The Wall Street Journal covers companies. Pirate Street Journal cover categories.
Each week, we pick three headlines worth paying attention to and break down the category underneath. Live every Tuesday at 7 am PST / 10 am EST. See the news through a different lens. Keep reading in the Deep Dive Reports.
Dear Friend, Subscriber, and Category Pirate,
Here’s what we covered in this episode:
1. Meta is not in the social media business. It is in the comparison business.
The Journal said Meta got hit with an $18 billion fine. The real number is $12.7 billion, paid out over ten years. Meta makes about $200 billion every single year. So this “punishment” costs them about 1% of one year’s money, stretched across a decade. Mouse nuts.
Calling it a fine misses what really happened. Meta bought something.
The other $5.3 billion only gets paid if TikTok and YouTube both agree to limit kids to one hour a day, and both pay $5.3 billion of their own. Meta turned its own punishment into a trap for its rivals. The states asked for $200 billion at the start. They settled in week two of the trial, right before Zuckerberg had to take the stand.
If you judge a company by its product, you will get it wrong. Judge it by how it makes money. Meta started as FaceMash, a website that showed two photos and asked which person looked better. Every Meta product since then is that same game, just bigger.
Facebook asks: is your life better than your friend’s life?
Instagram asks: is your body better than that body?
The feed asks: is your day better than the day on your screen?
Meta makes money when you compare yourself to other people. That is the whole business. This is why the new parental controls are real wins that still leave the machine running.
Two weeks before signing this deal, Zuckerberg published a 6,500-word letter promising everyone a personal super-smart AI that lives inside your health, your calendar, and your relationships. The press covered the settlement. The bigger question is whether you want to hand your AI life to the company that was already keeping score.
2. NVIDIA made $59.7 billion in 90 days. Apple, Walmart, Disney, and 9 other giants combined made less.
NVIDIA made $96.2 billion in three months, up 106% from last year. Data centers alone brought in $89 billion. Their profit was $59.7 billion in ninety days. That is more than the profits of Apple, Walmart, Coca-Cola, P&G, PepsiCo, Disney, McDonald’s, Costco, GM, Nike, Starbucks, and UPS added together. Next quarter, they expect $108 billion. The stock jumped 8.7% after four straight quarters where they beat expectations, and the stock dropped anyway.
History teaches a lesson here. The company that sells the hardware wins the first act of a tech revolution and loses the last one. Jensen has read that same history book.
One layer of the six layer cake is never enough. Smart players buy squares on the bingo card before they need them. Watch where NVIDIA’s cash went:
$6 billion for Poolside, an AI model built in America to answer China’s AI. Now every customer knows Jensen can compete with them if he wants to.
$500 billion in backup money with six Wall Street banks, to help customers who cannot afford the chips buy them anyway.
Half the data center business now comes from smaller companies, growing 100% a year.
The number that matters most is money made per gigawatt of power. The Hopper chip made $18 billion per gigawatt. Blackwell makes $25 billion. Vera Rubin will make $40 billion. Jensen’s stated goal is infinity.
Towns across America are voting to keep data centers out, while the company at the center of them makes more money than anyone ever has. If the backlash had real power, it would show up in these numbers. It does not. Every DoorDash order is now an AI job running in a data center. Jensen, take the job. Be the Prime Minister of AI, because the alternative is letting the Darth Vaders do the talking.
3. Pringles spent $4 million teaching AI to make every chip identical. It already paid off.
The most important AI story in the news this week: The Pringle Chip.
The price was four to five million dollars, for one production line in Poland. It took four years of work with Siemens. Sensors on the line check 200 pieces of information every millisecond, watching more than 200 things at once, from the size of the flour bits to where the potatoes were grown. The results: chips that are 10% better, 13% less wasted food, and the project earned back more than 40% of its cost. Belgium is next, then the US in 2027, across factories that make 705 million pounds of Pringles a year. This is regular, older-style machine learning. There is no generative AI anywhere in it.
Everyone is watching the trillion-dollar AI buildout. The four million dollar one already paid for itself.
Market the problem, not the technology. In a brand new category, customers want creativity. In an old category, customers want the chip to taste exactly like it did thirty years ago. People pay for sameness, and sameness is a chemistry problem:
Budweiser makes sure the beer tastes the same in Asia, Europe, Latin America and the US. August Busch III enforced it personally.
Cadbury pointed high speed cameras at melted chocolate twenty years ago to watch how it flowed, for the same reason.
Domino’s figured out how to stop water from moving out of the dough, so the crust rises. That is food science doing marketing’s job.
Give AI a job that says “match the recipe perfectly, and keep improving until every chip is identical” and it beats humans every time. Nobody working that line in Poland wants to be Lucille Ball at the chocolate belt, stuffing candy in her mouth because the conveyor got too fast. This is what embedded AI looks like when it works. No big keynote speech. Just a chip that tastes right.
3 conversations to have about the news with the Pirate Eddie Bot and Pirate Christopher Bot
The frameworks only pay off when you run them on your own category. Both bots come with the founding tier and jam 24/7. Take these in:
Name the category you actually monetize, not the one on your website. Ask the bots to separate what you sell from what your revenue actually rewards. Meta sells social media and monetizes comparison.
Map your six layers and pick the second square. Ask them where your competitors are already set up to compete with you, and what you would have to buy to be ready. Jensen spent $6 billion on that answer.
Find the place your customers are paying for sameness. Ask where variance is costing you, then ask what the boring machine learning fix is worth. Four million bought a 40% return on one line.
The through-line this week: every one of these three companies is being described by its product and understood by its category, and the two are not the same thing.
Not a founding member yet? You can join here.
What’s coming up on Pirate Street Journal
Tuesday episodes: three topics, thirty minutes, a couple of bongos. Mini-books and Deep Dive Reports every other Friday. Founding Members get the full DDR, everyone else gets a preview. Breaking News reports go to all paying subscribers.
New episodes drop every Tuesday: three topics, thirty minutes, a couple of bongos.
Our Deep Dive Reports go deep on one company or one shift, with the category read and the call attached. Founding Members get each one start to finish the day it drops, and everyone else gets the preview. Read the Deep Dive Reports here.
Arrrrrrr,
Category Pirates 🏴☠️
Eddie Yoon
Christopher Lochhead
P.S. - Every story this week came down to the same move: find the problem before you go shopping for a solution.
The Category Design Academy is where you do that at the category level. You name the problem only you can name, then design the category around it, so you stop competing on terms somebody else set.
The next cohort starts in October. Apply now to save your seat before it sells out.
