Decentralised News (@decentralisednews)

The AI Infrastructure Bubble Debate: $7.6 Trillion of Compute, Power and Risk The AI boom is usually framed as a chip story. It is not. It is becoming one of the largest infrastructure capital cycles in modern economic history. Amazon, Microsoft, Alphabet and Meta are guiding to roughly $725 billion in combined 2026 capital expenditure. Goldman Sachs models a broader $7.6 trillion AI infrastructure buildout between 2026 and 2031. That is not a normal software cycle. That is compute, data centres, memory, cooling, land, transformers, grid connections and electricity. Nvidia is the visible winner. Its data centre business has become the headline symbol of the AI boom. But the deeper bottleneck is no longer only GPUs. It is power. AI data centres are now colliding with the limits of the electricity grid. U.S. grid interconnection queues have become a major constraint. In some regions, data centre demand is already feeding into higher capacity prices and household electricity bills. That changes the politics of AI. When the cost of the buildout starts showing up outside the data centre, regulators notice. The revenue side is just as important. OpenAI and Anthropic are scaling quickly. But their revenue is still small relative to the hundreds of billions being spent on infrastructure to serve future AI demand. That is the real question: Can inference demand grow fast enough to justify the capital being deployed today? Not just chatbot demand. AI agents. Enterprise workflows. Coding systems. Search. Customer support. Data analysis. Robotics. Autonomous operations. This is why Decentralised News built the DN Query Cost Ledger. The tool focuses on the missing number in the public conversation: What does one AI query actually cost to run? And how much of what users pay is electricity versus the full capital stack behind the query? Chips. Memory. Cooling. Networking. Depreciation. Data centre financing. Cloud margin. Model operations. The electricity cost may be small. The infrastructure behind the query is not. For crypto and DePIN investors, this matters. If hyperscaler compute is scarce, expensive and capacity-constrained, decentralised compute networks may have a real economic opening. But only if they can prove real usage. Not hype. Not token emissions. Not AI branding. Real customers. Real compute revenue. Reliable supply. Verifiable workloads. Token economics linked to demand. The AI boom is real. The technology works. But the investment question is much harder: Who owns the rails that intelligence runs on? Full analysis on Decentralised News: https://decentralised.news/what-one-ai-query-really-costs-and-why-the-answer-matters #AI #ArtificialIntelligence #DataCenters #AIInfrastructure #Nvidia #BigTech #CloudComputing #Energy #Electricity #DePIN #Crypto #AICrypto #Compute #Hyperscalers #OpenAI #Anthropic #Macroeconomics #DecentralisedNews