Why TVs Got So Cheap, and Why AI Needs Them Everywhere

It started with a receipt in my hand: a TV cheaper than my phone, and a quiet trade I never agreed to.

Recently, I bought a 65-inch 4K TV for less than I spent taking "part" of my family out to dinner.

Not refurbished. Not a doorbuster with a line out front at 5 a.m. Just a normal smart TV, off-the-shelf, cheaper than one nice meal for a household our size. And I stood there holding the receipt with this little itch in the back of my head that I couldn't scratch.

Because here's the thing nobody says out loud: nobody gives away hardware. A glass panel that big, shipped across an ocean, with a computer bolted to its back, costs real money to make. I've run businesses long enough to know that when something is priced as if it's almost free, the product you're looking at usually isn't the one you're paying for.

So I got curious. Genuinely curious, the kind that keeps you up a little. Why do TVs keep getting cheaper every single year while everything else in my life (groceries, gas, insurance for nine kids' worth of eventual drivers) keeps going the other way? What's the catch? And why does nobody seem to be asking?

I do what I do now with anything that nags at me: I hand it to AI and tell it to go find the real answer. I want to walk you through what I found. But I also want to show you where the AI lied to me with a completely straight face, because after this year, I'm convinced that the second part is the more important lesson.

First, this isn't in your head

Before the "why," I wanted the "how much," because I didn't trust my own memory. Turns out the government tracks this. The Bureau of Labor Statistics keeps a price index for televisions, and it's the honest kind of number: it holds quality steady, so it measures the price of the same TV over time, not just "TVs got bigger."

By that measure, an equivalent television got roughly 30% cheaper in five years. And the year-by-year story is even better than the headline, because prices dropped every single year, except one.

That one exception, 2021, is my favorite part. For a single year, TVs got more expensive, about 6% up, because a pandemic snarled the supply of panels and shipping. And I love that, because it quietly proves the whole point: this isn't gravity. Prices don't fall because of some law of nature. They fall because someone, somewhere, decided to make them fall, and the one time the machine jammed, the price went the other way. When it un-jammed, prices didn't just resume falling. They fell harder.

So something is pushing. I wanted to know what.

Why the box got so cheap

Most of a television's cost was never the brand or the software. It was the panel, the sheet of glass itself. And over the last decade, that panel has turned into a commodity, much like flour or plywood. A handful of enormous, heavily state-subsidized manufacturers in China now make around 60% of the world's display capacity, and they'll sell panels at prices their competitors literally can't survive. Last year, LG sold off its final big LCD TV factory. The race to the bottom was over, and China won it.

Once the expensive part is a commodity, nobody can make money selling the box anymore. And they're not even trying. Roku, over a recent stretch, sold its hardware at a negative 7.6% margin, meaning it lost money on every device, on purpose. A former CTO at Vizio said the quiet part into a microphone at a tech conference: "I really don't need to make money off of the TV. I need to cover my cost… you sell some movies, you sell some TV shows, you sell some ads."

The TV isn't the product they're selling you. The TV is the bait they use to get the real product into your living room.

So what's the real product?

You. Or more precisely, everything your television can learn about what you watch.

Most modern smart TVs run something called ACR, automatic content recognition. Here's what it actually does, and this is the part that made me put my ZOA down: the TV takes tiny screenshots of whatever is on the screen, over and over, and matches them against a giant reference library to identify exactly what you're watching. Samsung's own advertising materials describe grabbing a snapshot of the screen every 500 milliseconds. Twice a second. All day.

And it doesn't care where the picture comes from. Cable, streaming, a DVD, and the news. Even when you're using the TV as a plain monitor for a game console or a laptop, the screen-watching keeps going. It only stops if you dig into the settings and turn it off, which almost nobody does, because almost nobody knows it's there.

I kept waiting for the part where this turns out to be a paranoid exaggeration. It doesn't. Back in 2017, the Federal Trade Commission caught Vizio doing exactly this on 11 million televisions, collecting second-by-second viewing, as many as 100 billion data points a day, then attaching it to household details like age, income, and marital status and selling it on. Not a rumor. A federal settlement. And this spring, it got current again: Texas sued Samsung, LG, Sony, Hisense, and TCL over the same kind of tracking, and Samsung actually switched ACR off, for Texas viewers only.

Where AI comes in, and where it lied to me

Here's the connection I was actually chasing. If the data is the product, then the business only works if the TVs are everywhere. Not in wealthy homes. Everywhere. Every income level, every spare bedroom, every kid's room. A loss-leader only pays off at saturation, which means the whole model quietly depends on the hardware being cheap enough that literally anyone can afford it. The low price isn't generosity. It's customer acquisition for a data business.

And AI is what turns that raw viewing data into money at a scale that wasn't possible before, by stitching viewers into detailed profiles and generating swarms of individually targeted ads, tuned to the person on the couch. The cheap TV feeds the data. The data feeds the AI. The AI makes the data worth more, which justifies making the TV even cheaper. Round and round.

Now, the part I promised. When I first pointed AI at this whole question, it handed me a clean, confident summary that said, in effect, AI made TVs cheap. And that's wrong. Backwards, even. Subsidized factories made TVs cheap. AI just showed up to feast on what the cheapness made possible. The AI's most confident sentence was its least correct one, and if I'd just published it, I'd have taught all of you something false with total conviction.

It happened again a few minutes later. Two of the juiciest numbers it gave me (the exact ones I most wanted to be true) traced back to some random blog with nothing behind it, delivered in the same authoritative tone it used for an actual federal court filing. Same confidence. Wildly different reliability. And it never flagged the difference. I had to.

AI is a phenomenal researcher and a terrible editor of its own certainty. The job didn't disappear. It moved: from finding the answer to interrogating the machine that hands you one wearing a lab coat.

I'm still figuring out how to do that well, honestly. But the pattern I'm landing on is simple: the more confident the answer sounds, the harder I now push on where it came from. The confident, tidy, quotable line is exactly the one most likely to be subtly wrong. That's not a reason to stop using these tools. They found me a peer-reviewed study and a decade-old FTC case I'd never have dug up on my own. It's a reason to stay in the chair as the editor, not hand the whole desk over.

What I actually want you to do with this

Don't throw out your TV. I'm not going to throw out mine. The thing is genuinely great, and the free ad-supported channels it unlocked are a real gift for a big family on one screen budget. Here's the honest version: for plenty of households, this might be a fair trade. Cheap hardware, free shows, in exchange for data you don't particularly care about. That can be a perfectly good deal.

The problem was never that a trade exists. The problem is that the trade is hidden: defaulted on, buried three menus deep, and made on your behalf before you knew there was anything to decide. With nine kids growing up with screens in this house, what bothers me isn't the ads. It's the quiet lesson that being watched is just the normal background hum of owning things. I'd rather they learn to notice the trade and choose it on purpose.

So that's the whole ask. Go into your TV's settings tonight and find the ACR option. It'll be called something friendly like "viewing information" or "smart interactivity." Look at it. Then decide, on purpose, whether you want it on or off. Somebody reading this needed to know that the switch existed. If that's you, that's the whole reason I wrote this down.

One more thing, because it's already happening again

Here's what I can't stop thinking about. The TV was the warm-up.

The exact same move (give the thing away, because the recommendation is the real product) is happening again, one screen closer to your face. For twenty years, searching for a plumber or a security company got you ten links and a choice. Ask an AI assistant the same question now, and you get two or three names and no page two. The box is free. The answer is the product. And which companies get named, or don't, is quietly becoming the whole game.

Most business owners have no idea whether the machine can even see them yet. That's the thread I'm pulling next, and it's a big one. But it started, for me, with a receipt in my hand and a TV that was somehow cheaper than my phone.

Come build alongside me. I document these rabbit holes as I go: what the AI got right, where it confidently steered me wrong, and what I'm still figuring out. If that's useful to you, join the lab. Someone out there needs the next one, too.

Where these numbers come from: FTC v. Vizio (2017) · Lobato, Convergence/SAGE (2025) on ACR · Texas ACR suit & Samsung (2026) · Roku/Vizio hardware margins · Vizio CTO quote · Chinese panel dominance · BLS CPI for televisions (quality-adjusted). Figures on AI-generated ad personalization draw on industry and advocacy reporting and are directional.