By Helena Glass at Global Research. Reposted with permission.
AI will liberate humanity!
Every household will have their private robot for free!
Humans will never work again!
And the mass hypnosis droned on and on and on.
Except AI agrees that we were lied to.
AI: the reality reveals that computing power, energy consumption, and high token usage make AI an exceptionally expensive line item.
In fact, so expensive that businesses are shutting down use of AI as it is pricing them out of business.
Nvidia’s Vice President of Applied Deep Learning publicly stated that compute costs for their AI engineering teams far surpass the actual cost of the employees’ salaries.
An economic analysis by MIT found that AI is only cost-effective in about 23% of basic jobs; in the remaining 77%, keeping a human worker is the more economically viable option, making all those 5,000+ data centers ‘obsolete’ and the new ones being built a fictional storyline. It is called ‘token usage’ and small business owners are staring down bills as high as $113,000.
Heavy users were burning $500 to $2,000 per month each. Microsoft is banning the use of AI by its engineers. The boom is going to bust after global spending reaching over $6.7 trillion.
Across the board, tech companies claim that human employees are much more cost-effective. But our scholarly education systems have failed to teach real world skills for engineers leaving 80% unemployable. Exasperating a problem that will become exponential as businesses scramble for humans to replace AI. Are we in the beginning stages of an AI bubble?
Many companies were deepening their debt as they bought into the data center hype. Offloading AI means that debt remains with no value to offset it. When the value is gone, companies have to raise their pricing to mitigate the debt. And the inflation cycle ramps up rather quickly. The debt attached via bank loans could see a chapter 11 fallout as companies can’t afford the data costs much less the loans. The Department of War engaging in many of the largest data centers across the country, including in Utah, have yet to book or analyze the costs. And given Hegseth’s extensive ‘math and accounting’ background, the Pentagon is going to go broke, making Trump’s military bunker under the ballroom even more suspect for its unsustainability.
Although, on a brighter note, the economic realization may be a pushback for the launch of the ‘digital credit system’ as proposed by the World Economic Forum. Building a global credit and financial system involves monumental outlays, including trillions in infrastructure and immense societal trade-offs. While the structure could be built in roughly three years, the integration would take another seven to ten years —variability being country-specific.
The land alone for these data centers is in the billions with military intelligence costs considered “classified”. One such data center is owned by the NSA and is located at Camp Williams, Utah. Perspective: The Camp David data center is 1 million square feet. 100,000 of the million is designated mission critical Tier III, while 900,000 square feet is designated technical support and admin – 90%. Completed in 2014, the 20-building complex includes water treatment facilities, chiller plants, electric substation, fire pump house, warehouse, vehicle inspection facility, visitor control center, and 60 diesel-fueled emergency standby generators and fuel facility for a 3-day 100% power backup capability. This was the source for Edward Snowden’s whistleblower release.
The electricity cost for Camp Williams data center is $70 million; maintenance is another $20 million; water usage is 1.5 million gallons daily. It is built on a desert. The purpose? Surveillance.
While government subsidies have eliminated property taxes on these facilities, their value to the state is a negative number given the depletion of electricity and water. Their employees are not local. And their spectrum of surveillance is on a global as well as national level. In that regard, it is curious why we would need more than one. According to Bloomberg, in 2026, half of all proposed data centers will face delays or cancellation. Why? Shortage of materials and supply chain delays.
Components are not manufactured in the US so contractors need to buy from Canada, Mexico, South Korea and China. But Trump has created riffs with these countries hemorrhaging land loans, inflating building costs, while interest rates on loans eat up the remaining cash on hand, which is why corporations are looking into bond sales, bridge loans, and other means as time frames lag.
With the realization that AI is cost-prohibitive, corporations scaling back will create a glut of outstanding loans against vacant land. Banks will look for bailouts. And the circle will stop short. Oracle has $100 billion in debt against an Abilene data center where clients are walking away. Outstanding debt has reached hundreds of billions with Morgan Stanley claiming the need for trillions in external funding in America alone.
The military funding on taxpayer debt? Top secret – super doper classified. Don’t even ask – whisper.
Helena Glass is Former CPA & Series 7, with emphasis in Real Estate and Financial Planning. Two brains in one: former Bronze Sculpter and Danseuse. Visit the author’s blog.
