'How High Can It Go?' OpenAI, SK hynix & AI M&A Prices. ARD #129

Today’s theme: ‘How high can it go?’

That question focuses on three events today.

Three companies to discuss OpenAI, nudging a billion weekly users. SK hynix, the leading global memory company, which has had its own amazing ride this year. And Stripe/OpenRouter on topic AI M&A where the prices are provoking the same questions on valuation.

This is the AI Tech Wave at the point where the fundamentals and the emotions stop keeping step. Three events, each with my Take, then my Overall Take. Plus a Gadget AI on delivery drones, and two questions. Let’s get started.

(1) OpenAI Nudges a Billion Weekly Users

MP TAKE: They were expected to cross the billion weekly mark back in December, and they’re only now approaching it. And by the way, Google claimed 950 million monthly users just a couple of days ago. So both are racing on slightly different metrics. The weekly metric is a little harder than the monthly one, but roughly in the ballpark.

And both have a lot of headroom, given there are eight billion people on the planet and at least half of them have smartphones.

ChatGPT is of course the AI ‘Kleenex’ brand, the fastest growing product in the history of technology, as some say. It’s been barely three plus years. It is the Kleenex of what people think of when they think of AI. So the momentum is impressive even if a tad late.

But what I wanted to talk about is that it’s not just that pure metric reaching a certain height. They’ve also been re-shifting priorities for the last few months, from the consumer side to the enterprise side. To catch up with Anthropic over there running ahead with Claude Code, Claude CoWork, and everything they’re building on top of that. And on a higher annualized revenue run rate than OpenAI.

OpenAI claims about 50% of their business is already enterprise. And remember the billion number. Only about five percent are paying ChatGPT subscribers. That’s about 50 million.

The metric by itself makes great headlines. What investors care about is how high the revenues go, and how much higher relative to the capex the company is spending, which is in the hundreds of billions of dollars. Those are the relative highs that matter. The pendulum swings that matter.

And that’s what’s going to be focused on, especially as the company races toward its IPO. Potentially not this year, unlike Anthropic, which is focused on a trillion dollar plus market cap. But next year. We’ll see.

They are almost at this holy grail number. Close to a billion. They didn’t say they crossed it. So we’ll give them that. Remember, a billion here, a billion there, eventually it adds up to something real and sustainable. Potentially.

These metrics are impressive by themselves. But it takes decades for these companies to really become what I call Internet Treasures. In this case, AI Treasure companies.

What is that? Amazon is an Internet Treasure. Google is an Internet Treasure. Companies that have stood the test of time across decades. Served billions of customers constantly, daily, with extraordinary focus on daily habits that billions of people have developed and rely on. And they’ve made a lot of money across a range of businesses, whether subscriptions, transactions or advertising.

That’s what I have always called, for over 30 years, Internet Treasures or Technology Treasures. PC Technology Treasures would of course be companies like Microsoft and Dell. Those have stood even greater tests of time. Through massive ups and downs. Through thick and thin. Through searing changes in technologies.

That’s what’s important in terms of the long, long-term metrics. A billion here is something to note in terms of ‘how high?’ And commend. But I wanted to highlight this business in the broader context of those metrics.

(2) AI Driven Memory Stock Volatility Continues

MP TAKE: We’ve talked a lot about this, and I talked a lot about it yesterday.

The three big memory companies are an oligopoly on a global basis. SK hynix and Samsung in South Korea have each been up several hundred percent in the last couple of years, are about 50% of the Korean market, and are each over a trillion dollar market cap. Then you have Micron over here. Other companies like Sandisk have also had huge runs. And Sandisk’s manufacturing partner in Japan, Kioxia, is also a big one. I’ve written about all of them, and they’re in the source links below.

SK hynix reported numbers, and that’s what I wanted to talk about. This past week we had Google report. Most of the big tech companies report starting literally today and through the week. Meta, Microsoft, Amazon and Apple. SpaceXAI and Nvidia later in coming days. So we should talk about SK hynix, because they’re the big daddy of the global AI HBM (high bandwidth memory) companies.

Their revenue is up over 250%. Operating profit over 550%. Gross margins 83%. These are software-like margins.

And the stock was down about nine, ten percent.

Now that typically happens. A sell on the news kind of thing. And the media will say they didn’t beat the expectations. The expectations game is a tactical short-term game that people play with Mr. Market. I was a former analyst, head of internet research at Goldman in the 1990s, so I know that tune very well. So you almost always have a little bit of a sell on the news, and volatility around this.

In this case it matters, because tens of millions of people in South Korea have been trading in and around ETFs, leveraged ones, on SK hynix and on Samsung in Korea. So they’ve had a lot of margin calls up and down. And with these kinds of moves, a lot of them are being wiped out. Over a million people had margin calls in recent days.

That’s the danger when you’re playing momentum trading games on a short-term basis, on these wild pendulum swings, trying to guess how high these markets can go. And today’s technologies and platforms like Kalshi and Polymarket basically encourage us to do more of the short term betting than the long term investments.

The core story on SK hynix is that their demand picture looks very good and they’re building the supply. They’re going to be investing a lot of money on additional capex. Which by the way also causes a reflex action on sell-side and buy-side professional institutional analysts, because they’ve had decades of experience with semiconductor companies especially overinvesting cyclically, and as a result having shortfalls when the cycles turn.

So there’s a lot more volatility at the bottom of the tech stack, where semiconductors live, versus the application companies on the top. That’s a dynamic to watch here.

And by the way, in the context of the other thing I’ve talked a lot about, which is ‘RAMageddon’. A phenomenon where memory supplies, both RAM and storage, are so tight that we’re going to see price increases continue through the end of the decade. These companies are basically sold out of their supply.

So regardless of the volatility on the quarter numbers, you’ve got to keep that in mind if you’re a longer term follower of Mr. Market.

(3) Another Potential High Price AI Acquisition Ahead

MP TAKE: There have been rumblings that Stripe, one of the most successful private payment companies, worth over $150 billion in the private markets, is looking at another deal. A few weeks ago they said they may want to buy PayPal. They made a bid and that doesn’t seem to be going anywhere.

But they’ve made another consideration that’s leaked. They’re looking at buying a hot AI company called OpenRouter. A company with about $140-ish million in revenues. And the valuation talk is about ten billion. So that would be a little over seventy times revenues.

That compares higher to SpaceX, with Elon buying Cursor, the AI coding company he bought right before the SpaceX IPO, for about 22, 23 times revenues.

Again, former analyst here. The multiples on revenue, on earnings, all these things are important. But what drives these things are strategic aspirations. Things you cannot easily measure and broadly communicate. Elements that are not in the numbers, not in the metrics. With the possibilities of combining A with B and creating ABC.

Most mergers actually don’t work out, for a whole host of reasons. There are qualitative issues and quantitative issues, and that’s what you’ve got to keep in mind.

But the numbers from a valuation point of view, when you take snapshots and the media talks about it, they’re going higher. And they can go higher, given that we’re still in the early days of the AI Tech Wave. Just year four. So much more to build.

And every company, incumbent or startup, knows from the lessons of the prior tech waves that if you consistently build ahead of your competition, and build on top of your own products and services, and build new products and services for the customers you serve, you can build a lot of value very quickly. Beyond just the momentum game of announcing acquisitions for acquisition’s sake.

Stripe is a serious, pragmatic company, very prone to thoughtful execution. Highly technical. Some of the co-founders of OpenAI came from Stripe.

So I would have higher aspirations for them in terms of this or any other acquisition they consider. But each one is a bespoke item and you have to analyze it in thorough detail. (Disclosure: Am an early shareholder).

We’ll come back to more on this later. But I wanted to use it as an example of how high it can go in terms of the M&A environment for AI.

MP OVERALL TAKE

We are still very much in the early days of the AI Tech Wave, and it’s a momentum-driven wave.

There’s a lot of bullish sentiment. A lot of fear and greed. A lot of fear of missing out. Whether it’s investors chasing things, or companies saying if we don’t do this now, this will not be as cheap tomorrow as it is cheap on a relative basis. And remember, inexpensive and expensive are relative terms for the most part.

We’ve seen it here in the context of single flagship companies like OpenAI. Flagship memory companies like SK hynix. And flagship companies on the M&A side, like SpaceX, the biggest IPO ever, and the biggest acquisition in technology ever with their AI coding acquisition, $60 billion for Cursor. And Stripe, one of the biggest, most successful private companies in tech history, still private over a decade and now looking to really lean into the AI Tech Wave.

All of this is important to note and understand in the broader context.

And remember, the pendulum always can swing wider. Not based as much on the fundamentals, more on the momentum. That’s the distinction one needs to figure out how to always make.

Gadget AI. Drone Deliveries May Actually Be Here Soon

MP TAKE: Here I’m going to talk about a weird kind of gadget. Drones that make deliveries. Food and goods.

This is something Amazon was really into, has been into over the last few years. Invested a lot in drones and drone deliveries, because they’re obviously a delivery business. They created one of the biggest infrastructure systems on planet Earth. Bigger than FedEx, bigger than UPS, USPS and all the other international delivery companies.

Well, flying drones, flying robots, is the next thing people have been focused on. Amazon has tempered its delivery ambitions on the drone side, but they’re still picking away at it.

A new company has jumped in. DoorDash, one of the biggest food delivery companies, is building its own delivery drones. And they just got a nod from the FAA, which is why I wanted to talk about it.

And it’s interesting. There’s a lot of detail in terms of what they’re finding. They’re finding this is not going to replace all their deliveries. Remember, the majority of their deliveries are in urban environments. This is for the rural places. The places where it takes 15, 20, 30 minutes each way to make a delivery. That’s where these things could potentially be helpful.

It’s early days technology. It’s very cool to see these things fly and take off. Plenty of videos on TikTok and YouTube of drone deliveries in Shanghai and China, because they love to show off these things, and they make a lot more of them than we do.

But it’s good to see the US companies in the action. It’s still a sideshow, but interesting nevertheless. And they do go pretty high and come back down with our goods and services. Hopefully safely and reliably.

Q1. Is Michael impressed with this mode of delivery?

Answer: Look, I’m a sucker for anything technology does that’s cool, and this is definitely one of those cool things. In a similar way, I’m very impressed with robot drones in hotels in China, where they literally make a delivery to your door in your hotel room. So drones are another version of that.

But recognize these are still early days and the technology has a long way to go. So it’s a novelty technology service, and I do like that.

Q2. What is the one area where the idea needs to be improved?

Answer: My biggest complaint with food delivery through DoorDash or Uber Eats or whatever is that every once in a while the order is wrong, and the delivery person is gone by then. You basically have to make do with whatever was delivered, and complain, and they’ll give you a refund. These companies have gotten very good at those things.

But with a drone, maybe I can send the order back and get a replacement order very fast, without feeling guilty that I’m making a human driver go all the way back and forth. So that would be an improvement in terms of being able to use drones in a different way, to correct delivery errors.

We’ll see, as and when these things become more mainstream. But they’re fun to talk about, to watch and discuss.

WRAP

Today’s AI-RTZ #1162, on Patents for AI and its agents getting going. Over a hundred thousand AI patents so far, now that organizations are counting them on AI agents. And it’s interesting that the OpenAIs and the Anthropics don’t have as many patents. An interesting ranking, with a lot of interesting implications, and early days. If you have an interest in these kinds of metrics, do take a look.

AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.

Tomorrow, ARD 130 and AI-RTZ #1163.

Thanks for joining us today, AI Curious Folk. Stay tuned.

— MP

✂️

Full Source Reading

For the broader context, see the canonical sources for ARD 129, in today’s narrative order:

Event 1. OpenAI Nudges a Billion Weekly Users

• The Information, OpenAI’s ChatGPT nears 1 billion weekly active users seven months after target

• AI-RTZ #1037, OpenAI trims AI Applications

• AI-RTZ #1125, OpenAI and Anthropic’s tight IPO clock

• AI-RTZ #1079, Anthropic and OpenAI, half of $2 trillion in cloud company backlogs

• AI-RTZ, Chatbot monetization is next race, from March 2024

• AI-RTZ #739, To a new AI powered Internet

Event 2. AI Driven Memory Stock Volatility Continues

• CNBC, SK hynix shares tank as exponential earnings growth fails to satisfy AI charged expectations

• AI-RTZ #557, SK hynix ready for its AI Close-up

• AI-RTZ #1145, ‘RAMageddon’ really here to stay

• ARD #128, South Korea’s leveraged chip trade

• ARD #104, Semiconductors remain a key AI gating factor

• ARD #108, Push Comes to Shove on the Global Memory Crisis

• AI-RTZ #1030, Global memory shortages of 20%+ through 2030+

Event 3. Another Potential High Price AI Acquisition Ahead

• The Information, OpenRouter financials suggest steep price for possible acquirer Stripe

• AI-RTZ #1066, SpaceX/xAI IPO filing outlines Elon’s boundless AI Ambitions

• ARD #62, Making ‘Hay While the Sun Shines’. AI M&A and Funding in Mega-AI IPO Year

• AI-RTZ #806, The ‘Frenemies’ nature of the AI Coding space

Gadget AI. Drone Deliveries

• Bloomberg, DoorDash is building its own Delivery Drones, gains FAA nod for commercial service

• AI-RTZ #669, Re-aligning Incentives in an AI Agent driven world

• AI-RTZ #348, On Land, Sea, and Air

Standing reference. The AI Tech Wave and the AI Tech Stack.

Clips from today

Clip 1. Stripe’s $10B Bid for OpenRouter. 70x Revenue

They’re looking at buying a hot AI company called OpenRouter, with about $140-ish million in revenues. And the valuation talk is about ten billion. So that would be a little over seventy times revenues.

MP Take: That compares higher to SpaceX, with Elon buying Cursor, the AI coding company he bought right before the SpaceX IPO, for about 22, 23 times revenues. Again, former analyst here. The multiples on revenue, on earnings, all these things are important. But what drives these things are strategic aspirations. Things you cannot measure, that are not in the numbers, not in the metrics. With the possibilities of combining A with B and creating ABC. Most mergers actually don’t work out, for a whole host of reasons.

Clip 2. The Pendulum Swings on Momentum, Not Fundamentals

We are still very much in the early days of the AI Tech Wave, and it’s a momentum-driven wave.

MP Take: There’s a lot of bullish sentiment. A lot of fear and greed. A lot of fear of missing out. Whether it’s investors chasing things, or companies saying if we don’t do this now, this will not be as cheap tomorrow as it is cheap on a relative basis. And remember, inexpensive and expensive are relative terms for the most part. Remember the pendulum always can swing wider. Not based as much on the fundamentals, more on the momentum. That’s the distinction one needs to figure out how to always make.

Clip 3. Why Delivery Drones Are a Rural Story, Not an Urban One

They’re finding this is not going to replace all their deliveries. Remember, the majority of their deliveries are in urban environments.

MP Take: This is for the rural places. The places where it takes 15, 20, 30 minutes each way to make a delivery. That’s where these things could potentially be helpful. It’s early days technology. It’s very cool to see these things fly and take off. Plenty of videos on TikTok and YouTube of drone deliveries in Shanghai and China, because they love to show off these things, and they make a lot more of them than we do. But it’s good to see the US companies in the action. It’s still a sideshow, but interesting nevertheless.

Clip 4. DoorDash Drones Get the FAA Nod. Cool, But Still a Sideshow

Am I impressed with this mode of delivery? I’m a sucker for anything technology does that’s cool, and this is definitely one of those cool things.

MP Take: In a similar way, I’m very impressed with robot drones in hotels in China, where they literally make a delivery to your door in your hotel room. So drones are another version of that. But recognize these are still early days and the technology has a long way to go. My biggest complaint with food delivery is that every once in a while the order is wrong, and the delivery person is gone by then. With a drone, maybe I can send the order back and get a replacement very fast, without feeling guilty that I’m making a human driver go all the way back and forth.

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text, a deeper written take on one idea, published by at least 5 AM EST. Today: post #1162.

AI Ramblings Daily is the afternoon video + podcast, my ad hoc takes and perspective on the day’s AI issues and news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #129.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)