MIT report says 95% of AI implementations don't increase profits, spooking Wall Street

[Daniel Sims]

The big picture: Warnings that the generative AI market may be forming a dangerous bubble have intensified in recent weeks. With big tech and AI investments playing an outsized role in today's stock market, many are starting to question where the returns are.

Sell-offs in numerous tech companies dragged global markets downward this week after a report from MIT's NANDA initiative estimated that almost no AI startups have achieved measurable gains from the technology. The findings add weight to growing warnings that an AI market crash could be looming.

The report, titled The GenAI Divide: State of AI in Business 2025, analyzed 300 public AI implementations, surveyed 350 employees, and interviewed 150 business leaders. Researchers concluded that only about five percent of companies piloting AI saw rapid revenue growth, while the majority stalled.

Following the report's release, shares of companies such as Nvidia, Palantir, Arm, Oracle, AMD, and AppLovin slipped by single-digit percentages on Tuesday. The news pushed the Nasdaq down 1.4 percent, the S&P 500 by 0.7 percent, Stoxx Europe by 0.6 percent, Japan's Nikkei 225 by 1.5 percent, and South Korea's Kospi by 0.6 percent.

Aditya Challapally, lead author of the MIT report, told Fortune that the few successful AI startups typically focus on one specialized area and maintain close collaboration with clients. Some have grown revenue from zero to $20 million in just a year.

The report also found that while most AI investment has gone toward marketing and sales tools, back-end automation delivers the highest returns. These include solutions for streamlining outsourcing and other operational processes.

Additionally, while most companies try to build their own AI tools, those that purchase solutions from specialized providers are far more likely to succeed. The most effective tools adapt to specific parameters, unlike general-purpose systems such as ChatGPT.

OpenAI CEO Sam Altman recently acknowledged that many new AI firms attracting venture capital are setting unrealistic expectations, warning that a resulting bubble could lead to historic losses when it bursts. His comments echo those of economist Torsten Slok, who has argued that the AI boom is more overhyped than the dot-com bubble of the 1990s.

Last year, Financial Times columnist Robert Armstrong noted that the S&P 500 was weakening outside of Nvidia and other AI-focused chipmakers. More recently, tech analyst Edward Zitron argued that sales of Nvidia GPUs – responsible for nearly eight percent of US stock market value – remain the primary source of AI-related profits. Any disruption in this feedback loop could have major repercussions for the handful of tech giants that make up more than one-third of the market.

Permalink to story:

MIT report says 95% of AI implementations don't increase profits, spooking Wall Street

 

[Theinsanegamer]

MIT saying what we already knew.

 

[Revolution 11]

MIT saying what we already knew.

The famous last words of every bubble ever: "This time is different."

When you hear that, go ahead and order a tombstone with the words on it for whatever is being talked about.

 

[Alfonso Maruccia]

Nice.

 

[trparky]

 

[Jack77]

MIT saying what we already knew.

Like always, corporate management is the last to gain insights. And they need reports of reputable organizations for that.

 

[GeoffreyA]

Get the popcorn and cool-drink out.

 

[redgarl]

The famous last words of every bubble ever: "This time is different."

When you hear that, go ahead and order a tombstone with the words on it for whatever is being talked about.

I am sure that every time this was brought forward, we were talking about the technology behind the 4th industrial revolution... right?

Ah yeah, I thought so...

Matter of fact, the order books at TSMC, Nvidia and AMD have never been so filled...

 

[redgarl]

People think AI was made to make more profits, when in reality, AI goal is to help alleviate administrative processes.

Ironically, AI is not really for the private sector, but for the public sector. Every government will develop their own custom AI cluster for national security reasons. If nobody can see the sign on the walls, then you are blind. When you have the world government having clear stance on the subject, which is in their priority list, then you understand what this is all about.

If you are mistaking AI and Generative AI as a simple chatbot, then of course you will call this a fad. However, AI will develop more than just a chat bot, it is already transcending into robotics which will shift the manufacturing business even more. It will also transcend government work to a level that no politician could do even with drastic policies.

Maybe you don't see the future because you are trapped into your safespace with your bias because you are mad that gaming hardware get more expensive as a collateral effect, but denying AI relevancy by just stating it is a bubble is ridiculous.

This bubble is on its 3rd year and it is driving the whole economy right now.

Matter of fact is Nvidia, AMD and TSMC will keep selling more and more chips year after year.

Lastly, the Nasdaq was down by 0,34% today, not 1.4%...

Not to mention what is really having the attention of the investors right now is Powell speech on inflation and the interest rate cuts.

 

[BuckarooBonzaii]

I tend to see any bubble as a way to separate the weak from the strong. The weak will fall, the little more intelligent ones will back out and reap their profits. The strong ones will continue on maybe acquire those they see that are good, and when the bubble pops the strong will have gotten what they want and have moved on. In a few months, years the bubble has been forgotten and it's business as usual until the next bubble. Corporate contracts many psychology firms to analyze and forecast that costs money, and those who can pay are the only ones who can play.

 

[Mark Fuller]

MIT saying what we already knew.

All AI every really was is a vehicle to tell anyone what they would find if they went ahead and looked it up for themselves. Dictionaries, encyclopedias and technical manuals were always readily available. I remember when the concept of fuzzy logic came from the selfsame institution of MIT. It was supposed to be a computer program that would tell you what to do about a problem by somehow taking into consideration all of the intangibles and then making the correct decision. Then someone realized that every once in a while a failure will surely occur given the fact that Murphy's Law has nothing to do with intangibles and the fact that if you want to go against pure logic you just make the opposite decision. Someone will always turn in the wrong direction at the wrong time or throw a wrench in the works. I think it was just an excuse to blame something else when something went wrong.
AI will never do anything for anyone that can't be done according to the laws of science anyway. It will never get you more profit from the strictest laws of accounting processes or make anything that can't be done. I dare say that it will never even make a Spiderman from a radioactive spider.
To me it was always overblown hype and IMO they have already spent more money on it than will ever do anyone any good. All of that money could have been spent for something to do all of mankind some real good. What happened to the problems of pollution, crime, government corruption and waste, just to name a few.
I think that makes AI a colossal failure. They need to address real issues in a pragmatic way that solves problems, and not just have the folks at MIT try to break the bank at Vegas.
Maybe I'm a lone hermit at the top of a mountain waiting for everyone else to show up.

 

[NumberNine]

It was always marketing something vaguely useful... sometimes... with human supervision... and a decrease in service quality, as a global neccesity.

Except for the TTS on drive-thru speakers, which is an absolute marvel that I fully support.

 

[yannus]

As much as I don't like it, AI is going to conquer every market it can. Obviously, a computer will always be cheaper than a worker. The real threat is that the cost for the individual (as a user or indirectly) won't be cheaper, as when AI will have the monopoly in corporations, there will be nobody left to compete. As always the tool isn't the problem, the corporation is.

 

[Theinsanegamer]

I am sure that every time this was brought forward, we were talking about the technology behind the 4th industrial revolution... right?

Ah yeah, I thought so...

Matter of fact, the order books at TSMC, Nvidia and AMD have never been so filled...

The blacksmiths were booked out for months making shovels during the Gold Rush too.

 

[Mark Fuller]

People think AI was made to make more profits, when in reality, AI goal is to help alleviate administrative processes.

Ironically, AI is not really for the private sector, but for the public sector. Every government will develop their own custom AI cluster for national security reasons. If nobody can see the sign on the walls, then you are blind. When you have the world government having clear stance on the subject, which is in their priority list, then you understand what this is all about.

If you are mistaking AI and Generative AI as a simple chatbot, then of course you will call this a fad. However, AI will develop more than just a chat bot, it is already transcending into robotics which will shift the manufacturing business even more. It will also transcend government work to a level that no politician could do even with drastic policies.

Maybe you don't see the future because you are trapped into your safespace with your bias because you are mad that gaming hardware get more expensive as a collateral effect, but denying AI relevancy by just stating it is a bubble is ridiculous.

This bubble is on its 3rd year and it is driving the whole economy right now.

Matter of fact is Nvidia, AMD and TSMC will keep selling more and more chips year after year.

Lastly, the Nasdaq was down by 0,34% today, not 1.4%...

Not to mention what is really having the attention of the investors right now is Powell speech on inflation and the interest rate cuts.

I can see what you mean by your statement that AI is for the public sector. But I don't know what or anything it will do for national security reasons. The make thermoneuclear warheads in Russia, Communist China and North Korea. It would only take one of them to dissolve a government.

 

[AnilD]

It’s funny how everyone is chasing the flashy front end like chatbots and marketing tools, when the boring back end automation is what actually pays bills. History repeating itself, because the internet bubble burst the same way when everyone wanted to be the next pets dot com instead of the next boring logistics platform.

 

[airbornex]

If Nvidia GPUs are basically the scaffolding propping up the stock market, then we are all one supply chain hiccup away from watching Wall Street cosplay as Silicon Valley Bank.

 

[ferrellsl]

The famous last words of every bubble ever: "This time is different."

When you hear that, go ahead and order a tombstone with the words on it for whatever is being talked about.

Just like VR goggles! "They'll be successful this year!", said by every VR goggle manufacturer at every release.

 

[Revolution 11]

I am sure that every time this was brought forward, we were talking about the technology behind the 4th industrial revolution... right?

Ah yeah, I thought so...

Matter of fact, the order books at TSMC, Nvidia and AMD have never been so filled...

Talking about full order books at TSMC and company as a way to reject that this is a AI bubble is as stupid and blind as saying that the tulip bubble was not a tulip bubble because sellers of tulip bulbs made lots of money.

Since when does a bubble prevent some providers of equipment and goods from making money????? Show us your logic, redgarl.

 

[OortCloud]

Matter of fact is Nvidia, AMD and TSMC will keep selling more and more chips year after year.

That's what everybody said about Intel until about 4 years ago and IBM until 15 years ago etc etc.
The world changes, things move on. I expect 50 years from now at least 3 of NVidia, Apple, Meta, Amazon, AMD, TSMC and Google will be a historic footnote.

 

[LemmingO]

So. The foundation of any C-suite is to cover their asses. They outsource decision-making so they're not held liable to investors. If you're using an AI tool and the AI tool right now is explicitly saying "double-check my maths because sometimes 1+1 comes out 3", then they're liable for their bad decisions. This is something the C-suite doesn't like: accountability (they're fine with taking credit for success, just not for failure).

I don't mind the AI bubble bursting. I mind that we've saddled society with the "hope" of AI bringing prosperity for all, when it's just a massive shortcut to a dystopian society controlled by corporations.

 

[VariableSpike]

My only surprise is that this has happened even faster than I imagined, sounds like the rumblings are well on their way to VCs starting to pull plugs and companies writing down losses for failed AI investments, like the dot com boom, the more specialised successful stuff and AI in general will survive but the blind "ai good" investing and inflated value of stocks in Nvidia, Microsoft etc. will go bang as the whole market starts to sell off based on those effects and the supply chains dry up (not like you need to invest in tons of gpus or server space if people and especially businesses focus instead on cost management and refinement on whatever AI they do keep) as AI investment will probably drop a notable amount where you will have businesses more likely to go "this needs to make clear business sense" vs the current / past "AI is the new thing, we can't miss the wave" blind investing

 

[wiyosaya]

Matter of fact, the order books at TSMC, Nvidia and AMD have never been so filled...

Of course they are since everyone that is ordering has yet to figure out that its part of the bubble.

The management at those you mention is not immune to $$$ on a stick waving in front of their faces, IMO.

 

[wiyosaya]

<...>the more specialised successful stuff and AI in general will survive

I'm not so sure that survival of AI can be assumed especially in light of evidence of cognitive decline in the more specialized fields of use like cancer detection as detailed here https://www.techspot.com/community/topics/doctors-became-worse-at-spotting-cancer-after-relying-on-ai-for-help-study-finds Cognitive decline is also noted in more generalized use of AI, begging the question, IMO, is the over-all cost worth any benefit derived from AI's use?

but the blind "ai good" investing and inflated value of stocks in Nvidia, Microsoft etc. will go bang as the whole market starts to sell off based on those effects and the supply chains dry up (not like you need to invest in tons of gpus or server space if people and especially businesses focus instead on cost management and refinement on whatever AI they do keep) as AI investment will probably drop a notable amount where you will have businesses more likely to go "this needs to make clear business sense" vs the current / past "AI is the new thing, we can't miss the wave" blind investing

Like moths to a flame, people have always been drawn to get-rich-quick schemes and they likely always will until society finds a way to understand that, in the words of Rodgers and Hammerstein "Money Isn't Everything".

 

[techstrike]

Am surprised ... not.