In What Universe is Thinking Machines Lab Worth $50 Billion

What Investors Are Buying

Founded in February earlier this year, Thinking Machines Lab is yet another player in the artificial intelligence startup scene. Structured as a Public Benefit Corporation, the company is aiming "to make AI systems more widely understood, customizable, and generally capable". In short, they're a for-profit company looking to make money doing "AI" and then publish annual statements about all the "AI" they've done for you.

Whereas most artificial intelligence startups are little more than wrappers around the ChatGPT API, Thinking Machines Lab is offering something a bit different. What they have brought to the table is Tinker.

What is Tinker?

According to Thinking Machines Lab, Tinker is

a flexible API for fine-tuning language models

In practice, what Thinking Machines Lab offers is some amount of GPU compute, readily available for training modified open-weight models. They manage the infrastructure (though it is unclear if the company actually owns the underlying hardware assets), and expose a convenient API for their target demographic of "researchers and hackers" to run experiments, do innovative things, etc.

Valuation? Soon to be $50,000,000,000.
source: bloomberg

The Apple Falls Not Far From The OpenAI Tree

The CEO of Thinking Machines Lab is none other than former OpenAI CTO Mira Murati. The very same person who, while commenting about the potential loss of creative jobs due to artificial intelligence, had this to say.

Maybe [the creative jobs] shouldn't have been there in the first place.

In starting Thinking Machines Lab, Murati reportedly hired, "a team of about 30 leading researchers and engineers from competitors including Meta, Mistral, and OpenAI".

The company was bootstrapped 10 months ago with a $2 billion seed round led by Andreessen Horowitz (a16z), at a $12 billion valuation. Even back then the company would have been one of the largest startup valuations in history. Participants of the initial funding round include Nvidia, AMD, Cisco, and Jane Street.
source reuters

They already need more money, with rumours placing the next valuation at $50 billion.

Yikes.

A Little Perspective

What else could investors buy for $50 billion? Instead of a fledgling startup company with an unproven product, zero market penetration, no revenue to speak of, and ultra stiff competition in the most over-hyped tech bubble of all time, one could own ...

• Target (TGT) currently worth about $40 billion
• Chipotle (CMG) currently worth about $41 billion
• Delta Airlines (DAL) and Southwest Airlines (LUV) worth about $52 billion combined

Who wouldn't want their own slop bowl food chain with $9 billion left over?

The First Hit Is Free

A month after launch of Tinker, Thinking Machines Lab announced Tinker Grants.

The purpose of Tinker Grants is to

make it as easy as possible for students and scholars to use Tinker

In this privately funded grant program one can apply for one of two types of grant. Teaching grants, worth $250 per student in an academic classroom setting, are designed to enable students in self-directed projects. And there are research grants, which start at $5,000 to support research projects and open source software that use Tinker.

While supporting open-ended research is commendable, this feels like a ploy to get users hooked into using the Tinker ecosystem. A blog post published shortly after is asking for submissions to be featured as community projects.

Some suggestions of what could constitute a community project include

• Re-implementation of a research project using Tinker
• Product prototypes built with Tinker, demoing a model that does something "fresh or delightful"

The blog post notes that any community project submissions should include a writeup and "preferably an open-source release of your code".

Full Speed Ahead

As the artificial intelligence hype train chugs onwards, one can't help but think this is all a massive, massive fucking bubble.

It was only a few years ago that venture capital was doled out in the millions of dollars, maybe tens of millions for an ambitious startup company. In 2023 the average seed stage valuation was around $2 million (with an m), and the average series A valuation was $30-50 million (again with an m).
source growthequity

And now? Now we have Thinking Machines Lab, a company with an API for building custom LLMs on top of existing open source models. And this company is about to claim a valuation of $50 billion.

With a b.