A bad week does not invalidate a long arc
OppenFolio is down about 5% right now.
That’s uncomfortable, but not unusual. We saw the same thing in December. A fast selloff, a lot of confident doom talk, and then — just as fast — a recovery that erased the losses within a week.
I didn’t sell then. I’m not selling now.
This post is my attempt to explain why, without pretending volatility doesn’t exist, and without dismissing the bear arguments as stupid. They aren’t. They’re just incomplete.
At the center of this discussion is Nvidia, which OppenFolio treats as a proxy for AI generally, not as a speculative lottery ticket. When NVIDIA sells off, it’s rarely about NVIDIA the company. It’s about anxiety around AI as a whole.
So let’s talk about that anxiety honestly.
Yes, this is a bubble — and that doesn’t mean what people think
A lot of the videos making the rounds right now are not wrong.
Valuations ran too far, too fast. Capital piled in aggressively. Narratives got ahead of execution.
That is, by definition, a bubble.
But “bubble” does not mean “everything goes to zero.” It means price discovery overshot reality and now has to recalibrate. The dot-com era is the right comparison — just not in the way most people use it.
The people who were wiped out in 2000 weren’t those holding broad market exposure and staying solvent. They were the ones concentrated in small, fragile companies with no real business underneath them.
Amazon didn’t disappear. Microsoft didn’t disappear. The internet didn’t disappear.
What disappeared were the Pets.coms — the companies whose entire value proposition was momentum.
So the real question isn’t “will the market crash?”
The question is:
Does anyone seriously believe NVIDIA is a Pets.com-style zero?
If the answer is no, then the rest of the conversation becomes about time, volatility, and allocation, not existential risk.
The real bear arguments — steel-manned
There are three serious bear arguments I hear over and over. I want to present them clearly, because hand-waving them away helps no one.
- “Models are getting more efficient. GPUs won’t be needed as much.”
This argument sounds technical, but it’s actually economic.
The idea is that as models become more efficient, the amount of compute required per unit of output drops, reducing demand for high-end GPUs.
The problem with this logic is that it ignores Jevons Paradox, which has played out over and over again in history: when a resource becomes cheaper, usage doesn’t shrink — it explodes.
We already see this happening. As inference costs drop, companies stop rationing usage. AI moves from batch jobs to real-time systems, from selective use to ambient use, from “should we run this?” to “why wouldn’t we?”
Efficiency doesn’t end demand here. It unlocks it.
- “Memory is the real bottleneck. HBM caps growth.”
This is the strongest bear case, and it’s real.
High-bandwidth memory is scarce, expensive, and currently one of the dominant cost drivers in advanced AI accelerators. It limits how fast complete systems can ship.
But scarcity is not fragility. It is a price signal — and one so loud that it’s already reshaping capital allocation, chip design, and software architecture.
Every time humanity has hit a constraint this valuable, three things have happened:
• capacity expanded • workarounds emerged • software adapted
Betting against this getting solved is not a bet against NVIDIA. It’s a bet against coordinated incentives across capital, governments, and engineering.
That’s a historically weak bet.
- “This is another dot-com collapse waiting to happen.”
This one is emotionally powerful, because it did happen before.
But again, the collapse didn’t destroy the winners. It destroyed the fragile intermediaries.
AI is not a handful of content studios deciding whether to invest in 8K cameras. AI is every individual, every team, every company, producing and consuming their own “content” constantly.
There is no centralized choke point that can say “we’re done investing.” Competitive pressure enforces adoption.
This is not a discretionary upgrade cycle. It’s an arms race in productivity.
Why the HDTV analogy fails
People often compare AI to HDTV: rapid improvement, then a plateau at “good enough.”
The analogy breaks for one reason: content control.
HDTV stalled because a small number of studios controlled the pipeline and didn’t see enough marginal return to justify the leap to 8K.
AI has no such gatekeepers.
Every user is a content producer. Every company is under pressure to grow. Every improvement feeds back into usage.
This creates an inverted demand curve. Better capability generates more demand, not less. That is the opposite of what happened with TVs.
What OppenFolio is actually betting on
OppenFolio is not betting that markets go up in a straight line. They never do.
It’s betting on something simpler and more durable:
• that humans will keep trying to do more with less time • that tools which multiply individual capability will keep spreading • that capital will continue to flow toward whatever enables that
NVIDIA sits at the center of that system today, not because it’s fashionable, but because it coordinates hardware, software, and developer ecosystems at the frontier.
Markets can panic about RAM. They can panic about model efficiency. They can panic about valuation.
None of that changes the direction of the underlying capability curve.
The dot-com lesson people skip
If you take any long-term market chart and cut it off at the crash, it looks like a cliff.
If you extend the chart another decade, it looks like a brief dip followed by a higher plateau.
That has been true of every major technological shock in modern history, assuming the broader world continues to function.
Corrections hurt. Over-leverage dies. Fragile companies vanish.
But the systems that matter recover — and then normalize at levels that once seemed absurd.
The close
I’m not dismissing the bears. They’re right that this is a bubble. They’re right that corrections happen. They’re right that some things won’t survive.
Where I disagree is the conclusion.
Betting against NVIDIA at this stage is not really a bet against a company. It’s a bet that:
• humans stop pushing capability forward • competition stops enforcing adoption • ingenuity fails to route around constraints
History is not kind to those bets.
Markets may drop. They may even drop hard. But as long as people keep building, learning, and competing, the curve underneath the noise keeps rising.
That’s the curve OppenFolio is aligned with.
And that’s why I’m still holding.
You can always reach The Architect at [email protected] if you want to go deeper.
Disclaimer: This post reflects personal opinions and is not financial advice. OppenFolio is not an investment advisory service. See site disclaimer for full details.