I stumbled into this rabbit hole completely by accident. There I was, scrolling through my research feed like any self-respecting blogger drowning in AI-startup announcements, when I noticed something weird. Every other post was about some miraculous new AI company that was going to “revolutionize industries” and “transform how we work forever”.

The thing is, that after diving deep into this world for months, I realized something that would make your head spin: More than half of these so-called AI companies aren’t actually building artificial intelligence at all. They are nothing more than really expensive middlemen with fancy websites.

Let me explain how I figured this out, and why it’s both hilarious and terrifying.

---

More rants after the messages:

Join me and Eliska Went on September 5th when we show you how to create your own automated AI tool to find, write and apply for jobs. Register here. Here, I mean: here

1. Comment, or share the article; that will really help spread the word 🙌
2. Connect with me on Linkedin 🙏
3. Subscribe to TechTonic Shifts to get your daily dose of tech 📰
4. Visit TechTonic Shifts blog, full of slop, I know you will like !

---

The pizza box revelation

Picture this scenario, and it will all make sense. Of course you know that incredible Italian restaurant downtown with the brick oven and the chef who trained in Naples. Now, for the sake of this piece, let’s call them “OpenAI’s Kitchen”. They make the most amazing pizza you have ever tasted. The crust is perfect, the sauce is divine, and they can make literally any pizza you can imagine.

Now, imagine I decide to start a business.

But instead of learning how to cook or how to build my own kitchen, I just go over to OpenAI’s Kitchen every time someone orders from me, I take their pizza, slide it into a box with my logo printed on it, and I deliver it to you for triple the price. I call myself “PizzaGPT” and tell everyone I’m “revolutionizing the food industry with artificial intelligence”.

Now, that is basically what a lot of people have seen happening with most AI startups right now, except for pizza, we’re now talking about intelligence, and instead of being honest about it, they’re raising millions of dollars and acting like they themselves invented it.

---

My accidental detective work

This whole revelation started when I signed up for this slick podcast editing tool. The marketing was gorgeous – they promised to turn my rambling 1-hour interview into social media gold, complete with viral clips, engaging captions, and newsletter content that promised to make my subscribers weep with joy, and all-o-that for the low, low price of $60 a month.

Being the curious (and slightly paranoid) person I am, I decided to peek under the hood, and what I found made me laugh so hard I nearly spit coffee on my new Chinese laptop.

The entire “revolutionary AI platform” was basically calling OpenAI’s API (programmer’s interface) with some fancy prompts. I could literally recreate their entire workflow in about 10 minutes for under $4.

Not $4 a month – $4 total.

For the same results.

This wasn’t innovation.

This was the equivalent of buying the pizza, putting it in a prettier box, and charging restaurant prices for delivery. Except somehow, somewhere, a company convinced investors that this was worth millions of dollars.

---

The great API gold rush

Once I started looking, I couldn’t stop seeing it everywhere. The market is absolutely flooded with these companies, and they all follow the same playbook:

1. Take OpenAI’s (or Anthropic’s or Google’s) AI brain
2. Write some clever prompts to make it do specific tasks
3. Wrap it in a clean, user-friendly interface
4. Add some buzzwords about “machine learning” and “neural networks”
5. Charge users 10-50 times what it actually costs to run
6. Profit (at least until someone notices)

The success stories are pretty wild though. . .

There’s PhotoAI making $77,000 a month, Chatbase pulling in $70,000, and InteriorAI bringing home $53,000 monthly. PDFai is “only” making around $30,000 a month, which honestly sounds like failure in this ecosystem (and I’m using PDFGear, which does the same and is for free).

But when I did the math on what these companies actually pay for the AI they’re using, the markups are insane. I am talking 1,833% to 3,800% markups in some cases, and one company was charging $199 a month for something that cost them $7.50 in actual AI usage.

Now that is very close to theft and fraud if you ask me.

The thing that really gets me though, is that many of these wrapper companies are being deliberately misleading about what they are actually doing. They use terms like “proprietary AI”, “advanced machine learning algorithms” and “neural network optimization” to describe what is essentially just a well-written prompt sent to OpenAI’s API.

It’s like if I bought a Tesla, painted it a different color, and then tried to sell it as “my revolutionary new electric vehicle powered by proprietary autonomous driving technology”.

Technically true. That’s a fact.

Their business model relies on users not understanding how simple the underlying technology really is. They’re banking on the fact that “AI” still sounds magical and complicated to most people.

---

The smart ones vs. The pizza box brigade

Now, before you think I’m completely cynical about this whole thing, let me clarify something important. There are actually two very different types of companies in this space, and the difference matters a lot.

The Pizza Box Brigade consists of companies that are literally just reselling someone else’s AI with minimal added value. They are the ones that will disappear fast once OpenAI or Google decides to add their “unique” feature directly to ChatGPT or Bard.

But then there are the Smart Ones.

These are the companies who are using AI as an ingredient, not as their entire recipe. They’re adding their own secret sauce in ways that actually matter, like um. . .

• Industry expertise. Say you develop a legal AI tool that understands the specific nuances of contract law and has been trained on thousands of legal documents like Harvey AI for legal precedents, or LawGeex for contract analysis.
• Proprietary data. For instance a company that has collected unique datasets from their customers that make their AI responses way more accurate for specific use cases. Think Netflix, or C3 for Energy sector data, Veeva for pharma data, or Grammarly that has been trained on billions of writing improvements.
• Deep integration. Think about tools that plug so into your existing workflow that switching would be genuinely painful. Think of Microsoft Copilot, or Hubspot AI or Salesforce Einstein, but also Workday’s or Notion’s AI features, etc.
• And multi-model magic. Platforms that intelligently combine different AI models depending on the task, optimizing for cost and performance, like Genspark and Manus AI, or Martian that automatically routes queries between GPT, Claude, Gemini and Llama based on complexity and cost.

The Smart Ones are taking the pizza, adding their own ingredients, cooking it further, and creating something you genuinely can’t get anywhere else.

---

The economics are bonkers

The investment numbers in this space are absolutely bonkers. In 2025 alone, AI startups have raised over $26.9 billion, which represents 64.1% of all venture capital funding. That’s not a typo – nearly two-thirds of all startup investment is going to AI companies.

But here it gets really interesting. Over 50% of that AI funding is going to companies that don’t build their own AI models. They’re funding the Pizza Box Brigade at ridiculous valuations.

I found companies being valued at 20-44 times their annual revenue, which would make even the most optimistic crypto investor feel embarrassed. For comparison, most normal software companies trade at 5-15 times revenue.

And the reason for all-o-that is of course FOMO (Fear of Missing Out) which now has reached epidemic proportions in the VC world because Nobody wants to be the investor who missed “the next big AI company”, so they’re throwing money at anything with “AI-powered” in the pitch deck, even harder than they did in the beginning of the hype.

---

The coming wildfire (and why it might be good)

Now here’s where things get really juicy. Multiple experts are predicting that 70-90% of current AI wrapper companies will be dead by 2026. That’s not me being dramatic – that’s coming from people way smarter than me who study this stuff for a living.

The reasons are pretty obvious when you think about it:

Platform integration: OpenAI, Google, and Microsoft aren’t stupid. They see these wrapper companies making millions by adding simple features on top of their APIs. Why would they let that continue when they could build those features directly into their own products?

Price wars: As AI gets cheaper and more accessible, those massive markups become impossible to justify. When users realize they can get the same results for 95% less money, the jig is up.

User education: People are getting smarter about AI. The early adopters who didn’t know any better are being replaced by users who understand what’s actually happening under the hood.

Open source revolution: Free and open-source AI models are getting really good, really fast. Why pay $60 a month for a wrapper when you can run similar models locally for free?

But this consolidation might actually be a good thing.

The Pizza Box Brigade disappearing will make room for the Smart Ones to thrive, and the companies that are actually building something valuable and defensible will have less noise to compete with.

---

What this means for us regular hoomans

If you’re not building an AI company but you’re using these tools – as a private person or as a company – this is what you need to know. . .

Before you subscribe: Ask yourself one simple question – “What is this company actually doing that I couldn’t do myself with one of the existing AIs out there” If the answer is “making it prettier and easier”, you might want to think twice about that $500/month subscription for your teams.

Look for real value: The best AI tools are the ones that solve specific, complex problems in your industry. They’ve been trained on relevant data, they integrate with your existing tools, and they get better results than you could achieve with general-purpose AI.

Plan for changes: If you’re building your business around one of these tools, have a backup plan. The company you’re relying on might get acquired, shut down, or steamrolled by a platform feature update.

---

My prediction game

Based on everything I’ve learned, and the research I’d done for this piece, here’s what I think will happen next:

2025-2026 THE GREAT DYING: Expect 70-90% of current AI wrapper companies to either shut down, get acquired for pennies, or pivot desperately to something else. The easy money is over.

2026-2027 THE REAL COMPETITION: The survivors will be companies with genuine moats and differentiation. They’ll compete on value, not on who can mark up APIs the highest.

2027+ AI GOES NATIVE: We will stop thinking about “AI companies” as a separate category, because “AI” will be a feature that every software company has, like search or user authentication today.

Now it is time for the grown-ups to build something that lasts.

Signing off,

Marco

---

I build AI by day and warn about it by night. It’s job security. Big Tech keeps inflating its promises, and I just bring the pins.

---

Think a friend would enjoy this too? Share the newsletter and let them join the conversation. Google and LinkedIn appreciates your likes by making my articles available to more readers.

To keep you doomscrolling 👇

1. The chicken that refused to die | LinkedIn
2. Vibe Coding is gonna spawn the most braindead software generation ever | LinkedIn
3. The funniest comments ever left in source code | LinkedIn
4. Today’s tech circus is a 1920s Wall Street reboot | LinkedIn
5. The Sloppiverse is here, and what are the consequences for writing and speaking? | LinkedIn
6. Why we’re all writing like shit on purpose now | LinkedIn
7. Creating a hero’s journey for your AI-strategy | LinkedIn
8. How corporations fell in love with AI, woke up with ashes, and called It ‘innovation’ | LinkedIn
9. The corporate body snatchers | LinkedIn
10. Screw your AI witch hunt, you illiterate gits | LinkedIn
11. AI perverts make millions undressing your daughters for fun and profit | LinkedIn
12. The state of tech-jobs 2025 | LinkedIn
13. Meet the sloppers who can’t function without asking AI everything | LinkedIn