The five layers between power and GDP

The UAE's Barakah complex generates 5.6 GW of clean baseload. Abu Dhabi's MGX fund targets £79bn ($100bn, €95bn) in AI infrastructure assets. Microsoft invested £1.2bn ($1.5bn, €1.4bn) in G42. The connection between energy generation and economic output is not theoretical in the Gulf. It is architected.

We have been tracking a pattern across our analysis. Two weeks ago, we showed how China compresses the gap between generation and compute. It integrates energy, grid, and demand under single authority. Last week, we showed that carbon accounting standards measure Layer 5 outcomes without crediting Layer 1 infrastructure. The pattern is the same. Every global AI competition is, underneath, an energy competition.

The infrastructure stack has five layers. Each depends entirely on the one below.

Layer 1: Sovereign energy supply. Reliable, domestically controlled power generation. Nuclear, renewables, gas. The foundation.

Layer 2: Grid infrastructure. Transmission, distribution, interconnection. Moving electrons from generation to demand.

Layer 3: Compute infrastructure. Datacenters, AI clusters, edge facilities. Entirely dependent on Layers 1 and 2.

Layer 4: AI capability. Models, training, inference. Entirely dependent on Layer 3.

Layer 5: Economic and strategic output. GDP growth, defence capability, industrial competitiveness. The output of the entire stack.

The global AI race is fought at Layers 3 to 5. The binding constraint sits at Layers 1 and 2.

The Lawrence Berkeley National Lab reports 2,300 GW of generation projects in the US interconnection queue. Globally, the IEA estimates over 1,650 GW in advanced development waiting for grid connection. Projects that clear Layer 1 permitting stall at Layer 2 delivery.

The average US interconnection study now takes five years. The generation exists or can be built. The grid cannot absorb it. This is a Layer 2 bottleneck constraining everything above it.

China's State Grid Corporation plans to invest £453bn ($574bn, €548bn) in grid infrastructure between 2026 and 2030. That is a Layer 2 investment at sovereign scale. The US has no equivalent programme.

China: Layers 1 through 3 under single authority. 62 GW operational nuclear, 43 GW under construction. State Grid spending £453bn on Layer 2. Datacenters co-located by central planning. The stack is vertically integrated. The binding constraint is Layer 4: AI capability, where talent and model development lag the West.

Gulf states: Layers 1 through 3 via sovereign capital. UAE's Barakah anchors Layer 1. MGX, Mubadala, and G42 fund Layers 2 and 3 directly. The model bypasses commercial financing entirely. The binding constraint is Layer 4: indigenous AI talent and model development remain nascent.

Europe: Rebuilding Layer 1. France committed to six new EPR2 reactors at an estimated £61bn ($72.8bn, €69bn). The UK started construction on Sizewell C in January 2026 (3.2 GW). Sweden reversed its nuclear phase-out. The energy sovereignty dividend from A66 is real: France exports 101 TWh while Germany imports. But Layer 2 queues and permitting delays slow the path to Layer 3.

United States: Layer 2 gridlock. 2,300 GW queued. Generation capacity exists or is planned. Grid delivery is the bottleneck. Hyperscalers bypass it: Meta signed 6.6 GW of nuclear deals. Microsoft contracted 34.7 GW. They are building Layers 1 to 3 privately because public Layer 2 cannot deliver. The binding constraint is governance, not physics.

The IEA projects global datacenter electricity demand will grow from 415 TWh in 2024 to 945 TWh by 2030. That is a 127% increase in six years. Every terawatt-hour of Layer 3 demand requires a terawatt-hour of Layer 1 supply, delivered through Layer 2 infrastructure.

India illustrates the challenge at scale. 8 GW of nuclear capacity operates today, with a target of 100 GW by 2047. Eight reactors are under construction. The Layer 1 ambition is clear. But India's grid loses 19% of generated power to transmission and distribution inefficiency. Layer 2 delivery will determine whether India competes at Layers 3 to 5. Japan faces a different Layer 1 question. It has restarted 15 reactors with 33 GW of combined capacity. Each restart displaces LNG imports and strengthens energy sovereignty. But grid queue bottlenecks are not uniquely American. Every major economy faces its own Layer 2 constraint. Grid delivery shapes outcomes differently in each region.

Strategic Purpose: Validate five-layer model with community data. Collect evidence on where constraint sits. Informs Articles 68 to 73 emphasis.

Most AI strategy documents start at Layer 3. Cloud capacity. Datacenter availability. Chip supply. These are real constraints. They are not the binding ones.

The binding constraint is always the lowest unfunded layer. In the US, that is Layer 2: grid delivery. In Europe, that is Layer 1: generation capacity. In the Gulf and China, Layers 1 to 3 are funded. The constraint is Layer 4: talent and models.

This explains why the same announcement produces different outcomes in different regions. A 1 GW datacenter in Abu Dhabi connects to sovereign energy within 18 months. The same datacenter in Virginia waits five years for grid interconnection. Same Layer 3 asset. Different Layer 2 reality.

Three predictions through the five-layer lens: First, Layer 2 investment will become the defining infrastructure theme. China's £453bn grid programme is the signal. Sovereign wealth funds, not utilities, will drive grid investment in the Gulf and Southeast Asia.

Second, hyperscaler bypass strategies will accelerate. Meta, Microsoft, Amazon, and Google are building private Layer 1 to Layer 3 stacks. Behind-the-meter nuclear, direct PPAs, and co-location are not niche strategies. They are structural responses to Layer 2 failure.

Third, the nations that control Layers 1 and 2 will set the terms for Layers 3 to 5. Energy sovereignty is AI sovereignty. France's 101 TWh of clean exports, Finland's 96% domestic coverage, and Japan's LNG displacement are not climate stories. They are Layer 1 positioning for the AI era.

The five-layer infrastructure stack is physics, not ideology. Each layer depends on the one below. The global AI race is fought at the top. The winners understood the bottom.

Where does your organisation break? If you cannot answer that in one sentence, the model just showed you why.

Next week: We apply the five-layer lens to financing. Why banks price Layer 1 as commodity risk whilst sovereign wealth funds price it as strategic infrastructure.

AI is moving fast and most people are falling behind.

The Rundown AI is a free newsletter that keeps you ahead of the curve.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.