Don't Buy Every AI Dip. Buy the Companies Controlling THIS — Transcript
Full transcript
- 0:00I think the market is like severely
- 0:02overreacting to this because there is a
- 0:05difference between
- 0:06saying that the new data center
- 0:08construction needs better oversight and
- 0:10saying that companies can no longer
- 0:12build data center
- 0:14data centers. And those are not the same
- 0:17thing because I was looking at Hinojos
- 0:19own website and the priorities she
- 0:22outlined are quite specific. She called
- 0:25for stronger labor protection, requiring
- 0:28developers to employ skilled American
- 0:30union workers and giving local
- 0:33communities a greater voice in whether
- 0:36these facilities should be built near
- 0:39them. She also proposed greater
- 0:41transparency including restrictions on
- 0:43non-disclosure agreements that can
- 0:45prevent residents from understanding the
- 0:48details of what's actually being
- 0:49developed in the communities. And two of
- 0:52her biggest priorities relate to
- 0:55electricity and water. She wants data
- 0:58center developers to demonstrate that
- 1:00surrounding households won't face higher
- 1:03electricity or water cost because of new
- 1:05projects, you know.
- 1:08And she wants developers to essentially
- 1:10take more responsibility to financing
- 1:12their own infrastructure rather than
- 1:15depending on tax paying dollars or
- 1:18subsidies.
- 1:19And finally she proposed stronger
- 1:21protections for
- 1:22you know waters local water supplies and
- 1:25particularly in situations where large
- 1:28industrial water contracts would
- 1:30threaten the resources available to
- 1:34residents. Is any of this crazy? Like I
- 1:37don't think what she's asking is crazy
- 1:39at all. I think these are pretty
- 1:42reasonable ask. Like if you want to
- 1:45build data centers in our community,
- 1:48you have to work with the community a
- 1:50little bit more better. You have to
- 1:52demonstrate that your projects won't
- 1:56place an unfair financial burden on the
- 1:58people living there. I mean, that is
- 2:00what everyone should realistically do,
- 2:04but some of these media outlets like to
- 2:06spin this, you know, spin these
- 2:09takes as, oh, they're shutting data
- 2:11centers down. No, I think
- 2:14data centers provide an enormous amount
- 2:16of jobs in the community.
- 2:18It's it's growing like it's the largest
- 2:22part part of our GDP like growth as of
- 2:25now.
- 2:26If
- 2:27the data center build out
- 2:29gets slower,
- 2:31I think we're going to severely because
- 2:33severely be
- 2:36potentially going towards a negative GDP
- 2:38growth and if you have two two
- 2:41consecutive months of negative GDP
- 2:42growth, congratulations, you're on
- 2:45you're on recession right now. You have
- 2:47a recession in your hands. So nobody
- 2:48wants that. I don't think Democrats or
- 2:52Republicans want a recession in their
- 2:53hands because
- 2:55now Democrats and Republicans will point
- 2:57fingers at each other and say, oh, you
- 3:00caused it, you caused it, but in reality
- 3:02both of them will cause
- 3:05cause this if if if the build out don't
- 3:08continue. So it is essential for this
- 3:11build out
- 3:12to take place and I think companies need
- 3:15to work with these communities to build
- 3:18these data center out. Now, here's why I
- 3:21don't think the AI build out is going to
- 3:24slow down anytime soon
- 3:26because AI build out is already well
- 3:29underway. We all know that. Amazon,
- 3:31Google, Microsoft, Meta
- 3:34collectively are spending billions,
- 3:36okay, billions of dollars. Look at these
- 3:39look at these three CAPEX estimates, you
- 3:42know. I think this just got revised just
- 3:44this week alone or last week.
- 3:47Meta, Google, Amazon, Oracle, CoreWeave,
- 3:49Nebius, SpaceX
- 3:51could increase from approximately two or
- 3:55did increase from approximately 280
- 3:57billion in 2024 to 951 billion in 2026.
- 4:02By 2027 that projected figure is going
- 4:06to reach 1.37
- 4:08trillion.
- 4:10Okay, and by 2028 we're talking about
- 4:121.5 trillion
- 4:15going into these data centers. Now, I
- 4:17think there is a reasonable possibility
- 4:19that spending projects could continue
- 4:21moving higher. Why? Because we're still
- 4:26relatively early in the development of
- 4:28AI.
- 4:29Think about how much AI has changed over
- 4:32the past few years. We have essentially
- 4:35gone from large language models that
- 4:37could answer questions
- 4:39and generate text to increasingly
- 4:41sophisticated AI agents that runs for
- 4:44days.
- 4:45And every new generations of these
- 4:47system requires enormous computing
- 4:50resources. And as more businesses
- 4:53integrate AI into their everyday
- 4:54operations,
- 4:56the computing power required to serve
- 4:58those application could increase term
- 5:00significantly.
- 5:02>> We're going to pause for a second.
- 5:04The AI trade is changing. The bottleneck
- 5:06stocks are either running out of steam
- 5:08or getting ready for another leg higher
- 5:10while the application layer is finally
- 5:11starting to show some real profits. Our
- 5:13analysts were early to the last wave and
- 5:15now they are buying for the next one.
- 5:18Catch your next call inside Milk Road
- 5:19Pro for just a dollar at the link below.
- 5:21>> And we're still so early. That's what
- 5:23I'm saying. We're probably in the second
- 5:26or the third inning of this AI
- 5:28infrastructure build out. And
- 5:30uh as investors, the profits may not
- 5:33come till the sixth, seventh, or the
- 5:34eighth or even the ninth inning of the
- 5:36majority of the profits. We're starting
- 5:38to see the profits come in, but the
- 5:40infrastructure has to get built out
- 5:42before before the profits come in. So
- 5:45another reason I think that we're not
- 5:48these this CAPEX number is going to go
- 5:49higher is because competition. We all
- 5:52know hyperscalers are competing against
- 5:54each other. If Microsoft decides to
- 5:57significantly reduce its AI
- 5:58infrastructure while Amazon continues to
- 6:00expand AWS,
- 6:02Microsoft essentially risk lose
- 6:05losing customers that need access to you
- 6:07know, additional computing capacity to
- 6:10Amazon. If Google slows down its
- 6:12investment in AI infrastructure while
- 6:14Microsoft and OpenAI keep pushing
- 6:16forward, Google risks losing ground in
- 6:19enterprise AI and cloud computing. If
- 6:22and if Meta decides to
- 6:25you know, reduce spending on AI models,
- 6:27somebody else is going to take take
- 6:29their position. These companies
- 6:31understand that and the because of the
- 6:33potential long-term market is so
- 6:35enormous, the TAM is like so so huge
- 6:39because we're talking about human
- 6:41workers are, you know, essentially
- 6:43getting replaced by AI. So the TAM is a
- 6:46lot here.
- 6:48So
- 6:49I don't think this CAPEX is going to
- 6:52slow down. Yes, it might slow down like
- 6:55in terms of like the dollar amount
- 6:56because it it's just a large it's just a
- 6:59numbers game. It's going to get smaller,
- 7:02right? It can't keep going higher every
- 7:04year.
- 7:05But what you have to understand is like
- 7:08companies are still spending
- 7:10billions and trillions of dollars to
- 7:12build out this AI infrastructure.
- 7:15And
- 7:16as I said, revenue is going to be a big
- 7:18deciding factor on all this. Revenue
- 7:20hasn't really come in yet.
- 7:23Okay, we're still early. So it will
- 7:26eventually start to come in because
- 7:28yesterday alone we had couple great
- 7:31earnings
- 7:32from companies like AMD.
- 7:35AMD reported
- 7:37significant growth in their data center
- 7:39businesses. I think they were at one
- 7:40point they were up like 6 or 7% after
- 7:42hours.
- 7:43AMD generated $13.4 billion in total
- 7:47revenue representing a 45%
- 7:50year-over-year growth in its data center
- 7:52business, generated approximately 8 and
- 7:551/2 billion dollars, up 97%
- 7:58from previous last year.
- 8:00It doubled in 12 months.
- 8:02And AMD isn't the only one benefiting
- 8:04from this expansion. Arista Networks has
- 8:07been
- 8:08reported strong demand for high
- 8:10performance networking equipment. They
- 8:11also had their earnings yesterday.
- 8:13Massive beat. Supermicro, another name
- 8:17experienced significant growth tied to
- 8:19AI server developments.
- 8:22There's so many earnings that came out
- 8:24yesterday all pointed that AI demand is
- 8:28stronger than ever before. And of course
- 8:30the stocks are down or a little bit flat
- 8:32because of weakness in semis and that
- 8:34that's fine.
- 8:36And even look at GPU rentals, okay?
- 8:39Nvidia B200 and H100 GPUs have had
- 8:43significantly
- 8:44like went up during the year and it's
- 8:46still continued to go up.
- 8:48And H100 is isn't even the Nvidia's
- 8:51newest generation. It's it's so old.
- 8:54If like these older computing hardware
- 8:57is still commanding rental high prices,
- 8:59suggest that there may be more
- 9:01underlying demand for capacity than
- 9:03compute is available. Okay? I I you guys
- 9:06need to understand it. Understand this.
- 9:09This is compute need for compute is
- 9:11going up. It's not slowing down.
- 9:15Now let's let me bring it back to the
- 9:16semiconductor sell off because I think
- 9:18investors this is where investors could
- 9:20be missing the opportunity.
- 9:22I think the market seems to be assuming
- 9:24that tighter restrictions on data center
- 9:26will translate into lower demand for
- 9:29semiconductors, but I don't think it's
- 9:31true. Even if data center approvals
- 9:34become difficult, that doesn't mean
- 9:35semiconductor demand disappears. It's
- 9:38this existing this facilities still need
- 9:40more powerful GPUs,
- 9:42bandwidth memory networking equipment,
- 9:45yada yada yada. And if developers can
- 9:47build any new facilities as they were
- 9:49originally planned,
- 9:50they have a greater incentive to
- 9:53maximize the computing capacity of the
- 9:54infrastructure they already have.
- 9:57Think about it. If Amazon or Microsoft
- 9:59already has an operational data center
- 10:01with access to electricity, cooling
- 10:04systems, yada yada yada, that facility
- 10:07becomes increasingly valuable if new
- 10:09construction becomes more difficult.
- 10:11So if you're a neo cloud provider that
- 10:13has
- 10:15these available capacity, your
- 10:18your spot price just went up
- 10:20because if there is less compute to go
- 10:23around,
- 10:24who benefits? People do that have
- 10:26computing capacity.
- 10:28And now look at this chart.
- 10:30This is the AI stack right here.
- 10:33Broadcom companies like Nvidia, AMD,
- 10:35Broadcom, Credo, Light, AAAOI
- 10:39could continue benefiting. Let's start
- 10:41with Nvidia. We all know Nvidia remains
- 10:43one of the most suppliers in the entire
- 10:45ecosystem. Regardless of whether
- 10:47Microsoft or Google or another company
- 10:49starts to slow down the AI race, Nvidia
- 10:51continues selling their accelerator
- 10:53networking technologies to help power
- 10:56their infrastructure. And then AMD is in
- 10:59the same position. Broadcom is another
- 11:01company I think particularly interesting
- 11:04because it's exposure to custom silicon
- 11:06and networking. And Marvell. Marvell is
- 11:09thinking about is going to acquire Light
- 11:11and they're going to that's going to
- 11:12create a massive moat in their business
- 11:14as well. So these are all some of the
- 11:17companies that are benefiting from the
- 11:19AI infrastructure demand.
- 11:21And then
- 11:22as I mentioned like Credo, Lumentum, Ply
- 11:25Electronics are like helping remove one
- 11:28of the bottlenecks in the industry,
- 11:29which is like networking. And then we
- 11:31have stuff like memory, companies like
- 11:33Micron, SK Hynix, Samsung has
- 11:36you know benefited from growing demand
- 11:38for high bandwidth memory and other
- 11:39memory products.
- 11:41So you get it. You have to understand
- 11:43there's so many bottlenecks right now
- 11:44like power, memory, substrates,
- 11:48um CPU, GPU like there's the entire
- 11:52supply chain is constrained right now.
- 11:55And you have to understand demand for
- 11:57all this is only going higher.
- 12:00And that brings me to the broader
- 12:02opportunity. I think the investors are
- 12:04still trading the AI infrastructure
- 12:07as though every company faces the same
- 12:09regulatory and financial financing
- 12:11risks, but the reality is much more
- 12:13complicated.
- 12:14A semiconductor selling into multiple
- 12:17hyperscalers is very different from a
- 12:19neo cloud.
- 12:20A or a lower tier, I should say lower
- 12:24tier neo cloud that borrows billions of
- 12:25dollars to fund build a new facility in
- 12:27Texas.
- 12:29And even within the neo cloud industry,
- 12:31I've told you guys this Core Weave and
- 12:33um Nevius still remains the strongest.
- 12:36They can actually get the funding. And
- 12:38so I think this this sell off could
- 12:40actually create some very attractive
- 12:42points. And um but I would definitely
- 12:45focus on companies with strong demand,
- 12:47real customers,
- 12:49reasonable financing needs and a
- 12:51credible path flow to generating cash
- 12:54flow.
- 12:55I think there's a lot of opportunities
- 12:57that you can buy here. This is a buy the
- 12:59dip moment. I'm going to buy some
- 13:02companies here. And if you want to know
- 13:04what companies I'm buying, you know, you
- 13:05can just subscribe to Milk Road for just
- 13:07just it's just a dollar. You can come
- 13:08check it out.
- 13:09>> Thanks for listening to Milk Road. If
- 13:11you enjoyed the show, make sure you like
- 13:12and subscribe. [music] And if you're
- 13:13struggling to find winners in the
- 13:15market, that's exactly what Milk Road
- 13:16Pro is built for. [music] Our analysts
- 13:18have called some of the biggest winners
- 13:20early and Pro lets you see what they're
- 13:22buying next, every trade they make, and
- 13:24the research behind every position.
- 13:25Check out Milk Road Pro at the link
- 13:27below. Everything you hear on Milk Road
- 13:29is for informational purposes only.
- 13:31These are our personal opinions, not
- 13:33financial advice, and we may own some of
- 13:35the investments we talk about. Always do
- 13:36your own research and make the decisions
- 13:38that are right for you. See you next
- 13:40time.
About this transcript
This page contains the full transcript of Don't Buy Every AI Dip. Buy the Companies Controlling THIS by Milk Road Stocks, generated from the public captions YouTube serves with the video. The transcript has 2,061 words across 362 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.