It's all just a gamble
3DP War Journal #85
I’ve been deeply exploring the topic of AI hype lately. On one hand, I’m reading financial analyses of the leading companies creating this market, and on the other, I’m trying to understand the essence of artificial intelligence and machine learning.
The conclusion of my research is brutal - the entire AI sector is one giant speculative bubble that will burst sooner or later with a loud bang.
Its symbol and stronghold - OpenAI - is like a sandcastle built on the very edge of the beach, waiting for the approaching ocean tide.
All these things that mainstream media writes about and world-class authorities emphasize:
the greatest civilizational revolution since the invention of the wheel,
job losses for millions of people who will be replaced by AI,
the collapse of existing education models, because humanity will use only homogeneous, hive-knowledge generated by ChatGPT, Grok or Gemini
well, this won’t happen. Or at least not in the described form.
And why not? Because all of this is exactly the same as what has been said dozens of times before…
In the context of 3D printing. In the context of augmented reality. In the context of cryptocurrencies. In the context of... graphene.
That’s right, hands up who still remembers graphene? The super material that 10 years ago was supposed to replace all materials? Has anyone heard anything about graphene lately...?
No Pawel, you don’t understand... You’re comparing this wrong. This can’t be compared. AI is something much, much more. There’s never been anything like this before.
This is the thing. This is the revolution.
Besides, look at the money! Someone who packs billions of dollars annually into this sector can’t be wrong.
They can’t, right...?
No. Wrong. These people are wrong all the time. They pack this money irrationally. This isn’t business. This is gambling.
Just like in the 2000s with mortgage loans. The Big Short. Margin Call. Too Big to Fail.
You’ve been sold short to a scam
In autumn 2024, I published on Medium one of my more prominent articles describing the mechanisms behind the creation of the great speculative bubble in industrial AM between 2018-2023 (the article was later republished on VoxelMatters after I left Medium).
On one hand, I presented the reasons why AM companies decide to push boundaries in terms of their actual capabilities, promising things that are impossible to fulfill. Origins of „3D printing poverty” and the inability to earn money adequate to the amount of work invested.
On the other hand, I described the short-sightedness and infantilism of investors who, in pursuit of easy, quick profit, don’t want to listen to all those difficult - and horror of horrors! - paid analyses exposing the lies or manipulations written in the prospectuses of aspiring unicorns.
Essentially, both sides deserve each other. They deceive each other. They deceive others around them, wanting to justify their behaviors and decisions.
They create narratives that with each subsequent round of financing become further and further from reality.
It all starts with a lack of understanding of what the product and technology behind a given startup are. That is, the investor knows “more or less” what it’s about, but in making decisions is guided by “growth rate” and the degree of “impact on the economy” of the technology. How it actually works is secondary to them.
The perfect example here is my beloved Desktop Metal. A true masterpiece of technological hype and the most detached from reality.
At the turn of 2010 and 2020, the company promised investors a revolution in industrial metal production. It was supposed to replace - or at least significantly displace - existing traditional manufacturing techniques. This was to be accomplished through an innovative method - BinderJetting, involving selective bonding of metal powder with a binder, and later sintering finished parts in a furnace.
Super fast, super efficient.
The problem was that:
this method wasn’t innovative... it was created in the mid-90s and was successfully developed by ExOne; which Desktop Metal immediately acquired after obtaining record funding from the stock market in early 2020s
this method was burdened with a series of limitations, the most important of which was significant part shrinkage that occurred in the sintering process; this forced upward scaling of parts before their production on the 3D printer, which was an extremely complicated mathematical operation; and in the case of some geometries made this method completely useless
this method wasn’t scalable at all... the cost of a single system was inadequately high relative to production output; the build chamber was large by 3D printing standards, but tiny in the context of metal parts production.
And finally, most importantly - in the context of Desktop Metal’s original R&D work, it’s metal BinderJetting never really worked... To bring machines to the market, company first resorted to licensing solutions from Taiwan’s XYZPrinting, and then the technology of acquired ExOne.
Eventually, this entire Mass Additive Manufacturing hype collapsed under its own weight.
Operating costs were disproportionately high compared to generated revenues. The entire metal 3D printing sector turned out to be simply unprofitable. Technologically limited for the expectations of the manufacturing sector.
I wrote about this here:
According to the definition: 3D printing is a manufacturing method that involves applying material layer by layer and selectively bonding it together. Compared to all other production techniques, it has three advantages and three disadvantages that significantly interact with each other:
Advantages:
is the fastest
is the cheapest
allows the production of geometries that would be impossible to produce otherwise
Disadvantages:
has the poorest accuracy and finishing quality
becomes problematic as part sizes increase
not profitable for mass production.
When do we use 3D printing?
When we need to produce one or several things.
When we want something done quickly and cheaply, and quality is of secondary importance.
When we came up with something so specific, unusual and complicated that only a 3D printer would be able to make it.
Let’s summarize:
low quantities
very specific
cheap
quality is not the most important thing.
It cannot be compared to injection molding. Neither for CNC milling. So you can’t have the same business expectations for 3D printers as you do for injection molding lines.
And exactly the same applies to AI…
A wonder no one understands
What is artificial intelligence? The problem isn’t even that few people know - the problem is that nobody tries to understand it. Nobody asks about it.
One day AI has become as obvious as electricity or the internet. People have decided it simply is here to stay.
And this shouldn’t be surprising. If you asked the average person where electricity or the internet comes from, the answers would probably be: from an electrical outlet and from the nearest 5G tower.
So where does AI come from? From OpenAI. Or some other “manufacturer”.
Some people are excited about it, and others are afraid of it. But neither group knows what it is. It simply appeared, it works, and depending on the context, it’s either wonderful or terrifying. So let’s start from the basics…
I often repeat that 3D printing is not a new technology because it was created over 40 years ago, and conceptual work on it dates back to the 70s (or 60s, depending on the definition of what was being worked on then).
Meanwhile, AI is even older. I recommend David William Silva’s article on this matter:
The term “artificial intelligence” was coined by John McCarthy in 1955. McCarthy, joined by Marvin Minsky, Nathaniel Rochester, and Claude Shannon (what a group!) submitted “A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence” to the seminal event for artificial intelligence.
Alan Turing proposed the Turing Test in 1950 to measure machine intelligence. Allen Newell and Herbert Simon built the Logic Theorist. Minsky advanced neural networks. Shannon gave us information theory.
Therefore, work on AI has been ongoing for 70 years. Nothing was created here overnight.
What’s more, this may shock you but I’ll write it anyway:
AI is not magic – it’s ordinary mathematics on a massive scale
at its foundation lies addition, multiplication, averages, and probability; algorithms gradually adjust numbers when the computer makes mistakes - that’s what “learning” means
the system processes enormous amounts of data (text, images, recordings) billions of times, fine-tuning itself; fast hardware transforms this into a pattern-recognition model
and so, for example, the popular AI Agents that allegedly are successively taking jobs away from talented people in corporations are even simpler: it’s a model with a task list and tools, executing ordered steps without internal consciousness.
AI has one main limitation: it’s a statistical summary of the past, and doesn’t think creatively. It doesn’t think at all in the sense of how humans think.
When you propose an unconventional idea, AI rejects it - not because it’s bad, but because it doesn’t resemble the training data. What it does is pattern recognition. This is not intelligence.
Due to gigantic databases (consuming gigantic energy and water resources), AI works very quickly, but... very sloppy. As DW Silva writes about it:
It is the most productive intern you’ve ever had, one who never sleeps, never complains, and many times fails to check whether the work is actually correct.
And we are building entire industries on top of this.
It’s an impressive engineering achievement, but not a miracle or an independent, creative mind.
Therefore, the hype around AI primarily serves obtaining further rounds of financing for further development (and survival).
Okay, but it’s a fact that AI is taking our jobs away from us humans…
This isn’t entirely true. I wrote about this two weeks ago:
A report by Oxford Economics shows that companies are not replacing workers with AI at any significant scale. Instead, the AI narrative is being used to mask financial difficulties and ordinary workforce restructuring.
Youth unemployment is indeed rising, but AI is not the cause. The real drivers are a surge in the supply of graduates - 32% to 35% in the US, 39% to 45% in Europe - and broader economic slowdown. In countries like South Korea and Japan, which are implementing AI but not experiencing recession, graduate unemployment is not rising.
This is a classic attribution error: correlation in time does not imply causation.
And even if it does, it ends in catastrophe. According to Will Lockett:
Microsoft turned around and laid off an astronomical number of employees, totalling 15,000 by the end of the 2025! Some of these branches were heavily impacted, with coders accounting for 40% of the layoffs. To many, it looked like Microsoft was replacing its software engineers with AI.
And then the glitches started to happen.
Over 2025 and into 2026, every Windows update seemed to have an even bigger bug than the last. Performance slowed, apps failed, cloud storage glitched out, core features broke, recovery tools failed, and updates completely bricked machines and caused boot failures.
Microsoft has had to issue a dramatic number of patches and workarounds to try to address this tsunami of screw-ups, but the damage has been done. Users are fed up and jumping ship from Windows to Mac or Linux. In fact, it appears that Windows has lost 400 million users since 2022!
Now imagine Ford, Toyota, Boeing, or Siemens factories that massively lay off their CNC operators and instead pack all halls with Production Systems from Desktop Metal. All parts stop being milled - they’re 3D printed and sintered.
You can’t imagine this...? But somehow you can imagine AI doing people’s work…
Why is so much money being invested in this thing then?
If someone like me, who until January of this year wasn’t very interested in AI, was able to obtain the above knowledge and recognize identical speculative patterns as in previous investment bubbles, why don’t people investing millions - billions of dollars see this?
The answer is banal... Because they don’t want to. Because they don’t need to. Because in this game it’s not about whether it really works or does not.
Because it’s a game. A casino. Gambling.
What counts is adrenaline. Prestige. Recognition from peers. And greed.
I’ll explain it this way:
for the average person - a representative of the so-called middle class, 1 million dollars is a very large sum of money
10 million dollars is a sum that will allow such a person not to work for the rest of their life (provided they maintain a slightly elevated standard of living than now and don’t spend half on cars, drugs, and parties in the first year)
100 million dollars is a sum that for the average person is already abstract; to lose such a sum requires exceptional creativity; 100 million dollars is also a fairly high sum even for a millionaire
500 million dollars is an unspendable sum
1 billion or more means that money is already lifelong infinite.
What to spend 10 million on when you have 500?
What to spend 200 million on when you have 2 billion?
On roulette in Las Vegas? Nah... You’d have to be a degenerate gambler. No, such money must be invested!
In what and how? Maybe in new technologies? Great idea! But which ones?
And then an investment advisor appears, or someone from an investment fund, or the founder themselves. And they present a plan. A plan that’s fucking clever! A plan with a high rate of return!
For example, in revolutionary metal 3D printing technology that will turn the entire industrial sector upside down. Or artificial intelligence that will replace humans at work. Or whatever…
And now the most important question: will a person who earned (or inherited) 500 million dollars delve into the technological nuances behind the metal BinderJetting method? Will they check whether some ExOne didn’t develop it some time earlier?
Or maybe they’ll start delving into scientists’ work on artificial intelligence from the 1960s or 70s? Or reach for great but difficult to read Nick Bostrom’s books?
No... It doesn’t work that way... After all, if a roulette player analyzed the essence of this game, sooner or later they would discover that its foundation is the casino’s victory. And then they wouldn’t play…
It’s all a game. It’s all a gamble.
And when the bubble burst, the vast majority will lose money.
Just like in the 2000s with mortgage loans.
The Big Short.
Margin Call.
Too Big to Fail.