US hyperscalers set to spend nearly $1.2 trillion in just three years, research shows - but I dread to imagine what will happen once the global data market cools down (www.techradar.com)

Forecasts from market analysts and a StockMarket.News post on X say US hyperscalers will pour roughly $1.15 trillion into data centers by 2027, building on $477 billion spent from 2022–2024. Industry-wide investment could hit $2.9 trillion through 2028, split roughly $1.6 trillion for chips and servers and $1.3 trillion for infrastructure (real estate, power, construction), with as much as $900 billion expected in 2028 alone. For scale, the entire S&P 500’s capex in 2024 was about $950 billion, and economists estimate data center and power spending could add up to 40 basis points to US GDP growth in 2025–26. For the AI/ML community this signals both opportunity and systemic risk: massive demand for accelerators, storage and power will accelerate chip, server and facilities development, concentrate capital (and environmental impact) in small geographies like Northern Virginia, and reshape supply chains and pricing for compute. But analysts warn the boom may be unsustainable—if global data demand cools, hyperscalers could face stranded assets, weaker returns and a market correction that crowds out investment in other industries (echoing 1990s telecom overbuild). Practically, researchers and operators should expect intense competition for hardware, regional power constraints and policy scrutiny, and must plan for volatility in infrastructure availability and costs as the market matures.