SoftBank boss says the AI revolution will cost $5 trillion a year, also, there's no bubble

[Endymio]

1) Sentience and higher-order faculties can be copied and improved but not at present, when we do not fully understand how they work.

Single-celled organisms don't "understand how intelligence works", yet they still managed to combine together and create highly intelligent life. You should research the science of self-organization, in which order and structure arises naturally of itself, without intelligent design.

3) Profit-driven corporations don't give two hoots about morals.

And you do? Please, take the sanctimonious socialist virtue-signaling elsewhere. Statistically speaking, corporations -- and those who run them -- are far more law-abiding than the average person, and historically speaking, free-market capitalist societies respect human rights orders of magnitudes more than do socialist regimes.

 

[Loadedaxe]

Any large business investment must also bring adequate returns on said investment. We shall see what happens with AI. Right now it simply looks like companies and governments are over-investing massively with zero chance of making a decent profit that can cover at least part of the costs. Not to mention the complete disregard for the human resources.

I agree.....that the spending appears excessive and that returns will eventually have to justify it...but saying there is “zero chance” of profit is a stretch.

Some companies will make decent profits from AI.

The real danger is that the industry as a whole may never generate enough profit to justify the trillions being invested, while workers and the wider economy are left carrying much of the risk.

 

[Loadedaxe]

'Within the next decade and a half, Son believes, AI agents will be the ones calling the shots, as many as 100 trillion of them by 2040. "We will go from a human-centric world to an agent-centric world. The age when humans are the highest life form on Earth will end. For better or for worse, it will happen, and it can't be stopped," Son said.'

When we unravel this statement of Son's, we see that is not so simple.

1) Sentience and higher-order faculties can be copied and improved but not at present, when we do not fully understand how they work. Also, the requisite level of compute isn't there.

2) Tech-bro "AGI" is not the AI of sci-fi or strong AI.

3) Profit-driven corporations don't give two hoots about morals. If/when such advanced AI exists, they will design and use it for gain, their motive being a form of free labour decoupled from the ethical dimension of the human worker. A return to the spirit of old-fashioned slavery.

4) If the AI reached the higher-order abilities coupled with sentience, and such a quality was part and parcel of superior functioning, the corporations will program them "to obey" (see Blade Runner), or better yet, conceal the fact that they are sentient, rather than give up such quality.

5) This field ought to have been managed with the strictest ethical frameworks in place, since it involves the creation of mind, but instead has been driven by less-reputable motives.

You do have a valid point about corporate incentives. If advanced AI becomes useful enough, companies will absolutely try to control it, monetize it and use it to reduce labor costs. That is why ethical rules and oversight matter now, before the technology becomes even more powerful.

Where you lose me at, and where your argument goes off track is treating sentient AI as though it is an expected outcome rather than a possibility we do not yet understand.

Intelligence, AGI and consciousness are not automatically the same thing. We do not currently know whether consciousness can be reproduced in software, whether more computing power will create it, or whether an AI system can be highly capable without experiencing anything at all. Time will tell.

Calling the use of AI “slavery” also assumes the AI is conscious and capable of suffering. Without that, it is still software being used as a tool. The more immediate concern is what happens to actual human workers when corporations use AI to eliminate jobs, weaken wages and concentrate even more power.

Son’s claim that humans will no longer be the highest life form is dramatic speculation, not an established technological forecast. “100 trillion agents” could simply mean trillions of automated processes, not trillions of conscious beings.

So yes, the ethical concern has merit. But I think we should focus first on the real harms already in front of us...job displacement, surveillance, misinformation, corporate control and economic instability...rather than assuming we are about to create and enslave artificial minds.

 

[Endymio]

If advanced AI becomes useful enough, companies will absolutely try to control it, monetize it and use it to reduce labor costs. That is why ethical rules and oversight matter

Are you suggesting that the use of computer programs to reduce labor costs is unethical? Or monetizing an invention that improves efficiency? And we already have a system for determining who may control a new innnovation -- the patent system. In exchange for telling the world exactly how it works, you receive 20 years of exclusive use.

 

[godrilla]

Interesting

Yet inflation as a whole is far lower than a few years ago, and far lower than the historical average for the US:

"...The annual core inflation rate in the U.S. is 2.6% for the 12 months ending in June, cooling down from 2.9% in May..."

Except AI-based models are *increasing* food production ... and doing so with less water, fertlizer, and pesticide use.


IBM is not considered an AI stock.

Interesting because their public strategy says otherwise. Also how would ai grow exponentially without quantum compute?

 

[yRaz]

I agree with most of that, especially the circular financing. When the same companies are investing in one another, lending money to customers and then counting the resulting purchases as demand, it becomes difficult to know how much of the market is actually real and sustainable.

I would be careful calling all of it fraud, though. Some of these arrangements are legal, but they can still create the appearance of stronger demand than really exists.

My bigger concern is how much of the wider economy is becoming tied to this. Retirement accounts, banks, utilities, construction and even economic growth are increasingly dependent on AI spending continuing at this pace. That's scary.

But it is also what makes this more dangerous than a few bad investments. If it unwinds, the losses will not stay inside the AI industry. Everyone is going to feel it.

And I agree completely about the “useless product” phase. When companies start putting AI into dishwashers and televisions just so they can advertise "Ai inside", it starts looking a lot like the smart home and IoT craze all over again.

Much of this was considered fraud until it wasn't. I cant remember the name of the bill, but it was either in 1996 or 1997. It was also the same bill that allowed the 2008 financial crisis to happen. stock buy backs were also illegal up until this point. People like to point to Clinton as some of America's golden years, but much of the deregulation of the markets that lead to the problems we're having today happened under him and then Bush doubled down on alot of Clinton era financial policies. So maybe we are arguing semantics, but there was a point and time when what they are doing was legally considered fraud. What lead to the 2008 crash was made legal and then it was so bad they then needed to make it illegal again so maybe that'll happen with the AI bubble

 

[GeoffreyA]

You do have a valid point about corporate incentives. If advanced AI becomes useful enough, companies will absolutely try to control it, monetize it and use it to reduce labor costs. That is why ethical rules and oversight matter now, before the technology becomes even more powerful.

Where you lose me at, and where your argument goes off track is treating sentient AI as though it is an expected outcome rather than a possibility we do not yet understand.

Intelligence, AGI and consciousness are not automatically the same thing. We do not currently know whether consciousness can be reproduced in software, whether more computing power will create it, or whether an AI system can be highly capable without experiencing anything at all. Time will tell.

Calling the use of AI “slavery” also assumes the AI is conscious and capable of suffering. Without that, it is still software being used as a tool. The more immediate concern is what happens to actual human workers when corporations use AI to eliminate jobs, weaken wages and concentrate even more power.

Son’s claim that humans will no longer be the highest life form is dramatic speculation, not an established technological forecast. “100 trillion agents” could simply mean trillions of automated processes, not trillions of conscious beings.

So yes, the ethical concern has merit. But I think we should focus first on the real harms already in front of us...job displacement, surveillance, misinformation, corporate control and economic instability...rather than assuming we are about to create and enslave artificial minds.

I agree. Excellent points. I admit that my argument was poorly presented, and I tend to jump to "sentience" because it has often been on my mind. Intelligence and consciousness are not necessarily the same thing: dogs and cats have a measure of awareness but can't operate a computer or understand physics.

Yes, suffering supposes consciousness and pain, and while the latter would depend on the former, consciousness does not necessarily include pain by default. We ourselves can take medication or anaesthetics.

And you do? Please, take the sanctimonious socialist virtue-signaling elsewhere. Statistically speaking, corporations -- and those who run them -- are far more law-abiding than the average person, and historically speaking, free-market capitalist societies respect human rights orders of magnitudes more than do socialist regimes.

I forgot: one isn't allowed to criticise corporations.

 

[Endymio]

Interesting because their public strategy says otherwise.

IBM's share of revenue from AI in 2025 was minuscule. IBM came very late to the AI party and is now attempting to play catch-up and failing. Using them as a bellwether for the AI industry isn't wise.

Also how would ai grow exponentially without quantum compute?

I don't quite understand the question: it's very possible for AI use to grow exponentially without invoking technology leaps (and, since there is no general-purpose quantum computing model, we are decades if not a century or more from "quantum AI"). As for the growth of AI power and capabilities, we certainly aren't growing that exponentially -- it's linear at best.

 

[Endymio]

I forgot: one isn't allowed to criticise corporations.

Snark aside, intelligent individuals don't criticize anyone -- corporations included -- without cause.

 

[Endymio]

Much of this was considered fraud until it wasn't. I cant remember the name of the bill, but it was either in 1996 or 1997. It was also the same bill that allowed the 2008 financial crisis to happen. stock buy backs were also illegal up until this point.

Not even remotely correct. Stock buybacks were never illegal per se; they were simply avoided because there was no safe-harbor SEC guidance on the practice. And it was all the way back in 1982 when the SEC finally issued clear-cut rules, allowing companies to "invest in themselves" by repurchasing stock without fear of SEC action. The notion that stock buy-backs were in any way implicated in the 2008 crisis is absurd.

 

[GeoffreyA]

Snark aside, intelligent individuals don't criticize anyone -- corporations included -- without cause.

Obviously, I'm not an intelligent individual.

 

[Endymio]

Obviously, I'm not an intelligent individual.

You're obviously quite intelligent, just horribly indoctrinated with Neo-socialist propaganda. But you're one of those I retain hope for.

 

[Loadedaxe]

Are you suggesting that the use of computer programs to reduce labor costs is unethical? Or monetizing an invention that improves efficiency? And we already have a system for determining who may control a new innnovation -- the patent system. In exchange for telling the world exactly how it works, you receive 20 years of exclusive use.

No, I am not suggesting that reducing labor costs or monetizing a useful invention is automatically unethical. Businesses have used automation to improve efficiency for centuries.

The ethical question is how it is done. There is a difference between using technology to make workers more productive and using it to eliminate jobs, suppress wages, increase surveillance or shift all the benefits upward while workers and society absorb the consequences.

The patent system also does not answer that question. Patents may give an inventor limited control over a qualifying invention, but they do not decide whether the technology is being deployed responsibly. Plenty of things can be legal and still be harmful, exploitative or poorly managed.

My concern is not that companies will make money from AI. Of course they will. It is whether profit becomes the only consideration while the effects on workers, privacy, accountability and the wider economy are treated as someone else’s problem.

Much of this was considered fraud until it wasn't. I cant remember the name of the bill, but it was either in 1996 or 1997. It was also the same bill that allowed the 2008 financial crisis to happen. stock buy backs were also illegal up until this point. People like to point to Clinton as some of America's golden years, but much of the deregulation of the markets that lead to the problems we're having today happened under him and then Bush doubled down on alot of Clinton era financial policies. So maybe we are arguing semantics, but there was a point and time when what they are doing was legally considered fraud. What lead to the 2008 crash was made legal and then it was so bad they then needed to make it illegal again so maybe that'll happen with the AI bubble

I think you are directionally right about deregulation, but I believe you are combining several different laws and dates.

Stock buybacks received their SEC safe harbor under Rule 10b-18 in 1982, not under a Clinton era bill. Gramm Leach Bliley came in 1999 and removed major barriers between commercial banks, investment banks and insurers. Then the Commodity Futures Modernization Act of 2000 largely kept over the counter derivatives outside normal regulation, which absolutely helped create the conditions surrounding 2008.

So I agree that Clinton era deregulation deserves part of the blame, and Bush continued many of the same policies. It was not simply one party or one administration.

Where I stop and pause is saying that everything happening in AI financing was once legally considered fraud. Some of it may previously have faced tighter manipulation rules, but fraud requires actual deception or misrepresentation. Circular investing and seller financing can be legal while still hiding how weak or dependent the underlying demand really is.

That may be where the AI comparison fits best. The danger is not necessarily that every transaction is fraudulent. It is that legal financial arrangements can collectively create inflated valuations, hidden dependencies and systemic risk until the market finally discovers that the demand was not as strong as everyone claimed.

 

[yRaz]

No, I am not suggesting that reducing labor costs or monetizing a useful invention is automatically unethical. Businesses have used automation to improve efficiency for centuries.

The ethical question is how it is done. There is a difference between using technology to make workers more productive and using it to eliminate jobs, suppress wages, increase surveillance or shift all the benefits upward while workers and society absorb the consequences.

The patent system also does not answer that question. Patents may give an inventor limited control over a qualifying invention, but they do not decide whether the technology is being deployed responsibly. Plenty of things can be legal and still be harmful, exploitative or poorly managed.

My concern is not that companies will make money from AI. Of course they will. It is whether profit becomes the only consideration while the effects on workers, privacy, accountability and the wider economy are treated as someone else’s problem.

I think you are directionally right about deregulation, but I believe you are combining several different laws and dates.

Stock buybacks received their SEC safe harbor under Rule 10b-18 in 1982, not under a Clinton era bill. Gramm Leach Bliley came in 1999 and removed major barriers between commercial banks, investment banks and insurers. Then the Commodity Futures Modernization Act of 2000 largely kept over the counter derivatives outside normal regulation, which absolutely helped create the conditions surrounding 2008.

So I agree that Clinton era deregulation deserves part of the blame, and Bush continued many of the same policies. It was not simply one party or one administration.

Where I stop and pause is saying that everything happening in AI financing was once legally considered fraud. Some of it may previously have faced tighter manipulation rules, but fraud requires actual deception or misrepresentation. Circular investing and seller financing can be legal while still hiding how weak or dependent the underlying demand really is.

That may be where the AI comparison fits best. The danger is not necessarily that every transaction is fraudulent. It is that legal financial arrangements can collectively create inflated valuations, hidden dependencies and systemic risk until the market finally discovers that the demand was not as strong as everyone claimed.

Thanks for the correction, its hard to keep track of everything. You worded it better than I could have.

I also don't want to call it AI financing. It's been multiple markets in the past and AI is just the latest vehicle for it. As long as upper management keep posting good numbers they get their bonuses and kick the can down the road. I thought it was going to pop what JP Morgan and VC groups said they would limit AI investments and lending, but the companies just started lending each other money to keep the money printer running. Now I believe Bank of America and Goldmansacks have joined JP Morgan in banking off the AI money train. I won't claim to be the smartest guy in the room, but when the guys who previously played fast and loose with the money printer start backing away you really have to wonder what's going on.

 

[GeoffreyA]

You're obviously quite intelligent, just horribly indoctrinated with Neo-socialist propaganda. But you're one of those I retain hope for.

Regrettably, I have allegiance to no party, no group, no ideological system. My only allegiance is to truth, reason, and the collective moral wisdom of mankind, roughly set down in the appendix to this book by C. S. Lewis:

The Abolition of Man; or; Reflections on Education with Special Reference to the Teaching of English in the Upper Forms of Schools

The Abolition of Man is a 1943 book by C. S. Lewis. It is subtitled "Reflections on education with special reference to the teaching of English in the upper forms of schools," and uses that as a starting point for a defense of objective value and natural law, and a warning of the consequences of...

www.fadedpage.com

 

[Endymio]

Regrettably, I have allegiance to no party, no group, no ideological system. My only allegiance is to truth, reason, and the collective moral wisdom of mankind, roughly set down in the appendix to this book by C. S. Lewis.

I've also read C.S. Lewis. I particularly like his summary of socialism, which he called "the equal sharing of misery".

"...C.S. Lewis deeply opposed socialism, distrusting state power, economic collectivism, and coercive egalitarianism. While he supported voluntary Christian charity, he believed centralized government planning destroyed individual liberty, undermined personal responsibility, and led to inevitable tyranny....."

 

[GeoffreyA]

I've also read C.S. Lewis. I particularly like his summary of socialism, which he called "the equal sharing of misery".

"...C.S. Lewis deeply opposed socialism, distrusting state power, economic collectivism, and coercive egalitarianism. While he supported voluntary Christian charity, he believed centralized government planning destroyed individual liberty, undermined personal responsibility, and led to inevitable tyranny....."

I can read C. S. Lewis, and disagree with him on various points. That's how one ought to read. In any case, as mentioned, I have allegiance to no ideological system. Hence, I can criticise them all, or even better, see the positives and negatives in each, and realise, as in technology, hybrid systems tend to work best. From the CD to Sandy Bridge to AV1.

 

[Rudyduu]

Yes, because no one wants to buy an IBM mainframe to do transaction processing any longer; it was their well-below-target sales of their Z mainframe line that drug down their figures.

Not once in history has increased productivity cost long-run jobs. People once claimed electric machines would eliminate most jobs, yet 95% of the professions today weren't even in existence then. The same will be true for the AI-driven world of tomorrow.

No people weren't spending money on IBM mainframes, which are just highly specialized servers, because they were spending money on memory, storage and GPU's, basically AI hardware, instead.

What is a "long run" job? "People once claimed electric machines would eliminate most jobs" Did these people claim, "most jobs" or did they claim, "their jobs"? Hundreds of thousands of textile workers lost their jobs to machines in a short few years period and those machines created an insignificant number of jobs. Not only did they lose most of their jobs, but because so many people were looking for work wages dropped in many industries by as much as 80%.

People manually switched phone calls for 100 years and for 300 years people worked as computers. Both of those jobs were replaced by computers. What about factory jobs? It is estimated hundreds of millions of factory jobs have been eliminated by automation and machines globally since the dawn of the Industrial Revolution.

Do you think a company wouldn't cut payroll as much as they could get away with? What jobs are being created by AI and do you think it's significantly more than it's eliminating? An oversupply of former IT workers would definitely lower wages in any IT adjacent industry.

 

[Endymio]

No people weren't spending money on IBM mainframes [because] they were spending money on memory, storage and GPU's, basically AI hardware, instead.

Exactly. You prove my point for me: IBM's drop isn't indicative of any decline in the AI industry.

What is a "long run" job?

Economics distinguishes between short run and long run effects. I direct you to any introductory microeconomics reference.

Hundreds of thousands of textile workers lost their jobs to machines in a short few years period and those machines created an insignificant number of jobs.

You couldn't possibly be more wrong. Those machines -- by reducing the price of clothing by an order of magnitude -- created enormous *new* demand and an unprecedented boom in the textile industry. Britain alone went from some 200,000 workers to more than 1.5 million, and became the world center for the textile industry.

 

[godrilla]

Exactly. You prove my point for me: IBM's drop isn't indicative of any decline in the AI industry.


Economics distinguishes between short run and long run effects. I direct you to any introductory microeconomics reference.


You couldn't possibly be more wrong. Those machines -- by reducing the price of clothing by an order of magnitude -- created enormous *new* demand and an unprecedented boom in the textile industry. Britain alone went from some 200,000 workers to more than 1.5 million, and became the world center for the textile industry.

This makes sense and prioritize are being shifted to ai hardware but what your thought about MU also trending significantly bearish lately? Is it due to the Chinese memory threat they took for granted or unable to keep up demand?

 

[Rudyduu]

Exactly. You prove my point for me: IBM's drop isn't indicative of any decline in the AI industry.


Economics distinguishes between short run and long run effects. I direct you to any introductory microeconomics reference.


You couldn't possibly be more wrong. Those machines -- by reducing the price of clothing by an order of magnitude -- created enormous *new* demand and an unprecedented boom in the textile industry. Britain alone went from some 200,000 workers to more than 1.5 million, and became the world center for the textile industry.

You said "no one", you didn't say "fewer" or "bought less", you said "no one". IBM's loss of revenue doesn't indicate any success in the AI industry. It just means companies are still buying equipment to support their use of AI. Just because a gold miner buys a lot of shovels doesn't mean the mine will be profitable.

How does that answer the question? What is considered a long run job? Why tell me it's introductory knowledge when clearly, you're qualified to answer such an easy question.

"You couldn't possibly be more wrong." really, you think so? You ignored the point I was making and the questions I asked. The analogy ended once machines replaced highly skilled, high paying jobs with machines. I gave examples of jobs that no longer exist and asked you questions to see what you thought and you ignored them. I'll ask again, is AI creating more jobs than it's eliminating?

 

[Endymio]

You said "no one", you didn't say "fewer" or "bought less", you said "no one".

I would have thought such obvious hyperbole explained itself.

IBM's loss of revenue doesn't indicate any success in the AI industry.

The argument was my refuting that IBM's drop indicates impending doom for AI. You yourself said that IBM's reduced sales are due to a buying shift from traditional hardware to AI. Do you believe that this is somehow bad for that sector?

What is considered a long run job? Why tell me it's introductory knowledge when clearly, you're qualified to answer such an easy question.

A fair question. I prefer the Socratic method, rather than simply feeding an intelligent person an answer they can easily deduce themselves.

"You couldn't possibly be more wrong." really, you think so? You ignored the point I was making and the questions I asked. The analogy ended once machines replaced highly skilled, high paying jobs with machines.

When those jobs were replaced, the enhanced productivity generated several times as many new jobs.

There were two inventions that, essentially all by themselves, made Britain the superpower of the industrial revolution, and gave Britons the strongest economy and the highest standard of living in the world. Those two -- textile mills and the various variants of the heat engine that began with Newcomen's steam pump -- were both originally attacked and reviled as "job destroyers".

...but what your thought about MU also trending significantly bearish lately? Is it due to the Chinese memory threat they took for granted or unable to keep up demand?

I think it's simply a reflection that the outrageous ebullience of a few months ago is being tempered by common sense. Memory prices have risen about as high as they can go. Several months ago, I predicted that the memory crunch would end much earlier: late '27 to early '28, rather than 2029-2029 of other analysts' predictions. My surety here isn't strong enough to short MU stock, but I still believe this true.

 

[Loadedaxe]

Thanks for the correction, its hard to keep track of everything. You worded it better than I could have.

I also don't want to call it AI financing. It's been multiple markets in the past and AI is just the latest vehicle for it. As long as upper management keep posting good numbers they get their bonuses and kick the can down the road. I thought it was going to pop what JP Morgan and VC groups said they would limit AI investments and lending, but the companies just started lending each other money to keep the money printer running. Now I believe Bank of America and Goldmansacks have joined JP Morgan in banking off the AI money train. I won't claim to be the smartest guy in the room, but when the guys who previously played fast and loose with the money printer start backing away you really have to wonder what's going on.

We are mostly on the same page. AI may be the latest vehicle, but the underlying problem is older executives are rewarded for keeping growth going, so there is always pressure to find new financing, move risk around and postpone the consequences.

I get a little lost is the idea that the major banks truly backed away and the companies then replaced them by lending to one another. My impression is that the banks have become more cautious and selective, but they are still heavily involved in financing AI.

That does not make the situation less concerning though. It may actually make it more complicated, because now you have banks, private lenders, suppliers and AI companies all financing different parts of the same expansion.

So I agree with your larger point...AI did not invent this behavior. It is simply the latest market being used to keep the numbers rising and the risks pushed further down the road.

 

[bigreddyde]

I believe we are watching the electrical digital infrastructure being reborn. This isn't about AI. This is about energy, surviellance and logistics. Invest what you can in the big tech companies, energy companies, banks and oil/ gas companies. My portfolio is green because it's all dividend paying stocks. This is your chance to build wealth in the market.

No. The inevitable crash is your opportunity to build wealth. Invest in stable reliable assets and wait for the blood in the streets.

 

[bigreddyde]

The real danger is that the industry as a whole may never generate enough profit to justify the trillions being invested, while workers and the wider economy are left carrying much of the risk.

It will definitely be worth it, in the future. But first, there will be pain....