There is a famous psychology experiment where researchers ask you to count basketball passes between players wearing white shirts. A person in a gorilla suit walks through the middle of the frame. Beats its chest. Walks off. And roughly half of all participants never see it.
Not because they’re careless. Because they’re focused.
How I Almost Missed the NVIDIA Bear Case
I almost made this exact mistake with NVIDIA.
I had just spent four full steps of the GAMMERS process building the bull case. And what a bull case it was. Revenue from $27 billion to $216 billion in three years. A CUDA ecosystem 20 years deep with 4 million developers and 1.5 million AI models that no competitor has replicated. A fortress balance sheet with $54.1 billion in net cash. Jensen Huang’s track record of surviving seven crises and emerging stronger from six of them.
Four steps. Dozens of data points. Every one of them legitimate. Every one of them sourced from public filings, earnings transcripts, and verified financial data.
And every single one of them bullish.
That’s the problem. Not that the data was wrong. The data was right. The problem was what the data wasn’t showing me, because I wasn’t looking for it.
When I got to the R-step of GAMMERS, the step that exists specifically to destroy the thesis I just built, I ran a bias audit. The audit requires you to count your findings across all prior steps. Bullish findings on one side. Bearish findings on the other.
The count came back approximately 25 bullish findings to 8 bearish. A 3:1 ratio. And in my process, a 3:1 bull-to-bear ratio triggers a formal flag: confirmation bias is likely at work.
I want to be clear about what this means. It does not mean the bull case is wrong. It means the pipeline has been building a one-sided case. The evidence I gathered was accurate. But the evidence I didn’t gather was the gorilla walking through the frame.
What was the gorilla? Several of them, actually.
Two customers represent 36% of NVIDIA’s $215.9 billion in revenue. Those same two customers are simultaneously spending billions building their own alternatives to NVIDIA’s chips. Google’s TPU is in its 7th generation with over a decade of production experience. Amazon deployed 500,000 Trainium2 chips for Anthropic’s model training. Broadcom’s custom ASIC AI revenue grew from $3.8 billion to $19.9 billion in three years.
These aren’t hypothetical threats. They’re happening now. And across four steps of analysis, they received a fraction of the attention that the bull case received.
Then there was the antitrust gorilla. Active investigations in four jurisdictions, the US, EU, France, and China, with France having prepared formal charges carrying theoretical penalties up to 10% of global revenue. That’s $21.6 billion. The analysis documented the existence of these investigations but never analyzed their probable outcomes.
And the insider selling pattern. Not a single discretionary insider purchase across the entire executive team and board over 31 months. Jensen Huang sold $1.78 billion in shares via pre-planned 10b5-1 plans. Proportionally small relative to his $162 billion position, sure. But the complete absence of anyone backing up the truck with their own money is a data point that deserved more weight than it received.
The gorilla was in the room the entire time. I was too busy counting bullish passes.
What You See Is All There Is
Think of it like building a house. You’re the architect, and you’ve designed a beautiful home. Strong foundation. Great layout. Perfect for the lot. You’ve spent weeks refining the plans, and you’re proud of them. Now someone asks you to find everything wrong with your own blueprints.
How hard are you really going to look?
This is the WYSIATI problem, a concept from behavioral psychology. “What You See Is All There Is.” Your brain builds the most coherent story it can from the information available, and then it treats that story as the complete story. It doesn’t flag what’s missing. It doesn’t raise its hand and say, “Hey, we haven’t looked at the antitrust outcomes yet.” It just presents the bull case as a finished picture and moves on.
(I’ve been doing this long enough to know that the moment you feel most confident is the moment you’re most vulnerable to exactly this mistake.)
The application is straightforward, and it applies to any investment analysis, not just NVIDIA. After you build your bull case, count your findings. Literally count them. Bullish on one side, bearish on the other. If the ratio exceeds 3:1, you haven’t disproved the thesis. You’ve proven that your process has a blind spot. The thesis might still be correct. But you don’t know yet, because you haven’t done the work on the other side.
Why This Matters for Investors
This is not an NVIDIA-specific lesson. It’s a permanent investing principle. [PERMANENT]
Every investor who has ever lost serious money (and I have, the kind that teaches you more than any book ever could) knows this feeling in retrospect. You look back and the warning signs were there. Not hidden. Not subtle. They were a gorilla beating its chest in the middle of the frame. You just weren’t counting gorillas. You were counting passes.
The bias audit I ran on NVIDIA caught the imbalance before it could cost me anything. After the full Rational Reversal process, the ratio improved from 3:1 to approximately 2:1, still tilted bullish, but with the bear case properly developed and weighed. The conviction survived the stress test, but it survived with caveats, not cleanly. And that distinction matters enormously.
A conviction earned through adversarial challenge is worth something. A conviction that was never challenged is just a story you told yourself.
In my process, I run this bias audit on every single company. The R-step exists for one reason: to try to destroy the thesis I just built. Not because I enjoy it. Because the alternative is discovering the gorilla after I’ve already invested.
One Thing to Watch
The next time you finish building a case for (or against) any investment, stop before you act. Count the findings. How many support your conclusion? How many challenge it? If the ratio is worse than 3:1 in either direction, the analysis isn’t done. It doesn’t matter how strong the evidence feels. Evidence that feels strong without opposition is exactly the kind of evidence that misses the gorilla.
The gorilla is always there. The only question is whether your process is designed to make you look up from the passes.
If you want to see this bias audit process applied to real companies in real time, the OIA Research Lab walks through the full GAMMERS methodology, including the Rational Reversal stress test, on live coverage universes. Learn more at oialabs.com.