What's slowing down the AI buildout (www.worksinprogress.news)

The AI buildout is facing significant delays primarily due to bottlenecks in the electrical grid that connects new data centers and energy-intensive facilities to power sources. A major project, the Stargate computing campus in Abilene, Texas—spearheaded by OpenAI and Softbank—is set to cost over $40 billion and is projected to require as much electricity as 313,000 homes at peak capacity. While the demand for electricity to power AI innovations is expected to soar, with worldwide AI computing power anticipated to hit 100 gigawatts by 2030, the interconnection process for new infrastructure is significantly backlogged, averaging 55 months in 2023 compared to 20 months in 2005. This inefficiency hampers the ability of major tech players like Nvidia and Meta to expand their AI capabilities, which are vital to driving economic growth. The current grid system operates on an outdated first-come, first-served basis that does not accommodate the urgency or scale of modern energy demands. The result is that many data centers are contemplating off-grid solutions to meet their power needs faster, which is not ideal given the reliability and cost advantages of grid power. To support the future of AI and electrification, grid infrastructure and processes must adapt to allow for quicker and more efficient connections. This transition is critical as the intersection of AI advancements and energy accessibility will shape the capacity for innovation and economic potential in the coming years.