What we’re reading: AI, China & the power of brand | Who said what? S2E35

Today’s Who Said What? episode is going to be a little different. I am not going to pick a comment from a business leader or a fund manager and contextualize things for you.

We recently started something called Subtext. The idea behind it is simple — there’s a lot of interesting thinking happening in markets, regulation, energy, technology, and everything adjacent to all of that, and most of it never reaches people who aren’t deep inside those worlds. So we’ve been trying to create a place where you can read extremely high-quality opinions from people who are actually in the thick of it. We have some exciting stuff lined up there, and we will start releasing it next week.

The other thing we’ve been doing on Subtext is something a little more personal. It’s called What We’re Reading, where every weekend, we send a long piece as to what all of us at Markets by Zerodha is reading. I wanted to experiment and see if this is something you like. If it is, let me know in the comments.

In the last edition, all of us went through a bunch of interesting reads, all of which you’ll find links to in the description below. I’m going to walk you through a few of them here today. The ones that I thought were worth talking through out loud. But if you want the full piece, it’s in the description and it’s worth reading in full.

Let me start with mine.

I’ve been reading a piece by Ed Zitron, who runs a newsletter called Where’s Your Ed At? It’s fun, it’s a little provocative, and he regularly goes places that mainstream tech journalism mostly won’t. This piece is called The Beginning of History, and it starts where you wouldn’t expect a piece about the AI bubble to start: the Strait of Hormuz.

The Strait of Hormuz is a 24-mile-wide channel between the UAE and Iran. That’s it — 24 miles at its narrowest point. And roughly 20% of the world’s oil and LNG flows through it every year.

Source

By now of course you know all of this. At the end of February, Iran’s Revolutionary Guard closed it off, telling merchant ships that passage was no longer allowed. Iran can’t legally do this — it’s international waters, governed by UN convention — but merchant captains don’t particularly want to find out how seriously Iran means it. Neither do their insurers. And so the tankers are sitting put.

What that means for oil prices is fairly direct. We saw a 30% overnight spike, with the price of a barrel crossing $100 for the first time since 2022. And oil prices feed into everything: shipping costs, construction materials, medicines, basically anything that has petrochemicals in it or needs to be moved anywhere. Inflation goes up. When inflation goes up, central banks raise interest rates to cool things down.

And here’s where the AI bubble enters the picture. The entire infrastructure of AI — data centres, chip shipments, the VC funds backing startups, the private credit sitting on top of all of it — runs almost entirely on debt. When interest rates rise, even by a point or two, financing the hundreds of billions this bubble requires gets significantly more expensive. Investors in VC funds can suddenly earn better returns with far less risk by parking money in a government bond. The debt spigot, which has been running wide open, starts to tighten.

Then there’s the physical layer. The data centres powering AI are largely gas-powered, and natural gas is up as much as 50%. Flying racks of NVIDIA chips from Taiwan to Texas becomes more expensive. Every contractor building these facilities will pad their fees to account for their own rising costs. The squeeze comes from every direction at once.

All of that is about external pressure — macro forces making the bubble more expensive to sustain. But Zitron’s second argument is different, and in some ways worse. It’s about whether the foundation that’s supposed to justify all this spending was ever real in the first place.

Press reports have been citing Anthropic’s annualised revenue at anywhere from $14 billion to $19 billion — but those figures take a single month of revenue and multiply it by 12. Then a court filing from Anthropic’s own CFO, submitted as part of a suit against the Department of Defense, stated that Anthropic’s lifetime revenue as of March 9th, 2026 is $5 billion. When Zitron adds up all the monthly figures that have been publicly reported, they already exceed $6 billion — which makes the $5 billion lifetime figure, and the $4.5 billion annual figure that’s been widely cited for 2025, very hard to believe.

His broader point is that journalism keeps reaching for the same historical comfort: dot com, crypto, now AI — and in each case building a narrative around the assumption that it’ll work out because things have worked out before. He thinks that’s lazy, and that this era is different enough that the comparison fails. I’m not sure I have a fully settled view on that yet. But the Anthropic numbers alone make this worth reading.

---

Now, Pranav has been reading two things, and I’ll tell you about both.

The first is a Paul Graham essay called The Brand Age. He starts with the Swiss watch industry. After World War Two, the Swiss were the best in the world at making watches. Then the Japanese started closing the gap, and within a generation they’d caught up, and a lot of Swiss manufacturers went bankrupt. The ones that survived did something interesting — they stopped competing as tool-makers and started competing as jewellers. They sold you not a watch, but a brand. And that brand saved them.

Graham’s conclusion isn’t a comfortable one. He argues that brand is often more important than actually making a good product. That good design — sensible, functional, honest design — can actually be bad for branding. That good branding might need to be loud and gaudy, which is the opposite of what anyone with taste would want to make. He’s not celebrating this. He’s just describing it plainly, which is what makes it sting a little.

However, one of our own stories might not entirely agree with Paul Graham. Last year, we narrated a tale of how the company that followed this playbook for natural diamonds to the highest possible extreme is now fighting for bare survival. And that’s because its fundamental business is becoming more commoditized every single day.

The second thing Pranav has been reading is Ruchir Sharma in the FT on China’s growth targets. Most countries discover their GDP growth rate after the fact — the year ends, the data comes in, and you find out how you did. China does it the other way around. They announce a target, and then they hit it. Ruchir’s point is about what it costs to do that, and who ends up paying.

Hitting a target, year after year, requires constant investment whether or not that investment makes economic sense. You build factories, you build capacity, you produce. Over time, this creates a structural glut — too much of everything, produced at a cost lower than anyone else can match. The first version of that reaches you as the cheap Chinese goods that have made life genuinely more convenient. The second version is the hollowing out of industries in countries that simply cannot keep up. That’s bred a deep resentment across much of the world. The targets make China’s problems the world’s problems — and there’s only so long the world will absorb that quietly.

And then there’s what Manie has been reading, which I think is one of the more interesting things in this edition — a long piece by Kevin Xu on China’s open-source software history.

Kevin Xu is someone worth knowing. He led GitHub’s international business strategy, worked at the White House during the Obama years, and runs a newsletter called Interconnected that covers Chinese technology and entrepreneurship. His piece starts from a question that I suspect most people haven’t thought to ask: DeepSeek, MiniMax, and Kimi — the Chinese large language models that have been making headlines — are all open-source. Did that come from nowhere? Or is there a much longer story behind it?

It turns out it’s the latter. China’s open-source community wasn’t created by a government mandate. It grew the same way it did everywhere else — spontaneously from the bottom-up, from people who cared about it. The government only truly committed to it once they realised how strong the movement already was, and how much it could add to China’s economic and technological strength.

The first big inflection point Xu traces is Alibaba. They were entirely dependent on Western providers — Oracle, IBM — for their server infrastructure. As their e-commerce business scaled, those bills kept growing. So Jack Ma brought in someone to fix it: a man named Wang Jian, who had a doctorate in psychology and, notably, did not know how to code. Wang Jian ended up playing the biggest role in helping Alibaba wean itself off that dependency.

What makes the piece stick is that it doesn’t treat open-source as just a software story. It puts it inside a frame of national security and sovereignty — who controls the tools, who can be cut off, and who can’t. Manie says he doesn’t like most of the long essays that get published on Twitter these days, but this one is a gem. I’d agree.

---

Thats it for this episode, do let me know if you liked this and if we should do more of this :)