EBRD: Its regions ship 0.8% of AI exports but see a 1.47 euro defence payoff
The EBRD's new Transition Report says its regions supply 0.8% of global AI-related exports and are heavily dependent on imported defence goods, so the gains are not automatic.

The European Bank for Reconstruction and Development (EBRD) says the economies it works in supply just 0.8% of global exports of AI-related products, and that the AI and defence booms will not lift them automatically. The finding comes from its Transition Report 2026-27, as covered by Reuters, and the bank published it on 17 November 2026.
Reuters headlined its story as a warning to developing economies. The EBRD’s own material speaks of the “EBRD regions”, the group of economies where the bank invests, and does not use the label developing economies. The report is titled “Frontiers of competitiveness”.

A thin slice of AI trade, a larger slice of goods
The EBRD says its regions account for about 8% of total global goods exports but only 0.8% of exports of AI-related products. According to the bank, the gap reflects a limited presence in upstream segments such as advanced chips, data infrastructure and frontier AI models. Power constraints also limit how many large data centers these economies can host.
The report sees the main opening elsewhere: AI applications and adoption. Under a moderate adoption scenario, the bank estimates AI could add up to 0.7 percentage points to annual labor productivity growth across its regions over the coming decade. Getting there, it says, requires digital infrastructure, skills, computing capacity, data governance and international cooperation. It also notes that the capacity to adopt varies, so the benefits are likely to be uneven.
Defence spending helps, but mostly goes abroad
On defence, the report finds that, on average, a 1 euro increase in defence spending is associated with a later increase in GDP of 1.47 euros. The bank describes this as an association, not proof of cause. It also says higher spending can raise debt and draw money away from welfare budgets.
The larger problem is who supplies the goods. The EBRD says most economies in its regions remain highly dependent on imported defence goods. It cites limited demand, state ownership and weak access to capital as constraints on domestic industry. Economies with strong industrial bases could expand output through retooling, joint procurement and licensing of foreign technology, according to the report.
EBRD Chief Economist Beata Javorcik said that “Geopolitical fragmentation, rising defence needs and advances in AI are creating new opportunities and challenges”. The bank says rising tensions have ended the post-Cold War peace dividend.
Why complex exports are hard to start
The analysis rests on how export baskets change. In a typical EBRD economy, about 35% of the value of the exports in which it specializes comes from products that were not widely exported 25 years ago. The executive summary adds that roughly half of the products with a comparative advantage today differ from those of 25 years ago. So change is possible, but slow.
The bank argues that complex products such as AI and military goods are harder to enter. New specializations tend to succeed where related upstream or downstream capabilities already exist, and large exporters play a key role through spillovers to other firms. That is a cautious reading for countries hoping to jump straight into chips or weapons.
The EBRD’s earlier work points the same way. A June 2026 edition of its Regional Economic Prospects, as summarized in search results, said exports tied to AI supply chains were growing faster than other exports across the regions, so the supply-chain route is not closed. Reading the two reports together, the opening is real but narrow: supplying parts of a chain rather than owning it.
The money gap behind the policy advice
The report says founders in these economies are often internationally experienced and build products outside established export baskets, but few scale up. Venture capital is thin and concentrated geographically, and later rounds come largely from abroad, particularly the United States, the United Kingdom and other advanced European economies. Javorcik said the founders “have the entrepreneurs and ideas; what they need are the financing and investment networks”.
The recommendations stress skills, research infrastructure, finance and supply-chain integration. The bank favors guarantees, co-investment and deeper domestic and regional capital markets over subsidies to individual firms.
Numbers in context: A race priced in gigawatts
The 0.8% figure sits against a very different picture at the top. Washington is reportedly preparing to name Nvidia, OpenAI and Anthropic among US national champions, a sign of how much weight leading governments put on owning frontier AI. Bloomberg reported the same day that Mexico is missing out on the AI data center boom because of an energy crunch, which echoes the EBRD’s point about power limits.
What the sources do not establish: the report’s 0.7 point productivity gain is a scenario estimate by the bank, not a forecast, and the 1.47 euro figure is an average across economies that may not hold for any single country. For readers outside the EBRD regions, the useful signal is the pattern. Where chips and frontier models are out of reach, the bank expects the gains to come from adopting AI, and that depends on skills, power and capital.
Sign Up for the Daily AI Pulse
One email a day. All the stories that matter.
By signing up, you agree to our Terms of Use and acknowledge the data practices in our Privacy Policy. You may unsubscribe at any time.
Alex CarterAI News & Big Tech CorrespondentAlex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.
You may also like

Blackstone’s data center REIT spends $5 billion on its first deal: 13 AWS buildings

Anthropic’s Nasdaq debut puts $26bn of chip and cloud money on the line

Cerebras triples cloud revenue to $141.9M and lifts its 2026 outlook to $940M+
