I use AI. We need to tax and regulate TF out of it.

Just a few months ago, I was trying really hard to keep everything straight in my head: every Bright Body formulation decision and reorder point, every Hypema'am client's positioning, avatars, and roadmap, my kids' childcare schedules, making one-on-one time with my husband, taking care of my pets, the house, when my prescriptions were about to run out, and the fifty-leven things that every other working mom has on her mind.

It should come as no surprise that I forgot at least 7 important things on a daily basis.

And I'm not talking about the "where TF did I put my keys?" kind of stuff, I mean "oh sh*t I forgot to order that raw material and now we can't restock that product for another week."

At home, that meant I wasn't pulling my weight - the dishwasher wasn't getting unloaded, laundry piled up, I wasn't spending enough time with my own husband.

I was letting someone down on the daily because my brain was at maximum capacity.

And for a while, I tried to just push through, Kool-aid-man style. Despite living with chronic illness and having learned this lesson many times over, I had internalized the hypercapitalist idea that productivity is just a matter of willpower.

Eventually, Jimmy sat me down and told me he was feeling neglected. Ok, it wasn't a sit down, we had a full-blown fight. But he was right.

I was conflating capability with capacity. And I couldn't do it anymore.

I would spend all day at work and then at least another 1.5 - 2 hours every night trying to get everything done between both businesses, but I STILL couldn't keep up at home or at work.

Of course, I was doing way too much. So I started by cutting back.

But I quickly realized it wasn't just about the amount of work, it was the fact that I outgrew every system I put in place when the businesses were in different stages.

And hiring someone wouldn't have solved it, because everything lived in my brain.

Sure, I have SOPs and documented processes. But anything else - especially if it had to do with more conceptual and intangible stuff like strategy - was locked in my head.

Nothing in my head about either business existed outside my own sh*tty memory. Nine years of hard-fought judgment about how small brands grow, stored in my sieve of a brain.

In 2026, I knew that the obvious way to get concepts out of my head and into a system was using AI. But I had a lot of f*cking feelings about that.

Two things I take seriously

I GAF about sustainability. After all, I've spent nine years at Bright Body making decisions that cost me money because I wasn't willing to do the cheaper but more wasteful thing: we package in glass and aluminum so we can reduce plastic waste. We offer refills and even pay for customers to ship back empty refill pouches.

I also care a lot about data. I come from a family of academics, and if you've been in my world for five minutes you know I don't run on vibes and I don't make sweeping generalizations or claims I can't back.

I had been watching plenty of TikToks about the impact of AI, and it made me queasy.

But I also knew that the internet is where nuance goes to die, and if I wanted to understand the ins and outs of it, I would have to roll up my sleeves, use the Google machine with AI overview off, and form my own opinions.

Data centers are not new (but that doesn't mean the boom is okay)

Every single thing we do on a phone or a computer already runs through a data center, and it has for twenty years.

Now before you call me a corporate simp, stay with me here because we are talking facts and data before we jump to conclusions.

Your email, your photos, your Shopify storefront, your Klaviyo account, your kids' school portal, your online banking, every episode of House of the Dragon on HBO Max, every Zoom call, every file backup uses a data center.

The internet is a mix of underwater cables (yes, really) and buildings full of servers that use power and water somewhere you'll likely never visit.

This is why data center numbers were already pretty high even before we get to AI. Data centers used around 485 terawatt-hours of electricity in 2025, which accounts for 1.5 - 2% of all electricity on earth, and the International Energy Agency's mid-range estimate projects that to double to 950 TWh by 2030.

AI usage is part of data center usage, but not all of it.

Gartner estimates that AI-optimized servers use about 31% of data center power in 2026, up from 20% the year before. And the chips inside AI-optimized servers draw 2-4X more wattage than conventional servers, so a dollar of AI computing costs the electrical grid a lot more than regular cloud computing.

In 2025, total data center electricity grew 17% while AI-specific facilities grew 50%, and that will compound into a bigger problem in a decade.

Then there's the water. Nationally, data center water use is close to a rounding error, but water stress is a local problem, not a national one. Approximately two-thirds of US data centers built since 2022 have been built in high water-stress areas.

None of that is good news.

But what I really wanted to understand was: why are so many data centers popping up right now? What are the "projections" of AI usage even based on?

Where do AI projections come from?

I Googled some form of this question almost daily for 4 weeks, trying to find a well-rounded answer from a trustworthy source (because lemme tell you, JPMorgan has a lot to say and I don't really care to listen).

We hear in the news that more data centers are being built "because of AI demand." But if you've been on my side of the internet, you've seen a lot of discourse from people who have opted out entirely (fair).

So where are these projections coming from?

Between 2023 and 2026, the largest tech companies committed ~$1.6 trillion to building AI infrastructure, and this was based on their most aggressive growth scenario.

Basically, the industry makes their own projections from their own assumptions and builds to the highest number: they're grading their own homework and slapping on a gold star sticker.

And there are real cracks in the demand when you take a closer look. Microsoft signed leases for a ton of data center capacity, then redid their projections and walked away from almost two gigawatts worth; they booked more than they ended up needing.

It gets even more confusing because data center builds are deals that run through several hands. One company develops the site, an investment firm finances it, an operator runs it, and the tech company leases it. Every link in that chain can show their signed contracts as "proof" that the market is hot.

This means that AI data center demand can easily be overstated, and we don't know by how much: the real end-user number is nearly impossible to see.

The data center boom is being built on an overly optimistic projection that assumes a colossal adoption rate that compounds for years on end.

But adoption rates have ceilings: the limit does exist.

But why would they build more data centers than they need?

In simple terms: there are so many financial incentives for these mega corporations to overbuild data centers.

First, the tax code. When a company buys servers, federal rules let them deduct that entire cost in the first year instead of depreciating it across the typical five or six year lifespan. Buy a billion dollars of hardware, write off a billion dollars this year. That setup rewards buying and building data centers regardless of usage.

Then the states line up to waive their sales tax on the same equipment. Virginia gave up roughly $1.6 billion in potential tax revenue last year doing exactly that, up 118% from the year before.

Meanwhile, Henrico County employees were asked to turn off lights, shut down computers, close blinds against the heat, and unplug idle chargers to save on power bills due to rising electricity costs. This is the same county that spent years courting data centers with generous tax breaks, and it's 15 minutes away from where I live.

Next, the accounting. Companies are legally allowed to tell one story to the IRS and another one to investors. To the tax authority, the equipment gets written off immediately, so they take the whole tax benefit up front. To shareholders, that same cost gets spread across the next five or six years, so reported profit stays high, at least for right now.

Microsoft, Google, Amazon, and Meta spent about $434 billion on property and equipment in the 12 months through March 2026, but their earnings reports counted only about $149 billion of that as a cost.

None of that depends on anyone actually using the computing power in those data centers. An underused data center depreciates the same as any other, so to their bottom line, it doesn't matter that much if the demand they've projected for doesn't show up.

Then there's the borrowing. A lot of data centers aren't financed by the original tech company, they're set up as a separate entity with an investment partner. That separate entity takes out the loan to build the data center, and the tech company signs a contract to use it. That means that the debt doesn't sit on the tech company's books, so again, they look "better" to investors.

Meta has an estimated $420 billion in debt structured like this, against about $140 billion in debt on their own balance sheet.

None of this is unique to data centers. It's what being a corporation in America looks like today.

To be clear, I can write off equipment for Bright Body in the same way, but what we can't do as a small business is issue $30 billion of bonds at a rate that makes our debt look trivial.

We can't set up a separate company with a private credit partner so that debt doesn't show up on our own books.

Technically, the tax code is the same, but the tax benefits that have caused the data center boom are only accessible to mega-corporations because of their scale.

Meanwhile, an individual earning a salary is taxed on the whole thing. We can't write off rent, child care, or food, even as those prices are skyrocketing.

When it comes to tech companies and their bottom lines, overbuilding data centers is cheap, and hesitating is expensive. They'll happily build to the most aggressive projections, because their money is protected.

In the USA, innovation comes before regulation (and I hate it)

Railroads took land and timber. Cars took the air, and leaded gasoline stayed in the fuel supply for most of a century after the men manufacturing it started dying. Refrigeration gave us CFCs, which ate a hole in the ozone layer for roughly 60 years before the Montreal Protocol shut them down. The chips inside every one of these data centers get made in factories that use millions of gallons of water.

Every one of those carried an enormous upfront environmental cost, and every one of those had regulations assigned after significant damage was done.

Don't get it twisted: this does not make any of this okay. It just makes it our American reality.

IMO, we need to tax and regulate AI companies and data centers

I'm a systems thinker, and I can't help but think about how our problems with data centers should be solved with systems. We can't expect to "individual choice" our way out of this situation, not when corporate power is what it is.

No more tax breaks. These are among the most profitable businesses in human history, yet state and local governments are handing them incentives and raising residential and school utility rates to cover this strain. That has to stop.

Data centers should pay to upgrade the grid they strain. Transmission, generation, storage, all of it. In November 2025, Virginia regulators approved an $822 million revenue increase for Dominion for 2026 and another $345 million for 2027 to fund the build out of our electrical grid. Most of that is driven by data center demand. And yeah, Virginians are expected to pay that. If your business requires infrastructure that doesn't yet exist, you should pay to fund it. That's exactly how it works when a home developer wants a new road.

We know this can be done because Ohio regulators approved a tariff in 2025 that forced large data centers to pay for a minimum share of the capacity they request, whether they use it or not.

Data centers' water usage should be public, permitted, and reviewed. Right now, local economic development agencies use water usage to sweeten deals inside of negotiations that residents aren't privy to.

Every single one of these mega corporations has the money. They're swimming in it.

I don't think it's a crazy concept to expect corporations to "pay their way" instead of screwing over individual taxpayers.

My personal lines in the sand

At this point I think it's pretty obvious that I've thought extensively about what using AI means from an environmental and ethical standpoint.

Ultimately, I believe that there are missing systems to make this more sustainable and more equitable.

This is not to minimize the impact of individual choices. But every one of us makes choices inside of systems.

The system we're living in now? End stage capitalism. And right now we've got a lil dusting of fascism on top.

My husband just left Henrico County Public Schools (yes, the same county where they told government employees to save electricity) after 10 years as a teacher. Now he's working in Bright Body and Hypema'am.

My family depends on me. And the amount of work I have to manage is - for better or worse - nearly impossible without strategic use of AI.

I'm not even asking you to agree with me. If you don't want to use AI, don't. Truly.

What I wanted to do with this blog is to illustrate that this is a complex and nuanced issue and that has gotten lost in sound bites on the internet.

Now, let me be very clear on a few things:

• I never, ever use AI to make art. Art is human and cannot be replicated by AI. You can quote me on that.

• I never outsource my critical thinking to AI. In fact, I spent months distilling my strategic frameworks and decision-making processes to create custom skills within Claude. Without my business context and thought frameworks, AI was almost useless to me.

• I never use AI in place of people. My goal in using AI skills is to help my team be more human, more creative, and to have more autonomy overall. I will never use AI as a reason for layoffs - no robot can replace human discernment.

So, how do I use AI?

• To systematize + democratize strategic frameworks within my business. Everything that used to swim in my brain about our strategic planning, growth trajectory, and marketing plan is now externalized in a system that anyone in my team can reference at anytime. I'm not the bottleneck anymore.

• To perform repeated tasks that drain my capacity or rely on my memory. The only way I know for sure certain things will get done is if they are systematized.

• To never start from a blank page. Sometimes I'm so stuck in big picture mode that I struggle to bring myself back down into the weeds, even when I need to. Having an external, interactive record of my frameworks and my business means my first drafts are pretty damn close to the final result, and I do less overthinking and more trusting my gut.

• To support clients when they need help at 11 PM on a Tuesday. I'm not on call 24/7, and let's be honest, marketing doesn't require the same working hours as an ER physician. But that doesn't mean that my clients don't deserve some extra support where I can offer it. I spent months training Claude skills on how I think about strategic growth so that they can have support on demand.
  
  Each of my clients gets password-protected access to a chat that incorporates my frameworks with their business context so that implementation is inevitable. I named her Maria Luisa, after my no-nonsense grandma and great-aunt.
  
  If I give you homework to rewrite your welcome series to better align with what your market research said? You can write the first draft with Maria Luisa. How about a level-set on ad performance over the past week? She can do that, too, and much more.

If you use AI, I'd tell you to use it as leverage, not replacement. I'd tell you to know where your data goes and to research how to protect your IP and sensitive data (especially if you're on an individual account).

And I'd also say don't take the shame from either direction.

You're not a dinosaur if you don't want to use it. And you're not a sellout if you do.

The throughline

I would love to not have to play this game at all.

I would love to slowly build a business that supports my family in a country that takes care of its people.

I would love to not have to think about mega-corporations that are too big and too wealthy to be forced to take accountability.

And these United States just ain't that.

Health care is tied to employment, childcare costs more than rent, this administration's organizing principle is deregulation, and they're destabilizing the economy so the richest can get even richer while the rest of us scrape by.

Sometimes you have to use the oppressor's tools to dismantle the systems of oppression.

AI is a tool, and how you use it makes all the difference. What I want is for the tool to be regulated such that it's not taking mental load off me while taking water from somebody else.

The water is real, the electricity is real, the energy costs are real, the IP theft is real, and the pacing of data center buildout is based on fairytale numbers.

And also, as a woman running two businesses inside a broken system, using the best tool available to compete against people who will never think twice about it isn't a moral failing.

I'm not saying that because it's "convenient," it's just the ugly gray reality. It's both-and.

IMO, the moral weight of any of this sits with the corporations first. And when it comes to individual choices, the question is not whether you use AI, it's how you use it and what it replaced.