$650 billion in annual revenue required to deliver 10% return on AI buildout investment, J.P. Morgan claims

[Anonymous]

$650 billion in annual revenue required to deliver 10% return on AI buildout investment, J.P. Morgan claims


https://www.tomshardware.com/tech-industry/artificial-intelligence/usd650-billion-in-annual-revenue-required-to-deliver-10-percent-return-on-ai-buildout-investment-j-p-morgan-claims-equivalent-to-usd35-payment-from-every-iphone-user-or-usd180-from-every-netflix-subscriber-in-perpetuity

Why are we so greedy in this country????

And if you add crypto this nonsense, we are seriously out of control....

[Anonymous]

Here is what is happening in my opinion...

We are seeing more and more of these kind of articles....

Do you guys know why?

Because the same usual suspects have secretly taken short positions....

Does it sound familiar? It should

[Anonymous]

Anonymous wrote:Here is what is happening in my opinion...

We are seeing more and more of these kind of articles....

Do you guys know why?

Because the same usual suspects have secretly taken short positions....

Does it sound familiar? It should

Maybe -- but they're not wrong. And yes, Michael Burry and other prominent investors have been open about shorting AI stocks, but that's an easy bet to make given their extreme valuations given that revenues/profits haven't yet caught up. What worries me is the composition of the S&P 500 -- over 1/3 is significantly tied to AI, yet people naively think it's a diversified portfolio and auto-invest their 401(k)s in ETFs tied to the S&P 500.

I think that valuations are precarious, and it will take very little to cause a broad-based sell-off in the markets. And since we've been in this K-shaped economy with the "haves" propping up consumer spending, my concern is that once the focus moves from the government spending shutdown and Fed rate decisions to valuations, the bottom will fall out as people head toward the exits.

Just call me Chicken Little!

[Anonymous]

Anonymous wrote:Here is what is happening in my opinion...

We are seeing more and more of these kind of articles....

Do you guys know why?

Because the same usual suspects have secretly taken short positions....

Does it sound familiar? It should

Yeah, they know AI isn’t going to deliver. Didn’t Meta layoff people in its AI division after that botched presentation?

[Anonymous]

Anonymous wrote:

Anonymous wrote:Here is what is happening in my opinion...

We are seeing more and more of these kind of articles....

Do you guys know why?

Because the same usual suspects have secretly taken short positions....

Does it sound familiar? It should

Maybe -- but they're not wrong. And yes, Michael Burry and other prominent investors have been open about shorting AI stocks, but that's an easy bet to make given their extreme valuations given that revenues/profits haven't yet caught up. What worries me is the composition of the S&P 500 -- over 1/3 is significantly tied to AI, yet people naively think it's a diversified portfolio and auto-invest their 401(k)s in ETFs tied to the S&P 500.

I think that valuations are precarious, and it will take very little to cause a broad-based sell-off in the markets. And since we've been in this K-shaped economy with the "haves" propping up consumer spending, my concern is that once the focus moves from the government spending shutdown and Fed rate decisions to valuations, the bottom will fall out as people head toward the exits.

Just call me Chicken Little!

The number is a little silly. More than half of the AI investments will not pan out. They will die off not at the same time. Some quickly, some medium term, and some long term. Look at the early days of the US car industry. Some remain (really just Ford) but many many others died. Some quickly -- some bought over the medium term by others and some as recently as the last 20 years. Same with AI. 10% return will need a lot less revenue when all is said and done.

Pretty comfortable with the S&P 500 index at this point.

[Anonymous]

We've been here before, remember the dot-com bubble? Lots of money going to try to find the winners. There will be winners in the end, but also losers. AI is here to stay, but not at the revenue/profit levels commensurate with the investment we're seeing. Just like in dot-com when we ended up with lots of dark fiber from all the efforts to wire up everywhere. It took years for that to all come to life, but it did.

The difference between now and dot-com days is the dollar amounts are much bigger and there's more urgency to find the winner, so that leads to more bubble than expected.

[Anonymous]

Anonymous wrote:We've been here before, remember the dot-com bubble? Lots of money going to try to find the winners. There will be winners in the end, but also losers. AI is here to stay, but not at the revenue/profit levels commensurate with the investment we're seeing. Just like in dot-com when we ended up with lots of dark fiber from all the efforts to wire up everywhere. It took years for that to all come to life, but it did.

The difference between now and dot-com days is the dollar amounts are much bigger and there's more urgency to find the winner, so that leads to more bubble than expected.

Yeah, except that fiber was actual infrastructure. All that's left after this will be a bunch of bullshit LLM models and millions of chips that in 5 years will be worthless. Meanwhile the planet burns up that much faster and people pay through the nose for electricity because the data center 2 miles away is running 24/7.

[Anonymous]

Anonymous wrote:

Anonymous wrote:We've been here before, remember the dot-com bubble? Lots of money going to try to find the winners. There will be winners in the end, but also losers. AI is here to stay, but not at the revenue/profit levels commensurate with the investment we're seeing. Just like in dot-com when we ended up with lots of dark fiber from all the efforts to wire up everywhere. It took years for that to all come to life, but it did.

The difference between now and dot-com days is the dollar amounts are much bigger and there's more urgency to find the winner, so that leads to more bubble than expected.

Yeah, except that fiber was actual infrastructure. All that's left after this will be a bunch of bullshit LLM models and millions of chips that in 5 years will be worthless. Meanwhile the planet burns up that much faster and people pay through the nose for electricity because the data center 2 miles away is running 24/7.

The other key difference is that back then the products had huge network effect benefits and economies of scale. When you do more LLM modeling, it just gets more expensive because you just need more chips and centers and electricity

The revenue like never crosses over the cost line because they have the same slope.

https://www.ft.com/content/fce77ba4-6231-4920-9e99-693a6c38e7d5

[Anonymous]

As the user base shifts toward mass adoption, the products will be enshittified and the cost per inference will fall.

Inference is too expensive right now but it is only 2x—3x too expensive.

[Anonymous]

Anonymous wrote:As the user base shifts toward mass adoption, the products will be enshittified and the cost per inference will fall.

Inference is too expensive right now but it is only 2x—3x too expensive.

Yeah people will.be really excited to pay $50/month for a product riddled with ads. Good luck with that

[Anonymous]

Anonymous wrote:

Anonymous wrote:As the user base shifts toward mass adoption, the products will be enshittified and the cost per inference will fall.

Inference is too expensive right now but it is only 2x—3x too expensive.

Yeah people will.be really excited to pay $50/month for a product riddled with ads. Good luck with that

I see ChatGpt or similar replacing Google in search. It's so much better, and they ahve 10 million users already paying $30/month or more.

But.. the jury's not out on this one. Google has been adding AI results as they are worried, and what we may see is an AI-search engine is bundled into something else people pay for like the Microsoft Office bundle.

[Anonymous]

Anonymous wrote:

Anonymous wrote:We've been here before, remember the dot-com bubble? Lots of money going to try to find the winners. There will be winners in the end, but also losers. AI is here to stay, but not at the revenue/profit levels commensurate with the investment we're seeing. Just like in dot-com when we ended up with lots of dark fiber from all the efforts to wire up everywhere. It took years for that to all come to life, but it did.

The difference between now and dot-com days is the dollar amounts are much bigger and there's more urgency to find the winner, so that leads to more bubble than expected.

Yeah, except that fiber was actual infrastructure. All that's left after this will be a bunch of bullshit LLM models and millions of chips that in 5 years will be worthless. Meanwhile the planet burns up that much faster and people pay through the nose for electricity because the data center 2 miles away is running 24/7.

Not as good for sure, but it means we'll be left with super-cheap AI compute power and as the price goes down, it opens up more applications for the use of AI.

As it is, AI pricing has dropped about 10x in the last 2 years, which is typical with new technologies, and that's mostly due to beter algorithms.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:As the user base shifts toward mass adoption, the products will be enshittified and the cost per inference will fall.

Inference is too expensive right now but it is only 2x—3x too expensive.

Yeah people will.be really excited to pay $50/month for a product riddled with ads. Good luck with that

I see ChatGpt or similar replacing Google in search. It's so much better, and they ahve 10 million users already paying $30/month or more.

But.. the jury's not out on this one. Google has been adding AI results as they are worried, and what we may see is an AI-search engine is bundled into something else people pay for like the Microsoft Office bundle.

Inference costs too much for the provider. If it cost one third as much then OpenAI would have positive unit economics with the paying customers they already have.

Reply above is spot on. I expect Copilot is going to be pushed into the E365 bundle. With confusing pricing to hide the way Microsoft will leverage their exiting distribution advantage to win in agents. Just like Teams and Slack.

[Anonymous]

Anonymous wrote:

Anonymous wrote:We've been here before, remember the dot-com bubble? Lots of money going to try to find the winners. There will be winners in the end, but also losers. AI is here to stay, but not at the revenue/profit levels commensurate with the investment we're seeing. Just like in dot-com when we ended up with lots of dark fiber from all the efforts to wire up everywhere. It took years for that to all come to life, but it did.

The difference between now and dot-com days is the dollar amounts are much bigger and there's more urgency to find the winner, so that leads to more bubble than expected.

Yeah, except that fiber was actual infrastructure. All that's left after this will be a bunch of bullshit LLM models and millions of chips that in 5 years will be worthless. Meanwhile the planet burns up that much faster and people pay through the nose for electricity because the data center 2 miles away is running 24/7.

Dude— the fiber turned out to be near worthless.