development aid, which is down by nearly a quarter in 2025. That takes it back to 2016 levels. Five of the world’s richest countries — Germany, US, UK, Japan and France — account for 95 per cent of the decline. That’s both short-sighted and shameful.

• Lori Heinel of State Street Investment Management has posited five “grey swans” for 2026 in a fun piece for the Official Monetary and Financial Institutions Forum think-tank. These are an AI slowdown; a China consumption pivot; US-Mexico-Canada (USMCA) negotiations generating a North American industrial renaissance; the new Japanese government’s policy agenda unwinding the yen carry trade; and a global bond shock.

• USTR has published its annual “National Trade Estimate”, a 534-page list of gobbets setting out the state of play (and plenty of gripes) with the US’s trade partners. You can download the whole thing here. Ctrl+F is your friend.

Trade Secrets is edited by Harvey Nriapia