Based on the amount of AI Data Centers they're building I came to the conclusion they are not primarily used for citizen surveillance. Although, that's one of the uses... the main reason THEY need so much compute power is because of the New Financial system that is to come. We're talking about huge computing capacity if it's to judge only by the amount of electricity consumption needed.
Global data center electricity consumption is projected to reach 565 terawatt hours by the end of 2026, up 26% from the year before. Goldman Sachs projects US data center power demand will climb from 31 gigawatt in 2025 to 66 gigawatt by 2027. That is more than doubling in 2 years. The US Department of Energy found that data centers consume 4.4% of all US electricity in 2023 and projects that figure would hit 12% by 2028. Over 5,000 active data center facilities are operating in the United States today with more than 700 under construction across 38 states. In Virginia alone, data centers consume more than a quarter of the state's electricity.
You do not build that much compute capacity for faster stock settlement. You build that much compute capacity when you intend to process, monitor and enforce rules on every transaction in a programmable financial system. My question: is the US data center buildout designed for domestic programmable finance or for global programmable finance? I don't know. As I understand a similar AI data center buildout is planned for Australia. Maybe that would be used to supplement the load required for US or Global demand. Because if you are pushing stable coins and digital tokens out to the entire world through the dollar system, and THEY're just starting to do, the compute requirements are orders of magnitude larger.
You're correct in saying AI training is one of the most computationally intensive tasks in commercial technology. Because it relies on distributed GPU clusters optimized for high throughput and memory bandwidth. The compute required to train the major AI systems has doubled approximately every 6 months. Training AI is computationally intensive because it involves performing numerous floating-point operations (FLOPs) to adjust billions of parameters across massive datasets. I suspect that's one area where they grossly missed estimating the compute requirements. But, that still doesn't explain the data center buildout, IMO.
However, consider this - the Depository Trust & Clearing Corporation (DTCC) is the institution provider that automates, centralizes, and standardizes the clearing and settlement of securities transactions for the global financial system. DTCC manages assets valued at approximately $114 trillion in its central securities depository subsidiary, DTC, as of June 2026. In terms of transaction volume, the DTCC processes over $2.5 quadrillion in securities transactions annually. Managing that transaction volume calls for a lot of computing power. DTCC has already begun live production trades using tokenized stocks, bonds, and treasuries, and most people do not even know it is happening. The data center buildout required to run this system is so enormous that it tells you something about the intended scale that no official will say out loud.
Based on the amount of AI Data Centers they're building I came to the conclusion they are not primarily used for citizen surveillance. Although, that's one of the uses... the main reason THEY need so much compute power is because of the New Financial system that is to come. We're talking about huge computing capacity if it's to judge only by the amount of electricity consumption needed.
Global data center electricity consumption is projected to reach 565 terawatt hours by the end of 2026, up 26% from the year before. Goldman Sachs projects US data center power demand will climb from 31 gigawatt in 2025 to 66 gigawatt by 2027. That is more than doubling in 2 years. The US Department of Energy found that data centers consume 4.4% of all US electricity in 2023 and projects that figure would hit 12% by 2028. Over 5,000 active data center facilities are operating in the United States today with more than 700 under construction across 38 states. In Virginia alone, data centers consume more than a quarter of the state's electricity.
You do not build that much compute capacity for faster stock settlement. You build that much compute capacity when you intend to process, monitor and enforce rules on every transaction in a programmable financial system. My question: is the US data center buildout designed for domestic programmable finance or for global programmable finance? I don't know. As I understand a similar AI data center buildout is planned for Australia. Maybe that would be used to supplement the load required for US or Global demand. Because if you are pushing stable coins and digital tokens out to the entire world through the dollar system, and THEY're just starting to do, the compute requirements are orders of magnitude larger.
Perhaps they are preparing for the huge amount of data to be aggregated to 'evolve' and upgrade AI rather than sheer surveillance.
That wouldnt be as per raw data storage per se, but more about computing all of it over and over through ML for AI
You're correct in saying AI training is one of the most computationally intensive tasks in commercial technology. Because it relies on distributed GPU clusters optimized for high throughput and memory bandwidth. The compute required to train the major AI systems has doubled approximately every 6 months. Training AI is computationally intensive because it involves performing numerous floating-point operations (FLOPs) to adjust billions of parameters across massive datasets. I suspect that's one area where they grossly missed estimating the compute requirements. But, that still doesn't explain the data center buildout, IMO.
However, consider this - the Depository Trust & Clearing Corporation (DTCC) is the institution provider that automates, centralizes, and standardizes the clearing and settlement of securities transactions for the global financial system. DTCC manages assets valued at approximately $114 trillion in its central securities depository subsidiary, DTC, as of June 2026. In terms of transaction volume, the DTCC processes over $2.5 quadrillion in securities transactions annually. Managing that transaction volume calls for a lot of computing power. DTCC has already begun live production trades using tokenized stocks, bonds, and treasuries, and most people do not even know it is happening. The data center buildout required to run this system is so enormous that it tells you something about the intended scale that no official will say out loud.
regional. see 1984. abt 3 or 4 regions.