Artificial Intelligence

[winjitsu]

https://openai.com/sam-and-jony/

https://www.cnbc.com/2025/05/21/openai-buys-iphone-designer-jony-ive-device-startup-for-6point4-billion.html

 

Buying a 55 person, pre-product start-up for $6.4bn?

Top.

[lnofeisone]

3 hours ago, winjitsu said:

https://openai.com/sam-and-jony/

https://www.cnbc.com/2025/05/21/openai-buys-iphone-designer-jony-ive-device-startup-for-6point4-billion.html

 

Buying a 55 person, pre-product start-up for $6.4bn?

Top.

How do yo figure it's the top?

 

OpenAI spends 13B in cloud compute for inferencing (not training and R&D) because everything gets routed back to its cloud. Shifting AI to local devices is a logical next step and if they are successful will reduce their 13B inference cost. 

[winjitsu]

12 minutes ago, lnofeisone said:

How do yo figure it's the top?

 

OpenAI spends 13B in cloud compute for inferencing (not training and R&D) because everything gets routed back to its cloud. Shifting AI to local devices is a logical next step and if they are successful will reduce their 13B inference cost. 

 

Just general observations for pre-revenue acquisitions and fundraisings. Last one that jumped out to me like this was Magic Leap, and before that the Katzenberg start-up Quibi. Windsurf I could at least understand based on user adoption. 

 

The more I think about it, the more I think of Yahoo in the late 90s -- first to market in many ways, very acquisitive, but a long term loser. Maybe there's a pump or two left on the way to IPO.

 

Wanting to bypass Apple / Samsung / Phones into their own user devices IMO is a much different game, and I think they will burn billions on the way to bankruptcy. Closed $40bn 2 months ago and already spent $10bn on these two acquisitions.

[lnofeisone]

7 hours ago, winjitsu said:

 

Just general observations for pre-revenue acquisitions and fundraisings. Last one that jumped out to me like this was Magic Leap, and before that the Katzenberg start-up Quibi. Windsurf I could at least understand based on user adoption. 

 

The more I think about it, the more I think of Yahoo in the late 90s -- first to market in many ways, very acquisitive, but a long term loser. Maybe there's a pump or two left on the way to IPO.

 

Wanting to bypass Apple / Samsung / Phones into their own user devices IMO is a much different game, and I think they will burn billions on the way to bankruptcy. Closed $40bn 2 months ago and already spent $10bn on these two acquisitions.

I like your perspective. The idea of Yahoo-like acquisitions crossed my mind. This is one of the main reasons why I have been slowly and steadily building up my Google position. For Google to gain the advantage, here is a simple software push. 

[mattee2264]

https://foundationcapital.com/the-ai-hype-600b-question-or-4-6t-opportunity/

 

This is a pretty good article. 

 

A lot of bears are citing record valuations as a reason to be bearish. But if a few years down the lines companies will be able to reduce their headcounts by 10-20% and a chunk of the money saved goes to Big Tech with the rest increasing profits then both S&P 500 earnings are going to be an awful lot higher.

 

There is a little uncertainty as to whether Big Tech will retain their market leadership or will be eclipsed by nimbler later entrants who aren't burdened by the huge initial capex investments and benefit from lower future chip costs and cost of computes etc. But given Big Tech are scooping up all the promising start ups and have a huge user base they can test their AI offerings on as well as synergies with their existing core products (e.g. Google/Meta can use AI to improve the return on their clients advertising investments, Microsoft can offer an AI office assistant and other productivity enhancing features, Nvidia may be the next Intel able to retain their technological leadership and capture most of the market share for each new generation of AI chips etc)  then I think they will get more than their fair share of the AI revenue opportunity and if it is big enough then there is room for new entrants to come in and quickly reach large valuations that will be additive to S&P 500 market cap and earnings.

 

But at 30x forward earnings Mag7 aren't exactly priced for perfection and their core businesses alone are still growing and you could almost argue that you are getting any AI upside for free and the upside could be considerable. 

 

 

 

[Saluki]

11 hours ago, mattee2264 said:

https://foundationcapital.com/the-ai-hype-600b-question-or-4-6t-opportunity/

 

This is a pretty good article. 

 

A lot of bears are citing record valuations as a reason to be bearish. But if a few years down the lines companies will be able to reduce their headcounts by 10-20% and a chunk of the money saved goes to Big Tech with the rest increasing profits then both S&P 500 earnings are going to be an awful lot higher.

 

There is a little uncertainty as to whether Big Tech will retain their market leadership or will be eclipsed by nimbler later entrants who aren't burdened by the huge initial capex investments and benefit from lower future chip costs and cost of computes etc. But given Big Tech are scooping up all the promising start ups and have a huge user base they can test their AI offerings on as well as synergies with their existing core products (e.g. Google/Meta can use AI to improve the return on their clients advertising investments, Microsoft can offer an AI office assistant and other productivity enhancing features, Nvidia may be the next Intel able to retain their technological leadership and capture most of the market share for each new generation of AI chips etc)  then I think they will get more than their fair share of the AI revenue opportunity and if it is big enough then there is room for new entrants to come in and quickly reach large valuations that will be additive to S&P 500 market cap and earnings.

 

But at 30x forward earnings Mag7 aren't exactly priced for perfection and their core businesses alone are still growing and you could almost argue that you are getting any AI upside for free and the upside could be considerable. 

 

 

 

I think the bump will be temporary. There is anecdotal info out there that some insurance company is using it for claims processing and that the AI can look at the photos of car damage and come up with a price and have a letter ready in minutes, with no people involved. 

 

Okay...then what? Let's say Progressive used it and has a huge cost advantage vs it's competition. Then GEICO and State Farm get it. And cut costs too. The margin vs competitors stays the same but consumers win by getting lower prices. But insurance companies might lose because less money from premiums means less float for investment.

 

Margins could get even lower too. What if instead of calling around you can have AI analyze your driving habits and speed, quick stops from your cell phone motion data and find you the best price?

 

I think AI plays like Google can be huge if they move from SaaS pricing to some kind of auction driven or other model to capture the gains. When mining in the UK started using coal powered engines, they were very costly and the machine owners didn't sell them or rent them, but charged a price that lower the labor it replaced. It was most popular in the places where labor costs were highest, so I see it (for now) taking hold in places like the US white colar industry rather than places like India where we outsource our labor to, because it's cheaper there so the savings will be less. In China, which makes farm machinery, they still have people farming by hand in rural areas because labor is so cheap that a machine is not worth investing in.

[mattee2264]

Most industries aren't that competitive in this day and age. If they were then margins wouldn't have expanded so much as companies pocketed tax cuts, interest cost savings from cheap debt, and retained pandemic price hikes and other things you'd expect would be competed away. 

 

Also AI isn't free. So to some extent it will represent IT budget replacing staff budget with perhaps limited benefit to profits in the short term and adoption driven more by the worry that if you wait for AI costs to come down you will be way behind the learning curve that early adopters benefited from. 

 

But if AI does become cheap enough fast enough for near term mass adoption (and mass layoffs) then at that point Big Tech margins will probably be a lot lower and they won't be recouping a large chunk of their initial investments made when NVIDIA chips were super expensive and data centre rents similarly so. 

 

 

 

 

 

 

[DooDiligence]

Does this mean we'll be able get rid of politicians? Let's go boys!!!

 

https://www.404media.co/github-is-leaking-trumps-plans-to-accelerate-ai-across-government/

[Dalal.Holdings]

 

The race to Spend Spend Spend on AI is as hot as ever. But how will AI ever generate the returns to compensate for all the spending? Meh, let's not think too much about that...

 

[Spekulatius]

I have some shares in ACVA which uses AI for their wholesale used car market place business. It’s used for pricing, but also to estimate the condition of the cos, as well as damage estimates based on fotos. It’s obvious to use AI if you have a lot of data and are a digital disrupter anyways.

 

For legacy business, the challenge is to get clean data in a way that’s suitable for LLM ingestion, the skill of the employee base, including the IT personal, integrating the AI in existing systems as well as training the users.

https://d1io3yog0oux5.cloudfront.net/_da26f5697c332c8c5b5e6db6df5ebb08/acvauto/db/937/8424/presentation/ACVA_Q1-25_Earnings+Presentation_FINAL.pdf

 

I think overall, AI makes it easier for a digital challenger to win against an incumbent who may not be able to move that fast on AI adoption.

 

[Dave86ch]

Actuators are usually very expensive, and Unitree (backed by Alibaba) built a robot with 26 joints for just $5,900.

 

 

 

[lnofeisone]

I'm going to leave this right here - https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

 

A lot of buzz in at least 2 publicly traded consulting firms about this one, and I got a few calls from clients wanting to discuss this.

[nsx5200]

So @theperksofbeinganINTJ from Askeladden Capital's been integrating AI into his investment and non-investment productivity workflow, and has shared his experience in his newsletters, which I find interesting.

 

Based on his experience, it seems like the task of digesting information on smaller companies are becoming easier.  I suspect this will start to eat away at the mismatch between the fundamental and the price for these small/microcaps.

 

AI is changing the way we communicate and think, based on articles like ChatGPT Is Changing the Words We Use in Conversation and How algorithms are changing the way we speak.  The implication of this, coupled with increasing adoption of people digesting information using AI-first is: this can lead to corporate(or worse, state agents) to 'hack'/influence AI systems by using certain languages to get AI to spit out more favorable or unfavorable summaries.

 

Super interesting times as AI is disrupting 'investing' in more meaningful ways.

[tnathan]

On 8/27/2025 at 3:17 PM, lnofeisone said:

I'm going to leave this right here - https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

 

A lot of buzz in at least 2 publicly traded consulting firms about this one, and I got a few calls from clients wanting to discuss this.

Yeah I think the article is spot on. It's the ease of use / implementation for specific use cases or workflows that really matters. Think of your average person at your average slow moving company -- there is no way these companies create their own agents effectively from both a functionality + implementation persepctive. 

 

A great example of things that are currently being done successfully in the finance world are companies that have sprung up to assist private equity firms in their M&A / due diligence process. Think: (1) upload VDR -> (2) query all of this data in a useful and streamlined way (3) do other cool stuff that used to take hours (e.g., pulling all the specific clauses from the 100 loan agreements you might have and put into a spreadsheet), (4) quickly output 80% finished reports of various types that used to take hours / days. Essentially now you can either go through more deal flow or reduce your headcount and get through the same amount of deal flow. Or just spend more time on the important stuff, whatever your goal.

 

AI will be huge, I think businesses are just waiting for AI-savvy young people to create the tools (e.g., AI software / agents) that they can easily plug and play into their current workstreams without having to actually know anything about AI. That is coming, first to the highest value / easiest use cases, then to everything else. Just going to take some time

[Eng12345]

The thing with all of this that I can't stop thinking about and don't see anyone else talking about is that this AI boom seems to be taking these asset light, FCF generative businesses and turning them into very capex heavy businesses that will have high ongoing maintenance costs, whether that be for the power generation, the data centers, or the cost of upgrading the chips/servers regularly. This is the main reason I have held out on alphabet recently. 

 

Honestly seems like long term there should be a multiple reduction for some of these companies like alphabet etc

Edited August 29, 2025 by Eng12345

[tnathan]

40 minutes ago, Eng12345 said:

The thing with all of this that I can't stop thinking about and don't see anyone else talking about is that this AI boom seems to be taking these asset light, FCF generative businesses and turning them into very capex heavy businesses that will have high ongoing maintenance costs, whether that be for the power generation, the data centers, or the cost of upgrading the chips/servers regularly. This is the main reason I have held out on alphabet recently. 

 

Honestly seems like long term there should be a multiple reduction for some of these companies like alphabet etc

But the TAMs will explode. No reason why the cloud providers can't charge substantially more when they are the backbone of all AI applications. In the example I gave above, the LLMs are currently being paid probably a few million to provide the backbone of these Finance AI companies. That is the current starting point, but the ability for them to take a much greater slice of the pie once AI applications start to make real money is clearly there. Now TAMs should include a bunch of service and labor costs (e.g., I'm an AI application company that can replace 20% of your labor in a specific area, so I will charge 50% of that replacement cost and a company keeps 50% of the savings -> I am the hyperscaler powering that application, I will price my services to the AI application company so that I take 50% of that -> Now all of a sudden hyperscaler TAM is based off of labor savings (e.g., 25% of total labor saved or whatever) which was not really in the realm of possibility before)

[Dalal.Holdings]

11 minutes ago, Eng12345 said:

The thing with all of this that I can't stop thinking about and don't see anyone else talking about is that this AI boom seems to be taking these asset light, FCF generative businesses and turning them into very capex heavy businesses that will have high ongoing maintenance costs, whether that be for the power generation, the data centers, or the cost of upgrading the chips/servers regularly. This is the main reason I have held out on alphabet recently. 

 

Honestly seems like long term there should be a multiple reduction for some of these companies like alphabet etc

 

Yep. Most of the Mag 7 are becoming worse businesses (more capital intensive) with likely lower future returns on capital. Then you also have the insane compensation (many are in “non-cash” equity so cash flows mask it) packages for people skilled in AI…

 

Mom & Pop who are shoving everything into an index fund (over 1/3 of investment in S&P now goes to Mag 7) might not be aware of that fact…

[Spekulatius]

On 8/29/2025 at 10:43 AM, tnathan said:

But the TAMs will explode. No reason why the cloud providers can't charge substantially more when they are the backbone of all AI applications. In the example I gave above, the LLMs are currently being paid probably a few million to provide the backbone of these Finance AI companies. That is the current starting point, but the ability for them to take a much greater slice of the pie once AI applications start to make real money is clearly there. Now TAMs should include a bunch of service and labor costs (e.g., I'm an AI application company that can replace 20% of your labor in a specific area, so I will charge 50% of that replacement cost and a company keeps 50% of the savings -> I am the hyperscaler powering that application, I will price my services to the AI application company so that I take 50% of that -> Now all of a sudden hyperscaler TAM is based off of labor savings (e.g., 25% of total labor saved or whatever) which was not really in the realm of possibility before)

I think in general, more than 80% of the value accrues to the customer. Also cloud is really a commodity st the core, so it seems that competition will prevent this from happening.

I think the real money will be made by a company that may not exist yet, just how Google came out of nowhere after the internet infrastructure was in place and took the search market. It’s probably some super adapted SI agent That can crawl around your own property data and come up with no AI build solution that can make the average employee that works with ERP systems way productive or does away with some specialists that just move around things in those systems (logistics, purchasing, finance, HR functions etc).

[Dalal.Holdings]

https://www.bloomberg.com/news/articles/2025-09-02/apple-s-lead-ai-researcher-for-robotics-heads-to-meta-as-part-of-latest-exits?srnd=homepage-americas

 

The tech companies are poaching each other's employees for tens or even hundreds of millions of dollars, spending 3 orders of magnitude more than that on datacenters/chips/etc (hundreds of billions of dollars+ collectively)

 

...and they're all producing near identical chatbots that are difficult to distinguish from each other...

 

Sure hope the margins and returns justify all the expenses & expenditures !

[tnathan]

On 8/30/2025 at 3:13 PM, Spekulatius said:

I think in general, more than 80% of the value accrues to the customer. Also cloud is really a commodity st the core, so it seems that competition will prevent this from happening.

I think the real money will be made by a company that may not exist yet, just how Google came out of nowhere after the internet infrastructure was in place and took the search market. It’s probably some super adapted SI agent That can crawl around your own property data and come up with no AI build solution that can make the average employee that works with ERP systems way productive or does away with some specialists that just move around things in those systems (logistics, purchasing, finance, HR functions etc).

Can you explain this? It's not really a commodity if the cloud companies have >30% margins ... 

[LC]

7 hours ago, Dalal.Holdings said:

...and they're all producing near identical chatbots that are difficult to distinguish from each other...

 

They're probably betting it will play out like search. In the early days of search, results were not really different between the various search engines. Eventually, one search engine took all the economics. 

[Dalal.Holdings]

6 hours ago, LC said:

 

They're probably betting it will play out like search. In the early days of search, results were not really different between the various search engines. Eventually, one search engine took all the economics. 

Search did not require all that capex though. 
 

The tech companies are betting AI will be winner-take-all as everything else in tech. There is no indication that will actually happen though…

[Dalal.Holdings]

Saved this gem of Seeking Alpha comment from a few weeks back:

 

 

A lot of the AI revenue...is coming from other AI/tech firms.

 

Silicon Valley is increasingly incestuous. I sure hope they can one day convince the average person to pay for all this investment they are making. For now, it looks like the money is coming from the tech companies (which are getting some of the money from issuing their overvalued shares as comp/etc).

 

Meanwhile, mom & pop holding S&P index fund are 1/3 concentrated in a handful of these stocks

 

Edited September 3, 2025 by Dalal.Holdings

[tnathan]

50 minutes ago, Dalal.Holdings said:

Saved this gem of Seeking Alpha comment from a few weeks back:

 

 

A lot of the AI revenue...is coming from other AI/tech firms.

 

Silicon Valley is increasingly incestuous. I sure hope they can one day convince the average person to pay for all this investment they are making. For now, it looks like the money is coming from the tech companies (which are getting some of the money from issuing their overvalued shares as comp/etc).

 

Meanwhile, mom & pop holding S&P index fund are 1/3 concentrated in a handful of these stocks

 

Yeah we all know they're investing ahead of the actual application layer. It will be totally fine if you think a substantial amount of value accrues to the cloud companies when the application layer does come(which I don't see how it won't)

[Dalal.Holdings]

5 minutes ago, tnathan said:

Yeah we all know they're investing ahead of the actual application layer. It will be totally fine if you think a substantial amount of value accrues to the cloud companies when the application layer does come(which I don't see how it won't)

 

This is like thinking AOL would accrue the value of folks wanting access to the internet back in 2000. AOL (and the telco companies that brought internet to your house) didn't become the big winners of the internet. AOL-Time Warner was a spectacular blunder by Time Warner. Companies like Google and Facebook/Meta barely existed/didn't exist back then. 

 

Deepseek showed a lot of AI might be heavily commoditized and easy to duplicate. At current levels of spending, the risks of this are not being adequately accounted for...