Ashmore Group, with $75 billion under management, published a research note in May 2026 titled "The AI Supercycle and Bottlenecks: A Multi-year Tailwind for EM." It is the most rigorous institutional analysis of AI infrastructure constraints produced to date — with detailed modelling of chip supply, grid capacity, transformer lead times and behind-the-meter solutions.
Their core conclusion aligns precisely with what GridReadiness has been documenting since launch: physical constraints, not capital or software, are the primary ceiling on AI infrastructure deployment. Their investment directive: own the constraints, not generic AI beta.
There is one significant gap in their analysis. France is not mentioned once.
WHAT ASHMORE GOT RIGHT
Several findings from the Ashmore report are worth examining in detail, because they validate and extend the GridReadiness thesis with institutional-grade data.
Goldman Sachs AI capex: $6 trillion by 2030 · $7.6 trillion by 2031
US grid queue (PJM/Virginia): average 8+ years 2025 · was 2 years in 2008
US data centre capacity on schedule: only 50% of 2025 planned capacity
Transformer lead times: tripled since 2022 · 150+ weeks (nearly 3 years)
Transformer inflation: +80% since 2020 · second highest in US PPI of 47 categories
Gas turbines: booked through 2028-2030 · GE Vernova, Siemens, Mitsubishi
US deployable AI capex by 2030: ~$2.75 trillion · constrained by 55 GW power ceiling
The transformer data is particularly important. Ashmore writes: "A $100 billion data centre campus can sit idle waiting on a $40 million transformer order." This is not a marginal observation. It is the central operational reality of AI infrastructure deployment in 2026.
THEIR RELEASE VALVE LIST — AND THE GAP
Ashmore identifies what they call "power-rich release valves" — markets that can absorb AI infrastructure investment redirected from a constrained US grid. Their list:
Notable absence: France · Western Europe more broadly
This is a significant analytical gap. Ashmore is primarily an emerging markets investment manager — their lens is naturally oriented toward EM power-rich regions. This makes their France omission understandable but not excusable for a comprehensive bottleneck analysis.
WHY FRANCE SHOULD BE FIRST ON THEIR LIST
The Ashmore report outlines four criteria for a viable release valve market: available firm power, accessible grid connection timeline, proximity to hyperscaler markets, and regulatory stability. France meets all four more completely than any market on their list.
vs GCC: gas-dependent · weather-cooling intensive · geopolitical exposure
vs Morocco: renewable intermittency · early-stage grid infrastructure
vs Scandinavia: hydro limited capacity · distance from EU hyperscaler hubs
Grid connection timeline: France brownfield 18–36 months
vs Scandinavia: comparable but fewer brownfield sites
vs Latin America: 3–7 years · regulatory complexity
vs GCC: fast but distance from US/EU hyperscaler operations
Regulatory stability: France EU member · Loi Industrie Verte · Choose France programme
vs Morocco: developing regulatory framework
vs GCC: improving but sovereign risk premium applies
Proximity to market: France 1–2 hours from London, Frankfurt, Amsterdam
Sub-10ms latency to major European financial and tech hubs
THE CHOOSE FRANCE VALIDATION — PUBLISHED AFTER ASHMORE'S CUTOFF
The Ashmore report was dated May 31, 2026. Choose France took place on June 1, 2026. The institutional capital did not wait for the research note:
Ardian/Verne: €5 billion · 500 MW · Île-de-France · "grid connection already secured"
Nebius: €8 billion · 240 MW · Béthune brownfield · operational end 2026
Digital Realty: 4 sites under simultaneous construction · Marseille + Paris
Mistral AI: €4 billion · France + Sweden
Ardian's exact words — "the grid connection is already secured" — are the operational translation of everything Ashmore documented. A €150 billion fund chose France specifically because it solves the bottleneck Ashmore identified. This is not sentiment. As Ashmore themselves write: "The sovereignty trade isn't sentiment — it's operational necessity."
THE TRANSFORMER ARBITRAGE — ASHMORE'S BLIND SPOT
Ashmore documents the transformer crisis comprehensively. They note lead times have tripled to 150+ weeks. They note 80% price inflation. They note that new US transformer factories are two to three years from meaningful output.
What they do not document: EU second-tier transformer manufacturers who have not been absorbed by the US demand surge.
Efacec (Portugal): 20–28 months · ANSI/IEEE 60Hz available for US projects
TMC (Italy): 20–28 months · industrial specification
Pauwels (Belgium): 24–32 months · large power transformers
vs US OEMs (ABB/Siemens/Hitachi): 48–60 months
vs GE Vernova: 60+ months · $100B+ backlog
The EU second-tier window is narrowing as Choose France commitments absorb capacity.
A developer sourcing transformers from EU second-tier manufacturers for a French brownfield site is compressing two bottlenecks simultaneously: the grid connection (brownfield 18–36 months vs US 8+ years) and the transformer procurement (EU 20–32 months vs US OEM 48–60 months). This is the operational arbitrage that Ashmore's report does not model.
ASHMORE'S BASE CASE — AND WHAT IT MEANS FOR FRANCE
Ashmore's base case scenario — "bottleneck-constrained cycle" — is the most instructive for understanding France's position:
"The US bears more of the capex-growth slowdown, while EM captures the bottleneck rents."
Replace "EM" with "France" and the sentence becomes more precise. France is not an emerging market. It is a developed market with the operational profile of the release valves Ashmore describes. The institutional capital is beginning to recognise this — Choose France 2026 is the evidence.
The gap between Ashmore's analysis and France's actual position represents a significant information asymmetry. The institutional research community has not yet fully mapped France onto the bottleneck thesis. GridReadiness exists to close that gap — with monthly data on transformer lead times, RTE connection timelines, and brownfield site availability.
Source: Ashmore Group. "The AI Supercycle and Bottlenecks: A Multi-year Tailwind for EM." May 31, 2026. Gustavo Medeiros, Ben Underhill. GridReadiness analysis and France-specific data are independent of and not endorsed by Ashmore Group. All Ashmore data points paraphrased from the original report. Original Ashmore report →