To The Moon

🐯 Hi Tigers,

the AI hardware trade is entering a new phase. For two years, the story was simple — more AI demand, more GPUs, $NVIDIA(NVDA)$ wins. But $NVIDIA(NVDA)$'s own numbers are starting to point somewhere else: the next wave of spending is spreading into memory, storage and networking. With $Broadcom(AVGO)$ reporting soon, this week's setup looks at where AI CapEx goes next — and why $Broadcom(AVGO)$'s report may be the real test of that shift.

📊 1. The AI Hardware Trade Is Changing

For the past two years, the easiest way to trade AI hardware was simple: AI demand goes up, Hyperscalers buy more GPUs, and $NVIDIA(NVDA)$ captures the biggest share of the spending.

I think that trade is now entering a different phase. 🔄

The clearest signal is not just $NVIDIA(NVDA)$'s revenue growth, but where the company is locking in future supply. $NVIDIA(NVDA)$ disclosed that its Supply and Capacity Commitments increased from $119 billion last quarter to $279 billion, with the increase primarily related to memory procurement.

That tells me the next stage of AI spending is becoming much broader. The bottleneck is no longer only compute. As clusters get larger, memory, networking, optical connectivity, storage, power and cooling all have to scale at the same time.

The question is therefore starting to shift from “How many more GPUs can $NVIDIA(NVDA)$ sell?” to something more interesting:

Where does the next dollar of AI CapEx go after the GPU?

🧠 2. Memory Is Becoming a Core Part of the AI Trade

The first place to look is memory.

A high-end accelerator is only as useful as the system around it. Blackwell, Rubin and future platforms can deliver enormous compute, but those processors also need enough memory capacity and bandwidth to stay fully utilized. That is why HBM has moved from a relatively niche DRAM product to one of the most strategically important components in the AI hardware stack.

The most direct beneficiaries are still $SK hynix(SKHY)$ and $Micron Technology(MU)$. $NVIDIA(NVDA)$’s disclosure does not mean the entire $279 billion is being spent on SSDs. What it does show is that memory is becoming important enough for NVIDIA to secure supply years in advance.

There is another reason this cycle looks different from the old memory cycle. Historically, stronger pricing led to aggressive capacity expansion, followed by oversupply. Today, a growing share of capital and advanced manufacturing capacity is being pushed toward HBM.

That creates an interesting setup: AI is increasing memory demand while simultaneously pulling capacity away from other parts of the memory market.

💾 3. HBM Is the First Layer. Enterprise SSDs Are the Second.

The memory story does not end with HBM.

Once AI servers are actually deployed, they begin consuming and generating huge amounts of data. Training datasets, model checkpoints, inference outputs and enterprise workloads all need to be stored somewhere.

That is where NAND and enterprise SSDs enter the picture.

$SanDisk Corp.(SNDK)$’s latest fiscal-year Datacenter revenue increased from $960 million to $5.153 billion, up 437% YoY. To me, that is one of the clearest signs that AI infrastructure demand is already moving further down the stack.

I would separate the opportunity into two layers: $SK hynix(SKHY)$ and $Micron Technology(MU)$ benefit directly from the growth of AI accelerators, while SNDK and Kioxia benefit from the explosion in data created once those systems are running.

This is why I increasingly think $SanDisk Corp.(SNDK)$ should not be viewed only as a traditional NAND cyclical stock. The AI component of its demand is becoming harder to ignore.

🚀 4. $Broadcom(AVGO)$ Is the Next Real Test

This is why the upcoming $Broadcom(AVGO)$ report matters more than a normal earnings event.

$Broadcom(AVGO)$’s AI semiconductor revenue increased from $8.4 billion in FY26 Q1 to $10.8 billion in Q2, while Q3 guidance calls for $16 billion, representing more than 200% YoY growth.

If $Broadcom(AVGO)$ delivers again, it would confirm that AI CapEx is not staying concentrated around one GPU ecosystem. It is spreading into custom ASICs and AI networking.

That is the real reason I am watching this report.

$NVIDIA(NVDA)$ has already shown that AI infrastructure demand remains strong. Broadcom now has to show that the spending is broadening.

🏗️ 5. Why $Broadcom(AVGO)$’s Position Is So Interesting

$NVIDIA(NVDA)$’s moat is still obvious: GPUs, CUDA and a deeply entrenched ecosystem.

$Broadcom(AVGO)$ sits in a different position.

As $Alphabet(GOOGL)$, $Meta Platforms, Inc.(META)$, $Amazon.com(AMZN)$ and other Hyperscalers scale their own AI infrastructure, custom ASICs become more attractive for workloads where efficiency, power consumption and cost matter more than maximum flexibility.

At the same time, larger AI clusters create another unavoidable problem: networking.

More accelerators mean more data moving between servers. More data means faster switches, higher bandwidth and more sophisticated network infrastructure.

That is why I think the usual “GPU vs. ASIC” debate misses the bigger picture.

The future is more likely to be GPU + ASIC, not one replacing the other.

And Broadcom doesn't need ASICs to outcompete NVIDIA — it simply needs AI infrastructure to keep expanding. Whether the compute layer runs on GPUs or custom accelerators, the data still has to move.

⚠️ 6. The Biggest Risk Is No Longer Demand — It Is Expectations

There is one obvious problem with the AI hardware trade today: everyone already knows the growth is strong.

That changes how stocks react. 📉

A company can report excellent numbers and still struggle if the market had already priced in something even better. $NVIDIA(NVDA)$'s post-earnings price action is a good reminder of that. Strong fundamentals are no longer enough by themselves.

For $Broadcom(AVGO)$, a few things stand out heading into the report: Can AI semiconductor revenue can hold up against the already high $16 billion expectation? Can custom-silicon customers keep expanding? and can the next quarter's guidance still move higher?

If $Broadcom(AVGO)$ delivers strong numbers and the stock still fails to respond, that would suggest the bigger issue is no longer AI demand — it would be valuation and expectations.

That distinction matters.

💡 7. The Next AI Trade May Be in the Infrastructure Around the Chip

For the past two years, the AI trade was concentrated at the top of the stack.

Now the chain is getting longer:

GPU / ASIC → HBM / DRAM → Networking → Optical → Enterprise SSD → Power & Cooling

$NVIDIA(NVDA)$’s jump from $119 billion to $279 billion in supply and capacity commitments shows how important memory is becoming. $SanDisk Corp.(SNDK)$’s Datacenter growth shows that AI demand is already reaching enterprise storage. $Broadcom(AVGO)$ now has the chance to prove that custom silicon and networking are accelerating as well.

So instead of asking:

Who replaces $NVIDIA(NVDA)$?

Ask yourself:

For every additional $1 of AI CapEx, which part of the hardware stack captures the next dollar?

If $Broadcom(AVGO)$ keeps delivering, the next phase of the AI trade may look very different from the last one.

$NVIDIA(NVDA)$ may still dominate compute, but the next source of Alpha could increasingly come from the infrastructure surrounding it — memory, storage, networking and everything required to keep AI clusters running.

🐯 Tiger's Corner: Your Turn!

Question: If Broadcom delivers another blowout quarter and the stock still doesn't move, is that a valuation problem — or a sign the whole AI infrastructure trade is getting ahead of itself?

Share your reasoning — thoughtful comments may receive Tiger Coins! 🪙

Investing is a long journey. Knowing when to move matters — and so does knowing when to recharge.

Markets may not slow down, but you can. Whether it’s a quick break between trading sessions or a long journey, staying comfortable helps you reset for what comes next. The new Night Trading Memory Foam Pillow features 360° ergonomic support and slow-rebound memory foam that gently conforms to your neck, plus a roll-up design that makes it easy to bring wherever you go.

https://laohu8.com/J/redeemGift?goodID=100540&type=delivery

Redeem the new Night Trading Memory Foam Pillow in Tiger Coins Mall, and stay supported for the journey ahead.