Content About Stocktwits
No Image Available

‘Wheels within Wheels’ in AI. Stripe, Nvidia & Unitree. ARD #145

Author
Michael Parekh AI: Reset to Zero
58 minutes ago
9 MIN READ

‘Wheels within Wheels.’ The old phrase, and today it is the wiring diagram for AI’s biggest deals: a payments giant buying the tollbooth for AI tokens, Nvidia funding the data supplier feeding its own open models, and China’s hottest IPO riding a robot data flywheel most investors have never looked inside.

Today’s ARD is about those wheels, in this AI Tech Wave: each deal looks like an open and shut case from the outside, but there is a smaller wheel spinning inside every one. My takes below, as discussed on the show.

(1) Stripe Buys OpenRouter, the AI Token Router

Stripe is paying a reported $7.5 billion for OpenRouter: a three-year-old company with about 90 employees and three founders with big prior successes in crypto, valued at just $1.3 billion as recently as May. OpenRouter routes developer requests across 400+ AI models from more than 80 providers, open and closed, picking the optimal model per request on task complexity, price and speed, and letting developers serve up and manage their tokens. Very different from Hugging Face, the platform where developers find and download models: on OpenRouter, you run them.

That is the attraction to Stripe, whose core business is payment services for millions of developers: close to $2 trillion in transactions run on its infrastructure, and OpenAI and Anthropic run their billing through Stripe as their revenues scale into the tens of billions. Patrick Collison told investors this year marked ‘the beginning of the singularity’, and Stripe is staying private for it.

My take: developers managing token payments out and in is the same worry as managing cash payments out and in, which is Stripe’s core business. There is a lot of synergy on paper, and I am generally skeptical about big M&A transactions. But the technical chops of the Collison brothers are real; I have known and tracked Stripe for a very long time, and full disclosure, I am an early investor. Their big bet on stablecoins, another way to do micropayments on the internet, points the same direction. Early days, a lot of wood to chop, but the wheels within wheels are here.

Sources:

For longtime readers:

(2) Nvidia Funds Its AI Data Supplier Mercor

Nvidia, already the AI ‘Kingmaker’ investing in neoclouds like CoreWeave and Nebius and open-model companies like Thinking Machines and Reflection, is now backing data labeling leader Mercor as part of a round at a $20 billion valuation, per The Information, double its October mark. The backdrop: Scale AI, the prior number one founded by Alexandr Wang, saw Meta take a 49% stake at a roughly $14 billion valuation and hire Wang away. Mercor became the new number one, with gross revenue expected to reach $2 billion this year.

My take: data labeling, data organizing and synthetic data services are the pieces that make this tech wave different from every previous one, which typically ran on five boxes in the stack. With AI, data is an ever-running flywheel needed to train the models and then run inference on them: Data is the critical scaling element after Compute and the Frontier models, open and closed. For Nvidia, the wheels within wheels are strategic: stay close to the data piece of the puzzle for its open source ambitions around its Nemotron models, and the open-model coalition it is assembling. Geopolitical, business, strategic, financial and definitely technical wheels, all turning at once. And note the echo of item one: both Stripe and Nvidia are catering to the developers building with these models.

Sources:

For longtime readers:

(3) The Wheels Behind Unitree’s Spectacular IPO

China’s Unitree Robotics raised $904 million in Shanghai and closed its debut up 460%, after spiking as much as 629% intraday: a roughly $50 billion market value for mainland China’s first listed humanoid robot maker, at about 36 times sales. This weeks after memory maker CXMT’s ~450% debut surpassed Tencent, the venerable company behind the WeChat super-app, as China’s largest market cap. China has a robotics ecosystem of nearly 370 startups formed in two years and a manufacturing base the US lacks, one reason I have been more measured about the humanoid ambitions of companies like Tesla.

The core piece for robots is data, again, tying back to the Mercor story. Data is relatively scarce in robotics: unlike the LLMs, where the entire internet was the training source, robots in the physical world with ‘world models’ need constructed synthetic or real-world data. The FT details the complex ecosystem in China of state- and province-backed training centers, over 90 by June, that buy the robots, generate teleoperated training data, and sell that data back to the robot makers: circular deals of their sort. And per Reuters, the original quadruped breakthroughs came out of US Defense Department funded research, openly published, then commercialized at scale by Unitree, whose founder came out of drone maker DJI. Defense applications are of deep interest to both governments.

My take: investors in both China and the US are leaning into humanoid robots, and Unitree is getting the white hot IPO response. But these are not yet viable mainstream commercial products, humanoid or quadruped, and the data machinery underneath is the ecosystem to understand if you are keen on robotics globally. Wheels within wheels indeed, even for robots with no wheels.

Sources:

For longtime readers:

My overall take: early in this AI tech wave, there are lots of strategic and tactical drivers, both financial and technical, that are greasing the wheels for the deals we are reading about, and a lot of layers of participants, especially the developers around the world, who take all of these components and create new products and services like Lego blocks around them. Understanding that is the story, away from just ‘Stripe paid billions for a company barely three years old’ or ‘China’s robot company had a spectacular IPO.’ Seldom have we seen this many layers of opportunity in the fourth year of a major tech wave.

Gadget AI: AI Smart Glasses’ Trust Problem

The Verge details the abuse of AI smart glasses in the workplace, often by young content creators farming clips for their social media presence: retail workers pranked on camera, staff filmed by customers up to ten times an hour, comedians recorded mid-set without consent. Meta is the current leader with millions of pairs sold, and Apple, Google, Amazon, Snap, and possibly OpenAI with Jony Ive, are all coming. Meta is now moving to disable recording if the indicator light is tampered with, and regulators are taking keen interest in the category’s abuses.

My take: the privacy and trust issue with AI Smart Glasses that has ailed the category since ‘Google Glass’ of over a decade ago still remains a skidding wheel, despite the billions being invested. The societal ‘creepiness’ issue pervades both consumer and enterprise applications. These are early growing pains of the segment, and a very different hardware platform opportunity relative to PCs and smartphones in their early years.

Sources:

For longtime readers:

Q&A

Q1: What is my most used feature of the AI Smart Glasses? ANSWER: I still just use them as Bluetooth headphones. They are kind of cool, open-ear, nothing to stick in your ear, and they work with iPhones and Android alike. I almost never use the cameras.

Q2: What is my least used feature? ANSWER: The cameras, for pictures and videos, and ‘Meta AI’ for questions, in that order. Other form factors, like the sensor-based AirPods we discussed yesterday, may sit easier with society than cameras on faces.

Today’s companion AI-RTZ, #1184, is on China’s AI momentum at Alibaba, in Qwen’s downloads and those recent white-hot IPO valuations: AI: China’s AI Momentum, in Downloads & Valuations (part 1). AI-RTZ #1184

Full Source Reading

Stripe and OpenRouter

Nvidia and Mercor

Unitree and China robotics

Gadget AI: smart glasses

Clips from today

Stripe Buys the AI Token Router

Stripe’s Bet on the Token Economy

Nvidia Funds Its AI Data Supplier

AI Smart Glasses’ Trust Problem

An impact of the AI Tech Wave worth tracking. Stay tuned.

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





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

BROWSE ALL ARTICLES