저자: ShaiBalure · 외국인
요약 보기
AI 투자 테마가 GPU(엔비디아)에서 반도체 제조·네트워킹 인프라(TSMC, ASML, 브로드컴)로 확장 중이며, 전력·냉각 등 AI 유틸리티 병목이 새 테마라고 주장. 기존 SaaS 소프트웨어는 AI 네이티브 기업의 자체 구현으로 위협받으니 고객 수보다 NRR·사용량 지표를 봐야 한다고 강조. 다만 일부 AI 인프라 종목이 6개월간 600% 급등해 과열 우려가 있으나, 이번 붐은 닷컴과 달리 하이퍼스케일러의 영업현금흐름으로 자금 조달된다고 설명.
다듬은 글
It's time to spotlight the future of AI and discuss the AI trade. Joining us now is Shai Balure, the chief market strategist of Futurum Equities. Shai, thanks for being with us today. You know, first we're going to talk about Futurum's AI5. You know, can you tell us first what the AI-15 is and what the process is for picking out who makes the list?
Yeah, I mean first I want to say thank you for having me on and I think that the future of AI5 began three months ago where we wanted to create a list that was going to be rebalanced every 3 months on where new money in the AI economy is going right now at this moment. So the first AI wave we clearly saw it was all about hardware, servers, compute that's why Nvidia just continues up to the right because they are truly the GPU engine of this new digital economy.
Then after the Nvidias of the world, we start seeing the Mag 7 names kind of bolt uh have the software sprint where they're bolting on Gen AI into their whole product suite and the their ability to do it at the rapid pace that they can was why they benefited right away. So we're talking Google putting Gemini in search, we're talking Microsoft co-pilots and office.
We're talking Meta with smart smarter ad targeting. However, we at future believe that this is this chapter is largely priced in. So, we wanted to create a list that was outside the max 7 which 15 names we believe the new money is currently going to in this AI economy that's rapidly changing and the pace is way greater than we initially anticipated.
And so you mentioned that first wave of Gen AI, sort of what we all thought of when all the chat bots came out, but we've seen the AI trade change so much and and pivot really into the the AI infrastructure of all of it. So how many of the plays are in the the data components and the the physical component of AI?
Well, I mean, there's two monopolies outside of Nvidia that a lot of people don't really discuss as much as they should. We're talking TSMC. They are truly the AI factory of this new digital economy. Their wafers are the only reason we Nvidia is able to have these advanced GPUs. Then we also have ASML who essentially creates a lithography tools in order for these advanced GPUs to also be available in the market.
So we have the king AI keymaker and ASML and the AI factory and TSMC monopolies that nobody's touching. Then we also have Broadcom which I think is the most hidden $2 trillion company in the world where they control 70% of the custom AI chips for Google and the high-speed like gear moving the data between all the hyperscaler centers.
So when you think about all the data AI data centers that all these hyperscalers are creating we get the news story every week that Google Meta uh Microsoft they're creating these data billion dollar data centers. Broadcom is the AI highway for those data centers to even be able to reach their match maximum capacity.
So when you think about AI becoming intelligence, I view Broadcom as the Uncle Sam of AI a bit because they tax on all the workloads that are happening in AI. Then we also have a lot of like the software names that have kind of had a bad rap because nobody really want wants to live on SAS Avenue anymore in this new digital economy where uh the AI tsunami is coming for that whole sector and it's going to be a noisy for next couple years.
But you have a couple names like the Palunteers, the snowflakes, the MongoDBs that are actually providing the data rails for these enterprises to feed proprietary data into their models. So when you think about what I just said right there, it's really these models that all these LLMs are having right now.
They won't be able to learn from the business data without these three infrastructure data infrastructure software names. Uh two great terms in there, SAS Avenue and the AI tsunami. I want to focus on the second one there. What would you say is is the biggest mountain in the way of the AI tsunami in terms of a roadblock scenario?
The biggest mountain is making sure you're not a layer. I think it was really okay for a lot of these tech companies to become a layer into the tech stack because a lot of these CTOs and IT firms that we talk about like there was just so much spend in general IT of the past 5 years in the postcloud era to be to be exact but also when the chat GBT moment in happened in 2022 I think a lot of CTOs were just overspending on a lot of software companies that were just providing a layer in their stack because they felt the pressure from the CEO or the board or whatever reason why what are you doing to make sure we're ahead of this AI race.
So now these AI tools that are out there it's very easy or it's a lot more simplic simplified for these hyperscalers to create a competing product. I think a perfect example is these software companies have a lot of AI native customers open AAI especially where four years ago they were probably a lot smaller than the actual software companies that they're being used from.
Now with open eye being a $500 billion uh company, they are creating replicated services for themselves that they don't have to really go to these vendors anymore. So I I'll give you a perfect example, data dog. Data dog was a company that I was very confident in a couple years ago. Now I believe it's gen general it's a gen AI vulnerable name because uh they specialize in observability and making sure their clients don't overspend on the cloud.
Essentially I see a lot of overlaps what an AI agent can do. Open AI you was using data dog was one of the data dog's biggest customers and now they're kind of creating an observability product that replicates what data dog was doing and now they aren't going to be using data dog as much and I think the key metric to track on a perfect example of what the relationship that happened at data dog and open AI is do not track the customer count because I don't believe customers will be leaving these software companies I think they're going to be utilizing less of their services So whenever you hear a firm talk about oh the clients are increasing like we have this many clients growth year-over-year that is uh in my terms just corporate jargon you really need to track the utilization the NRR the how much they're actually improving year-over-year on the revenue front because that's the real tell of if you have an AI lifeboat in the AI tsunami or you're just on the boat screaming we're safe we're safe but really you're going to drown in a couple years.
I these analogies are fantastic. So if we're in the AI lifeboat and you know AI the driver of this fourth industrial revolution that we're seeing, how is AI as as a complete movement in changing the landscape of our everyday lives going to impact just the general public but also the broader economy.
So I think the initial uh gut reaction to that is our grid is not ready for what's about to come. And I think there is somewhat of a worry that our the AI head is going to hit the wall in the near term a lot sooner than people anticipate because of the rapid growth we're seeing in inference right now and we're going to have these data centers continue start to go live in a couple years at a rapid pace that we won't be ready for it.
So, I think the near-term bottleneck right now that we're seeing in the AI economy is these AI utility players. And that is a new theme that um we spotlighted in the future of May 15 where I think there's a brand new um thematic that's forming right now and that's the bottleneck in physical power racks cooling whoever can whichever firm can secure the data rails needed to run AI global scale.
And I think that you're seeing names like Nebius be able to supply a thousand GP300 GPUs under their Microsoft Azure deal they got u announced a couple years ago. You have names like iron who's able to uh provide cheap renewable power. I think it's three and a half cents plus the infrastructure to host all of those GPUs as well.
So like I think that there is a uh a bucket that's kind of gaining a lot of popularity where these firms used to be miners, used to be neoclouds, but now they're turning to this metered service with usage link cash flow that's being funded by the operating cache of these hyperscalers. So it's very different than the dot era where I think a lot of people are seeing the market moves on a lot of these names are just like plus 600% in the past 6 months.
I agree that causes me to like have some heartburn of like too much too quickly. Also, it's very different where back then in 20 in the dotcom era, a lot of the growth we saw in the names, they're ballooning because of VC money based on monetization of eyeballs and attention where this is purely being funded by operating cash flow of these hyperscalers that they know will eventually be able to turn the faucet on once they have all the tools necessary to max uh to to meet the capacity on their compute and cloud.
And then eventually they're gonna all the workloads are gonna be monetized. So there is a business plan almost with all this spend capback spend where back then 25 years ago I think it was more of a shot in the dark and a hope knowing that we're building the structural arteries of the internet eventually all these applications be built on top of it but nobody knew which applications and that's why it just ballooned at such a rapid pace.
This one's a little more tailored. It's just video game like numbers some of these deals. So, I think that's what's causing a lot of people to step back a bit, but I think we're all seeing how real AI is and scalability of it. Uh, and every 3 months you get reminded that the AI pie is getting bigger and bigger and there's going to be more seats at the table with uh the necessary utensils to eat that AI pie.
Well, Shai, you said it. You know, I'll throw my own analogy out here. It's the AI freight train and it's not stopping any soon, anytime soon. We appreciate you being with us at Shai Balour, the chief market strategist at Futurum Equities. Thanks again.
Comments 0