Defense is going modular, and private capital is setting the terms
NATO's transformation chief says adoption, not invention, is the bottleneck. Launch capacity and power contracts show where the money is going.
- Admiral Pierre Vandier says the primary challenge for NATO's transformation is the cultural and behavioral adoption of new technologies, rather than the innovation process itself. m:…1105
- Ukraine replaced artillery with off-the-shelf drones and AI-driven software to counter the shelling advantage it faced. m:…1105
- Anthropic agreed to purchase approximately 3.4 billion dollars of Caterpillar power-generation equipment. m:…8387
- CoreWeave says it expects approximately 9 billion dollars in customer prepayments and deposits over the next three months. Self-reported m:…8387
- SpaceX lists three upcoming missions: USSF-317, NROL-136 and USSF-266. m:…7192
- SoftBank canceled 13 gigawatts of turbine reservations for AI data centers, taking a 3.7 billion dollar earnings hit. m:…8387
Adoption is the bottleneck
NATO's senior transformation commander says the hardest part of modernizing defense is not inventing new technology but getting armies to use it. Admiral Pierre Vandier, NATO's Supreme Allied Commander Transformation, argues the war in Ukraine has forced a shift from crisis management to a protracted war that current Western stockpiles were not designed to support, and the primary challenge now is the cultural and behavioral adoption of new technologies rather than the innovation process itself. m:…1105
Ukraine shows what adoption looks like when it works. The country replaced artillery with off-the-shelf drones and AI-driven software to counter the shelling advantage it faced. Vandier's prescription is to buy that way on purpose: modular marketplace models, assembled in Lego-like fashion, are the most effective way to integrate IT-centric capabilities such as drones and software, rather than relying on traditional consolidation of the defense industrial base. m:…1105
The obstacle is political as much as technical. Europe's defense industrial base is fragmented, and political pressure to maintain domestic jobs hinders consolidation. That is where Vandier sees private capital doing work governments cannot: it can accelerate defense modernization by enforcing new performance measures that prioritize future-oriented investments over legacy hardware. m:…1105
The launch layer
The launch capacity behind that model is already organized around national security work. SpaceX runs three primary launch sites: Starbase in Texas, Kennedy Space Center in Florida, and Vandenberg Space Force Base in California. Starbase is the site for development, manufacturing, testing and launch of the Starship and Super Heavy rocket, and it also serves as the company's headquarters. m:…7192
The Florida and California sites split the work. Launch Complex 39A at Kennedy supports Falcon 9 and Falcon Heavy commercial satellite launches, and also supports human spaceflight, space station resupply, National Security Launch and Starlink missions. Launch Complex 4 East at Vandenberg provides access to high inclination and polar orbits, and supports satellite communication constellations, National Security Launch, Earth-observing satellites and lunar missions. Three missions are listed as upcoming: USSF-317, NROL-136 and USSF-266. m:…7192
The power bill
The same capital logic is visible in the power contracts behind AI. Anthropic agreed to purchase approximately 3.4 billion dollars of Caterpillar power-generation equipment, a sign that frontier developers are securing their own electricity supply rather than waiting for the grid. CoreWeave says it expects approximately 9 billion dollars in customer prepayments and deposits over the next three months. m:…8387
Not every bet is being doubled. SoftBank canceled 13 gigawatts of turbine reservations for AI data centers, taking a 3.7 billion dollar earnings hit. OpenAI reportedly raised 15 billion dollars at a 1.25 trillion dollar valuation, according to reports cited in the Mayfield newsletter. Nearly 6 in 10 enterprise executives plan to increase their use of open-weight AI models to manage costs. m:…8387
What buyers should watch
Put together, the pattern is a procurement story, not a technology story. The buyers that matter are learning to assemble modular capabilities quickly, and the suppliers that matter are the ones that can plug into that model: a launch provider with national security missions on its manifest, a power equipment order that secures electricity supply, or a software vendor that sells into a marketplace rather than a long program. The risk sits on the other side: fragmented industrial bases, political pressure to protect domestic jobs, and the security questions that come with agents operating inside enterprises. m:…1105 m:…7192 m:…8387
The regulatory clock is also ticking. New York requires major frontier AI developers to register by January 1, 2027. Microsoft researchers demonstrated that editing fewer than 10 neurons in an AI model can create hidden backdoors, and United Airlines moved to block unauthorized third-party AI agents from booking flights. The implication is that the same enterprises buying modular AI capability will have to police which agents are allowed to act on their behalf. m:…8387
What is still unresolved
- What mechanisms will push nations to prioritize adoption over protecting domestic jobs in their defense industries.
- The timeline for moving from legacy stockpiling to modular marketplace procurement.
- Whether power and launch capacity can scale fast enough for the demand implied by the Anthropic equipment order and the CoreWeave prepayments.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.