Does generative AI have value? Yes, clearly. Do the economics of AI make sense? Not really, not yet. Those are separate questions, and mixing them up is why most AI arguments go nowhere.
One person says AI is overhyped and points to the money: huge spending, thin profits, a bubble waiting to pop. Another says it's the real thing and points to their own day: they used it four times before lunch and it saved them an hour.
They're both right, because they're answering different questions.
The first question is about value: does this thing do something useful for the person using it? The second is about economics: can the companies building it charge more than it costs to build and run? A product can be very useful and still be a bad business. It happens all the time.
So when someone asks “is the AI hype real?”, the honest answer is: which part?
I've worked in design for 20 years. The change is not subtle.
Things that used to take a day now take minutes: a first draft of copy, a visual direction for a brand, a working prototype, the code for a landing page. None of it comes out finished. All of it comes out as something you can react to, and reacting is much faster than starting from nothing. Most creative work isn't stuck on skill. It's stuck on the blank page.
The simplest test of value is whether people would miss it. Hundreds of millions of people now use AI chat tools every week, for work nobody forces them to do with it. People don't keep coming back to a tool that wastes their time.
The limits are real too. AI is weak at final quality, at taste and at facts it can't check. It makes a confident mess when you hand it a whole job instead of a step. We wrote about where that line sits in what AI is actually good at for branding. But “not good at everything” is not the same as “not valuable.” Spreadsheets can't write your strategy either.

This is where the skeptics have a point. Look at it from the side of the companies building AI, not the people using it.
The largest tech companies are spending hundreds of billions of dollars a year on data centers, chips and power. In 2024, David Cahn at Sequoia Capital called this “AI's $600 billion question”: the revenue needed to pay back that spending was nowhere close to what AI was actually earning. The gap has not obviously closed since.
Normal software is cheap to copy. Once the app is built, the next customer costs almost nothing. AI doesn't work like that. Every answer, every image, every line of code runs on expensive hardware, every time. The more people use it, the bigger the bill. That looks less like a software business and more like a utility.
In January 2025, OpenAI's Sam Altman said the company was losing money on its $200-a-month ChatGPT Pro plan, because people used it more than expected. Free tiers, generous plans and cheap API prices are partly a land grab: win the users now, work out the margins later.
A new model that costs a fortune to train gets matched by competitors, sometimes by free open-weight models, within months. Prices keep falling. That's great for you. It's hard on whoever paid to be first.
This pattern is old. A new technology is clearly useful, money floods in faster than the revenue can justify it, a lot of investors lose, and the technology stays.
In the 1840s, Britain went through a railway mania. Investors poured money into lines that would never pay off, and many were wiped out. The tracks got built anyway, and trains ran on them for the next hundred years.
In the late 1990s, telecom companies laid far more fiber-optic cable than anyone needed. Several of them went bankrupt, including WorldCom and Global Crossing. Their cable got bought for cheap, and a few years later it carried streaming video and the rest of the modern internet.
In both cases the bubble was real and the value was real. The people funding it lost. The people using it won. A bubble is a statement about prices, not about whether the thing works.
If you run a business and use these tools, the economics question is mostly someone else's problem, with a few practical consequences.
And remember which part is cheap. Generating an image is getting cheaper every month. Deciding what your brand should look like is not. The output is a commodity. The decision is the valuable part, and it belongs to you, whichever tool made the pixels.
We see both questions every day. The value side: people describe a project in a sentence and get a full visual direction back that they can actually use. The economics side: every board we generate runs on paid image models, and we pay for each one.
That's why MoodyBoards isn't free forever. We'd rather charge a fair price for something useful than give it away and hope the money works out later. It's the same question the big AI companies are facing, just at a smaller size.
Is the AI hype real? The value is real. You can see it in what people do with these tools every day. The economics are shaky, and some of the money going in now won't come back out.
Those two facts don't cancel each other out. Use the tools for what they're good at, and don't confuse a company's stock price with how useful its product is.
In the financial sense, possibly. Spending on AI is far ahead of the revenue it brings in, and some investors will likely lose money. That says nothing about whether the tools are useful. They clearly are.
Yes. It turns hours of first-draft work (copy, visual directions, prototypes, code) into minutes. It's weak at final quality, taste and checking facts, so it works best as a fast starting point you then judge and edit.
Unlikely. The models already exist, open-weight models run on hardware anyone can rent, and prices have been falling. Expect plans and prices to change, not the technology to go away. Railways and fiber-optic cable outlived the companies that went broke building them.
Because every answer or image runs on expensive hardware each time it's made. Normal software costs almost nothing to serve to one more user. AI costs real money for every use.