👀 What is going on at OpenAI?

Most of the top players have left OpenAI. CTO Mira Murati announced her departure on Wednesday, alongside VP of Research Barret Zoph and Chief Research Officer Bob McGrew. OpenAI is turning into a for-profit benefit corp and Sam Altman will get equity worth billions for the first time. The truth is, we don’t know what is going on at the organisation. But today I want to share some considerations to help us think about state of the firm.

Sep 28th: I’ve added a note about reporting on OpenAI’s revenue growth.

1. OpenAI is NOT falling apart.

The claim that OpenAI is en route to a corporate calamity is, at best, clickbait. OpenAI is one of the fastest-growing companies in history. It doubled its ARR to $3.4 billion in the first six months of the year. If you’ve ever run a company, you’ll know how difficult it is to get anywhere near this growth rate. It’s got revenue resiliency from corporates and consumers directly. It signed a deal with the US government on AI research and testing. If that’s a train wreck, I’d like to know what a well-run train looks like.

2. OpenAI is becoming part of the national-security complex.

In January, the company changed its terms of service to remove the prohibition on using its products for the purposes of “military and warfare”. That same month OpenAI announced they were working with the Pentagon on open-source security software. We may not like it, but OpenAI is changing. There are many, many talented technologists who would not want to work at Palantir. The same may be true for OpenAI in 2024; Sam, and the board around him, may be willing to break the culture to complete this transition.

3. Sam’s part and the OpenAI dream.

Do employees trust Sam Altman? We don’t know. Perhaps his top executives don’t trust him, don’t like his choices, or just wanted a bigger slice of the pie.

While some successful tech firms, like Google and Microsoft, had stable leadership for years, others were full of drama. Consider Apple and Uber.

Karen Hao in The Atlantic characterises the latest departures as “Altman’s consolidation of power nearing completion.” That makes some sense… OpenAI has essentially abandoned its initial not-for-profit mission and the checks on its unalloyed power that structure offered.

It’s unseemly, of course. I was among many who heard directly from Sam about this novel ownership structure (see our public conversations in 2020 and 2023). Perhaps the stakes are too high for OpenAI to innovate in corporate governance that way. And if not the stakes, then it’s clear the potential payday is big enough to erase those promises from history.

via @GavinSBaker on X

4. No one knows how to run this kind of business, actually.

OpenAI and its ilk are a new kind of business. No one, not even Sam Altman, knows how to run it (yet). The company is building a technology that could impact initially every single one of the billion or so people who work at or near a desk. What other company has forced a reinvention not just of its industry, but also the semiconductor and energy industry? It’s ChatGPT that is a direct line to the revival of nuclear as a real contender in the US energy system (I’ll write more on this in the Sunday newsletter). There is no precedent.

5. Solo albums are more prolific.

The reports of burnout and turf wars are not surprising for a company that’s changed so drastically within two years. But there’s another way to look at this. Key players leaving could be reflective of a vibrant market, a market with more precedent for people to go off and start their own thing. We didn’t see this happening in the dot-com bubble where the cycle time was 5-7 years.

Cycles are much faster now – and there’s excitement around the opportunity to build something of your own at a time when AI use is shaping up. The decision worked out for the employees who left to found Anthropic (the latest rumours have its valuation floating around $40 billion), Ilya Sutskever who raised $1 billion for his new startup. Even Fei-Fei Li, best known for her academic and non-profit work, raised $230 million for her new AI startup (already valued at more than $1 billion).

Notes: The New York Times reported some data on OpenAI’s revenue growth after I wrote this.
Key numbers:

• 350m monthly actives on ChatGPT in June, up from 100m in March.

• $3.7bn in revenue annual revenue of which $2.7bn in consumer and $1bn API based.

• Forecast $11.6bn revenue next year, more than a tripling.

• Losses of $5bn this year; cost of $8.7bn by implication.

• My view: A firm with +200% annualised revenue growth rate should be running losses. This is pretty normal. You invest ahead of revenues.

• OpenAI’s metrics show a relatively higher burn compared to other SAAS companies, but the growth rate is exceptionally high to compensate. For one thing, OpenAI passes the “rule of 40” test.