Denzel Double Down

Tech Stocks, Ball and Drake Lyrics

[Denzel Double Down @DenzelDubleDown]

I expect Micron $MU to become a key name in the marketplace over the next few years. Micron is one of 3 companies that make HBM (high bandwidth memory). Why does this matter?

HBM is key to help GPUs overcome traditional memory limitations for high compute applications (model training, inferencing, edge computing, simulations/modeling).

Only US based HBM manufacturer (Samsung and SK-Hynix are both Korea based).

Micron's entire 2024 HBM supply and over 50% of 2025 supply sold out by March 2024. Supply can not keep up with demand.

Micron's HBM has recently entered the supply chain for Nvidia and Google.

2024 revenue grew to $25.1B +62% YoY, and returned to profitability with a net income of $778M. Looking ahead, the company is well positioned for continued growth as tariffs and onshoring of advanced hardware technologies create more favorable conditions for Micron.

[Denzel Double Down @DenzelDubleDown]

@littlejay53 What do you think they should do?

[Denzel Double Down @DenzelDubleDown]

The way back is always faster

[Denzel Double Down @DenzelDubleDown]

Lines are being drawn between $NVDA + Neocloouds vs hyperscalers.

As hyperscalers continue to try and penetrate GPU market share with ASICs, I expect $NVDA to partner more with neoclouds and eventually try to flip the script and release some of their own software products. We are already seeing this with Nvidias rev share partnership with neoclouds to sell and host compute.

If this does go back and forth, only vertically integrated hyperscalers that are actually producing cash from AI can win (survival is different), imo. These are $GOOG and $AMZN. Others will have to withdraw and play both sides.

Another interesting point here is that independent model companies will have to be very strategic since Nvidia is more in the open source camp and hyperscalers obviously want to deploy their own models.

[Denzel Double Down @DenzelDubleDown]

Very strong argument for a $MU multiple re-rate. Memory is more structural than cyclical now. Anyone telling you otherwise is in denial. Should become even more cemented once robotics takes off per management in the earnings call:

"Humanoid robots carry 10 times the amount of memory as an average L2+ vehicle, and we expect a sustained, substantial, multi-decade memory demand cycle to begin in the latter part of this decade".

Also, more revenue is being locked in through their SCAs (Strategic Customer Agreements) and thus of higher quality and predictability than one time spot purchases. Management has said up to 50% of revenue could come from these agreements compared to ~25% contracted today.

[Denzel Double Down @DenzelDubleDown]

Top 10 IPPs. Goal was to map out current functioning GWs across all power sources and get a replacement value based on that.

Each power source valued by an average between cost to build and precedents.

Also looked at an EBITDA multiple to account for uptime and efficiency since Total GWs assumes 100% functionality.

This is obviously very preliminary but $TLN, $VST, and $CEG look interesting. Let me know thoughts

[Denzel Double Down @DenzelDubleDown]

@Midnight_Captl Easily a beat and raise. Mag7 is raising capital at scale for the first time in years

[Denzel Double Down @DenzelDubleDown]

@zephyr_z9 Feel Google is more positioning towards consumer with their full stack across office products? I think they can gain in consumer while Anthropic and Open AI go after enterprise. Not a good look regardless though

[Denzel Double Down @DenzelDubleDown]

Good $PLTR bull case. Also hints that models will become commoditized.

Value accrues to how models will be harnessed and applied.

[Satya Nadella @satyanadella]

https://t.co/vLmiBKTtX3

[Denzel Double Down @DenzelDubleDown]

Almost everything everyone is talking about as "next up", $NVDA is already working on. Robotics, autonomous vehicles, world models and simulation, CPUs for agentic workloads, LLMs enabled to run native on edge devices, and photonics.

On top of all of this they grew top line at 65% on $215B and are the 3rd cheapest MAG 7 on a forward P/E.

[Denzel Double Down @DenzelDubleDown]

Recent @SemiAnalysis_ report supporting our take from December.

[Denzel Double Down @DenzelDubleDown]

2/ The market has two pure plays, $SNPS and $CDNS. Both companies are worth <$100B, but their customers and end markets are worth trillions. While I am long both stocks, I like $CDNS better. I believe their strategy, shown below, is better positioned to capture the value from the AI revolution.

[Denzel Double Down @DenzelDubleDown]

@damnang2 You can't vibecode semiconductors. Tough hold through the SaaS drop but am still long

[Denzel Double Down @DenzelDubleDown]

Currently diving deep into $ANET. "AI infra" name that has not run up nearly as much as others and is down a good chunk off highs.

Sell side consensus bear case for $ANET right now is too short term and does not consider long term architectural changes. Bull case is also not taking into account the disaggregation of inference and the sliding scale of traditional inference vs agentic inference workloads.

Lots of opportunity for understanding the technicals here in my opinion

[Denzel Double Down @DenzelDubleDown]

Anthropic is paying SpaceX $1.25B a month just to use their GPUs and people don't want to own $NVDA and other infra names

[Denzel Double Down @DenzelDubleDown]

Donovan Mitchell is Top 10, and still better than Cade Cunningham

[Denzel Double Down @DenzelDubleDown]

Thoughts on $GOOG earnings:

Obviously the smallest of the 3 clouds, but growth rate is insane and I think they continue to take more share through vertical integration, priority on security (Wiz), and I also liked the mention of their custom CPU with agentic workloads driving CPU demand being such a hot thing right now.

Feels like they are one of the first big tech companies beginning to see actual B2B and B2C monetization from CapEx. Growth rates are up across business segments - Cloud, Youtube, Search, Gemini, etc

Newer focus on UCP is super interesting too, laying foundations for agentic commerce inside the Google ecosystem

Waymo stuff continued to look good. I like the partnership with DoorDash.

CapEx obvious raise, the entire market is compute constrained, another datapoint supporting $NVDA as an easy hold.

This is one of the best companies on earth and maintains a sizable position in the portfolio. Up 115% the last year. Sometimes simple ball > fancy ball

[Denzel Double Down @DenzelDubleDown]

$TSLA cited pricing pressure in their energy storage business as part of the reason for the weak quarter. They are the top player in the space in North America. This directly supports UBS recent trim on $FLNC citing potential pricing pressures from EV battery supply getting redirected to utility scale. But nobody is talking about this connection?

$FLNC bear thesis does seem to be overstated? Can we really expect alternative battery suppliers to redirect supply away from EV batteries and into mass, utility scale batteries? Also factor in Fluence's vertical integration with their own software stack? Lots of these players aren't even American companies and foreign companies are having an extremely hard time getting new integrations into American power/tech stacks. We have heard this "commoditized" argument before with HBM and LLMs.

Putting on the watchlist for sure and would love any additional color on this. If bear case is weak, $FLNC seems like a buy right now.

[Denzel Double Down @DenzelDubleDown]

$TSLA stock not buyable until they show step function change in FSD, Robotaxi, or Optimus. As much as I hate this to be true, it is. Not selling but not buying either.

[Denzel Double Down @DenzelDubleDown]

If SpaceX IPOs at 2T, or even 1.75T, I don't want to hear a lick about $PLTR valuation?

[Denzel Double Down @DenzelDubleDown]

The path to success often requires not doing certain things as much it requires doing certain things.

“Via negativa” - growth by subtraction

[Denzel Double Down @DenzelDubleDown]

My contrarian take is that it still makes sense to learn to code. This is because the core purpose of coding is not task completion, but problem solving. As AI makes coding more accessible, a good coder becomes exponentially more valuable. They are able to solve problems at a higher level and faster rate. There are always more problems to solve for a business, while tasks are finite. It is more likely to me that those that know how to code eat into every other vertical rather than vice versa.

I believe much of the SWE market that got laid off was due to corporate bloat. Jobs have bottomed and software engineers are one of the only roles increasing in HC right now. Look at every leading edge tech company, they are still hiring endless engineers, and well positioned trad corporates are expanding their tech departments. At the end of the day the pre requisite knowledge and systems level thinking of good coders make them one of the most asymmetrically positioned roles to benefit from mass AI deployment.

This is a good thought excitement for how AI would impact a specific job. Is the job task completion (will be automated) or is it problem solving (endless opportunity)?