GVI Provisional Nowcast

“The scarce asset is not simply compute; it is financeable compute.”

The Weighted Global Velocity Index holds at 108.0 (2025=100), with a provisional unrounded movement of −0.01 that rounds to unchanged. The regime is clear: AI acceleration remains intact, but the cost and physical feasibility of financing it are becoming the binding constraint. Model confidence is 96%.

Today produces one admissible country revision: Norway falls −0.1 to 77.1 after Norges Bank raised its policy rate to 4.50%. The more consequential development is in the United States, where the 10-year Treasury yield has reached 5.19%, the 30-year rose above 5.50% intraday — its highest since 2004 — Brent is back around $105.75, and Oracle's Project Jupiter AI campus has hit a reported one-year power delay.

1. 01 United States — the financing constraint becomes physical

   Oracle issued a force-majeure notice on its 1,400-acre New Mexico AI data-centre campus, citing power delays; Reuters reports a roughly one-year delay. The campus has attracted about $18 billion in loans, with Blue Owl committing roughly $3 billion of equity, within the broader Stargate programme. Because the project remains committed and on schedule, the GVI does not downgrade the US today.

   The wider evidence is material: Morgan Stanley estimates AI infrastructure could require ~$1.5 trillion of external financing by 2028, while Moody's projects ~$1 trillion of capex from the six largest US tech firms in 2027. signal: US stays at 94.3, but Smart Capital momentum deteriorates.

2. 02 Smart Capital — 5% is no longer the ceiling

   The 10-year is around 5.19% (after touching 5.225%), the 30-year at 5.50%, the two-year near 4.90%. Markets imply roughly a 71% probability of another Fed increase next month and more than 90 basis points of additional tightening. Japan's 10-year has reached ~3.12%, its highest since 1996; French and German yields sit at 2007–08 levels.

   At a 5%-plus sovereign benchmark, the question shifts from "Can we fund AI?" toward "Which AI projects can still clear the cost of capital?" signal: A meaningful upward change in the AI hurdle rate.

3. 03 Norway — monetary tightening becomes measurable

   Norges Bank raised its policy rate by 25 basis points to 4.50%, saying inflation remains above target. Energy abundance no longer fully offsets the higher domestic cost of money. signal: Norway moves from 77.2 to 77.1, entirely through Smart Capital.

4. 04 China — Enterprise Transformation accelerates, overcapacity appears

   In H1 2026, 221,900 new AI-generated shows launched on Douyin; production costs for AI short dramas fell from ~5,000 yuan per minute to a few hundred, per CCTV data cited by Reuters. Yet only 1,055 passed 100 million views, per DataEye. AI eliminates production scarcity while creating attention scarcity and margin compression. signal: China holds 91.1; Enterprise Transformation momentum remains exceptionally strong.

5. 05 Discovery Capacity — AI tests the terrestrial boundary

   Google will launch a prototype satellite next week carrying its Tensor Processing Units under Project Suncatcher, testing radiation resistance, thermal management and reliability in low-Earth orbit. It receives zero numerical weight, but signals a system exploring alternatives to grid, energy, land and water constraints. signal: Discovery Capacity in its purest form — changing boundary conditions.

Sweden's Riksbank held at 1.75% but flagged stronger activity, a weaker krona and higher energy prices — forward guidance, so no adjustment. Switzerland offers the opposite: the SNB held at 0%, with August inflation at just 0.8%, gaining relative Smart Capital momentum but no numerical upgrade. As the advanced world tightens, low structural inflation becomes a competitive asset.

Brent trades around $105.75, up roughly 3% following another attack on Saudi Arabia. The transmission is clear: oil up, inflation up, central-bank reaction up, yields up, AI hurdle rate up. The physical and financial sides of the AI economy are now reinforcing each other negatively.

The main accelerator is the falling marginal cost of intelligence — tasks that cost thousands of yuan per minute now cost hundreds. The main decelerator is the rising marginal cost of everything around it: capital, energy, grid access, permitting and risk transfer. This is why the AI Supercycle can continue while the economics of individual projects deteriorate.

Key takeaways

for investors, executives and policymakers

• For investors: favour AI exposure with secured power, contracted demand, high utilisation, strong sponsors and conservative financing — the scarce asset is financeable compute.
• For executives: evaluate every major AI programme against a fully loaded hurdle rate covering compute, power, infrastructure, financing and deployment.
• For policymakers: AI competitiveness now depends on monetary credibility, power availability and permitting as much as on algorithms.
• Scarcity is moving from inside the machine to outside it: electricity, grids, water, capital and permitting.

The next winners will not necessarily spend the most on AI — they will be those capable of orchestrating the entire system so that intelligence compounds faster than the cost of financing and powering it.