I used to think semiconductor investing was about predicting the future.
You know—the big, dramatic calls. Spotting the next NVIDIA before it explodes. Timing the downturn before everyone else panics. Riding the wave, getting out at the top, and then casually pretending it was all part of the plan.
Turns out, that’s mostly fantasy.
What actually matters—what really separates people who get destroyed from people who quietly win—is something far less exciting and far more uncomfortable:
Understanding demand… and watching how capital gets allocated when nobody knows what demand actually is.
Welcome to semiconductor cycles. Where certainty goes to die, and spreadsheets pretend to be crystal balls.
The Illusion of Predictable Demand
Let’s start with demand, because that’s where all the stories begin.
Semiconductors power everything—phones, data centers, cars, AI, your fridge if it’s feeling ambitious. So logically, demand should be steady, right? Growing, maybe even predictable.
Nope.
Demand in semiconductors is not just cyclical—it’s emotional.
It swings based on:
- Consumer upgrades (or lack of them)
- Enterprise spending cycles
- Government incentives
- Hype waves (looking at you, AI)
- Inventory corrections that feel like hangovers from decisions made two years ago
Demand isn’t a straight line. It’s a mood.
And the industry responds to that mood with the kind of overreaction that would make a day trader blush.
The Classic Boom-Bust Loop (That Everyone Pretends Is New Every Time)
Here’s how it usually goes:
- Demand starts rising
- Companies report strong earnings
- Everyone assumes demand will continue rising forever
- Capital floods into the industry
- Capacity expands aggressively
- Demand slows (because it always does)
- Suddenly there’s too much supply
- Prices collapse
- Everyone acts shocked
Repeat.
Every. Single. Time.
And yet, every cycle is treated like it’s unprecedented. Like this time is different. Like we’ve finally figured it out.
We haven’t.
We just have better PowerPoint decks now.
Capital Allocation: Where the Real Story Lives
Demand gets all the headlines, but capital allocation is where the real decisions happen.
Because when demand looks strong, companies don’t just celebrate—they spend.
They build fabs. They expand capacity. They invest billions into infrastructure that takes years to come online.
Which means they’re not investing based on current demand.
They’re investing based on what they think demand will be years from now.
Let that sink in.
You’re making multi-billion-dollar decisions today based on a guess about the future, in an industry where the present is already hard to understand.
What could possibly go wrong?
The Timing Problem Nobody Can Solve
Here’s the fundamental issue:
By the time new capacity comes online, the cycle has often already shifted.
So you end up with this bizarre situation where:
- Companies invest heavily during peak optimism
- That capacity arrives during declining demand
- Oversupply crushes margins
- Companies pull back investment
- Demand eventually recovers… but now there’s not enough supply
It’s like trying to fill a bathtub while blindfolded and constantly turning the faucet based on what you think the water level is.
You’re always late.
AI: The New Narrative, Same Old Behavior
Right now, the industry is riding the AI wave.
And to be fair, it’s real. The demand for AI chips, GPUs, and data center infrastructure is massive. Companies like Advanced Micro Devices and NVIDIA are seeing growth that looks almost absurd.
But here’s where I get suspicious.
Not because AI isn’t important—but because of how the industry behaves around any strong narrative.
When demand spikes in a specific segment, capital doesn’t just flow there—it floods.
Everyone wants exposure. Everyone wants capacity. Everyone wants to be part of the story.
Which raises an uncomfortable question:
How much of this demand is sustainable… and how much is just the early phase of another cycle?
The Inventory Ghost
One of the most underrated forces in semiconductor cycles is inventory.
Not exciting. Not headline-worthy. But incredibly powerful.
Because demand isn’t just about end users—it’s about what’s sitting in warehouses.
When companies overestimate demand, they build inventory. Lots of it.
And when reality doesn’t match expectations, that inventory has to be worked through.
Which means even if end demand is stable, orders drop.
Because customers aren’t buying—they’re consuming what they already have.
This creates the illusion of collapsing demand, even when the underlying market hasn’t changed that much.
And it messes with everyone’s expectations.
The Capital Discipline Problem
In theory, companies should learn from past cycles.
They should invest more cautiously. Allocate capital more efficiently. Avoid overbuilding during peaks.
In theory.
In reality, incentives don’t work that way.
When demand is strong, executives are rewarded for growth. Investors demand expansion. Missing an opportunity feels worse than overextending.
So capital flows aggressively during the best times—exactly when it should probably be most restrained.
And restraint only shows up after the damage is done.
Governments Enter the Chat
As if the cycles weren’t chaotic enough, now we have governments heavily involved.
Subsidies, incentives, national security concerns—all pushing for more domestic semiconductor production.
Look at programs like the CHIPS and Science Act.
Billions of dollars flowing into the industry.
Which sounds great—until you remember how cycles work.
Because now you’re layering political incentives on top of already volatile capital allocation decisions.
Which could mean:
More capacity.
More investment.
And potentially… more oversupply.
But with a patriotic spin.
The Myth of Perfect Timing
Every investor wants to time the cycle.
Buy at the bottom. Sell at the top. Capture the full wave.
Good luck.
Timing semiconductor cycles perfectly is like predicting weather patterns three years in advance.
You might get close. You might even get lucky once or twice.
But consistently? No chance.
The better approach—if there is one—is to understand where you are in the cycle and adjust expectations accordingly.
Which is less satisfying than calling the exact top, but significantly more realistic.
What Actually Matters (Even If It’s Boring)
After years of watching this industry, I’ve realized something:
The companies that survive cycles best aren’t the ones with the best narratives.
They’re the ones with the best discipline.
- Strong balance sheets
- Flexible capital allocation
- The ability to scale up and pull back
- A focus on long-term returns, not short-term hype
Which is not exciting.
There’s no viral post about “disciplined capital allocation.”
But it’s what keeps companies alive when the cycle turns.
The Emotional Side of a Technical Industry
We like to think semiconductors are purely technical.
Engineers. Physics. Precision.
But the market around them? Completely emotional.
Fear. Greed. FOMO. Panic.
You see it in how quickly sentiment shifts.
One quarter, demand is “unstoppable.”
The next, it’s “collapsing.”
The reality is usually somewhere in between, but the narrative swings wildly.
And those swings drive capital decisions.
Why I Stopped Trying to Be Smart
At some point, I stopped trying to outsmart the cycle.
Not because I got worse—but because I realized how much of this is fundamentally unknowable.
You can analyze demand. You can model scenarios. You can study past cycles.
But you can’t control:
- Timing
- Sentiment
- Unexpected shocks
- Policy changes
- Technological shifts
So instead of trying to predict everything, I focus on understanding the structure.
How demand behaves.
How capital flows.
How cycles repeat, even when they look different.
It’s less glamorous. But it’s more useful.
The Real Game
Semiconductor investing isn’t about predicting the future.
It’s about surviving uncertainty.
It’s about recognizing that:
- Demand will fluctuate
- Capital will overreact
- Cycles will repeat
- Narratives will change
And positioning yourself accordingly.
Not perfectly. Just… intelligently.
The Part Nobody Wants to Admit
Here’s the uncomfortable truth:
Most people don’t lose money in semiconductor cycles because they don’t understand the technology.
They lose money because they believe the story.
They believe that this time is different.
That demand will keep growing.
That capital allocation will be rational.
That the cycle has been tamed.
It hasn’t.
It never is.
Where We Are Now (Probably)
If I had to guess—and it is a guess—I’d say we’re somewhere in the middle of a cycle that feels like the beginning.
AI demand is strong. Investment is accelerating. Optimism is high.
Which is exactly when things feel the most certain.
And certainty, in this industry, is usually a warning sign.
Final Thought: Respect the Cycle
I don’t try to predict semiconductor cycles anymore.
I respect them.
I assume:
- Demand will surprise me
- Capital allocation will overshoot
- Narratives will mislead
- And the cycle will turn when it’s least convenient
Because it always does.
And the moment you think you’ve figured it out…
That’s usually when it reminds you that you haven’t.
Comments
Post a Comment