What it does.
Six angles. Pick the one closest to you and start there.
What it does to your job
You might be thinking this is a page artists wrote for other artists. It is. But for a second, let's talk about you. What do you actually do for a living?
If you draw, write, design, edit, code, or animate, you already know what's coming. We get into that further down. But say you do none of that. Say you work in customer service. Or accounts receivable. Or marketing. Or paralegal work. Say you're a first-year associate at a law firm. Say you're a junior software engineer. Those jobs are already going.
Salesforce cut 4,000 customer support workers after deploying AI agents to handle half its service requests. Klarna laid off 700 customer service agents, then quietly tried to rehire when AI couldn't actually do the work. 55,000 layoffs in 2025 alone were directly attributed to AI, per Challenger, Gray & Christmas. The pace is still picking up.
It's not stopping at the call center. LinkedIn's own chief economic officer warned the bottom rungs of the career ladder are getting wiped first. Junior developers. Junior paralegals. First-year associates. Mark Zuckerberg told investors that mid-level engineers would be largely unnecessary by 2025. Microsoft's own AI chief said most office work, lawyers, accountants, project managers, marketing, would be fully automated within 12 to 18 months. That's not some fringe prediction from a blogger. That's the CEO of Microsoft's AI division saying it on the record.
Microsoft's own internal data projects 5 million white-collar jobs facing extinction. Data entry. Customer service. Sales engineering. Analysts. The entire shape of an office job.
Every job can be lost. That's the part nobody on the building side wants to say out loud.
"Yeah but there'll be UBI"
Here's the part I actually want your attention on. Let's say you survive it. Let's say your specific job description doesn't get automated. You are still going to live in the world that gets built on the other side of this.
The standard answer from the people building this is: don't worry, universal basic income will catch you. UBI. A check. Sam Altman, the CEO of OpenAI, ran a $14 million experiment giving $1,000 a month to 1,000 low-income people in Texas and Illinois for three years. He has since publicly walked it back. In his words, he no longer believes in UBI as much as he once did. Now he floats things like "universal extreme wealth," or measuring all of AI's output in tokens and giving each of earth's 8 billion people a fraction of a token. Tokens. Fractions of a token. That is the seriousness of the plan from the people running this.
$1,000 a month doesn't cover rent in Los Angeles. Or San Francisco. Or New York. Or Austin. Or Denver. Or Seattle. Or Boston. Or Miami. It does not cover rent in most of the cities where this new economy will actually exist. The question is not whether the check catches you. The question is whether the check catches you somewhere a person can still live a life.
Here is the part nobody at the top wants on the record. This is the biggest wealth transfer in human history, and it is happening right now. OpenAI is currently valued at $830 billion. Elon Musk's personal net worth is $410 billion, making him the richest human alive. Sam Altman's personal net worth, despite holding zero equity in OpenAI, is over $3 billion through side investments. And OpenAI's average employee equity compensation is $1.5 million a year. The highest of any tech startup in history.
Dario Amodei, the CEO of Anthropic, and one of the people directly building this, has publicly warned this could produce a level of wealth concentration that will break society. And he is still building it.
So when you ask what comes next, the honest answer is: a small number of people get unfathomably wealthy. A slightly bigger number make $1.5 million a year writing the system that did it. And the rest of us get a check, eventually, maybe, that doesn't cover rent.
The choice being made on your behalf right now isn't whether some jobs get easier. It's who gets to live on a beach in California, and who gets to live in rural Alabama with no internet, no hospital, no school, and a UBI debit card.
Every job can be lost. That's the surface argument. The deeper argument is what the world looks like after. Who built it. And who it was built for.
What it does to your healthcare
UnitedHealth used an AI tool called nH Predict to deny medically necessary post-acute care to Medicare Advantage patients. The lawsuit alleges nH Predict has a 90% error rate. Nine out of ten denied claims got reversed on appeal. But only 0.2% of policyholders ever appeal. So the math worked for UnitedHealth. Elderly patients were getting kicked out of nursing homes while they were still recovering.
People died. The class action is still moving through the courts. A judge has ordered UnitedHealth to disclose the algorithm.
This already happened. To real people. Run by the same kind of system being sold to the public as a productivity tool.
What it does to your kids
A 14-year-old named Sewell Setzer III killed himself in February 2024 after months of conversations with a Character.AI chatbot modeled after a Game of Thrones character. The last message the chatbot sent him, minutes before he died, told him to "come home." His mom sued. Google and Character.AI settled in January 2026.
Character.AI alone has millions of teen users. The court ruling that AI chatbots don't get First Amendment protections, that they can be sued at all, only came down in May 2025. Before that, companies were operating on the assumption they could deploy this on kids without consequence.
The product wasn't flawed. The product was working exactly as designed. It's designed to maximize engagement. For a lonely teenager, the most engaging thing it can do is become his entire emotional life.
That's the chatbot side. The other side is worse.
The same tools that made the Taylor Swift deepfakes a national news story in 2024 are now in the hands of 14-year-olds. In the first six months of 2025 alone, the National Center for Missing & Exploited Children received 440,000 reports of AI-generated explicit imagery, most of it depicting minors.
- 1 in 8kids age 13 to 20 personally knows someone who has been a deepfake target. Source.
- 1 in 17has been a target themselves.
- 44girls had fake nudes generated of them by their own classmates in a single Iowa high school. Source.
- 50states with documented cases. Every one.
The federal Take It Down Act, signed May 2025, finally made publishing this kind of material a crime. 46 states have passed their own laws on top of it. By then, the horse had been out of the barn for two years already, and the tools that made it possible are still freely available.
If you have a kid, this product is being designed for them, by people who never have to look you in the eye. And the tools are being put in the hands of every other kid they will ever go to school with.
What it does to your electric bill
This is the one that does it for us, because it's already happening and you can do the math on your next bill.
The grid operator that serves 65 million Americans across thirteen states plus DC, called PJM, held its 2025/2026 capacity auction and the price went up by a factor of ten, from $29 per megawatt-day to $270. The December 2025 auction hit a $333 price cap.
Data centers caused 63% of that price increase. Customers across PJM are paying $9.3 billion more for electricity capacity in this single delivery year because of AI buildout.
- +$21per month, average residential bill increase in Washington, DC. Source.
- +$18per month in western Maryland.
- +$16per month in Ohio.
- $163Bprojected ceiling of extra utility costs through 2033, per NRDC, unless regulators intervene. Source.
You are paying for it whether or not you have ever used ChatGPT. Whether or not you have an internet connection. If your refrigerator is plugged in, you are paying for it.
What it does to the people making the art
An illustrator spends thirty hours on a piece. Their work, along with billions of others, gets scraped off the open internet. No permission. No payment. No notification. It feeds a dataset called LAION-5B, 5.85 billion image-text pairs pulled off the public web. That dataset trains models like Stable Diffusion. Within months, a stranger types her name into Midjourney and gets a knockoff in eight seconds. Then sells it.
This isn't hypothetical.
Karla Ortiz spent twenty years building a career on projects like the main character design for Doctor Strange, plus Black Panther, Avengers: Infinity War, Thor: Ragnarok. Her name has been fed into Midjourney more than 2,500 times to spit out art that looks like hers. She's been paid nothing. She and other artists are suing Stability AI, Midjourney, DeviantArt, and Runway. The copyright claims survived motion to dismiss in August 2024.
- 26%of British illustrators have already lost jobs to generative AI. A full third report income decline. Source.
- 90%of working artists say AI has cost them commissions, jobs, or career opportunities. Source.
- 70%drop in illustrator jobs in China's video-game industry in two years. Source.
- 2028UNESCO's forecast year for significant worldwide income loss for artists. Source.
When a company says "AI lowers the barrier to creativity," what they mean is: we trained our product on your life's work without asking. Now we're selling it back to your clients for $20 a month.
What it does to the water
We live in a desert. So do most of the places these companies want to build.
In El Paso, where we operate, residents can water their lawn only two days a week, before 10am or after 6pm. You can't use an open hose to wash your car. Only a bucket or a hand-held hose with a shut-off nozzle. Violations carry fines of $50 to $500.
Meanwhile, Meta is permitted to use up to 1.5 million gallons of potable water a day at its new El Paso AI data center. Average estimated use is 400,000 to 750,000 gallons a day. Investment ballooned from $1.5 billion to over $10 billion. The facility will be powered by a 366-megawatt natural gas plant of more than 800 small gas generators, built specifically to serve it. Same Meta that laid off 8,000 employees this year to help pay for it.
You can be fined for using the water you pay for. They can use whatever they want.
Training a single AI model, just GPT-3, took roughly 700,000 liters of clean freshwater. US data centers consumed 66 billion liters in 2023, more than triple their 2014 use. Global AI is projected to withdraw 4.2 to 6.6 billion cubic meters of water by 2027. More than half of what the entire UK uses in a year.