Is OpenAI Worth $150B? (September 2024)

Bill Trench had a great question in his recent application for a new investor role at First Round Capital:

Would you invest in OpenAI at a $150B valuation today? Why or why not?

There are a lot of opinions going on about OpenAI, but I've struggled to find much rigor written about the recent investment round in public, so I took a wack at Bill's question.

I'm curious as to how it holds up (or doesn't) in the months and years to come:

Would you invest in OpenAI at a $150B valuation [in September 2024]? Why or why not?

I would invest in OpenAI's $150B valuation in September 2024, because I believe in:

1. the market
2. the team’s ability to win
3. the biggest risk is mitigated
4. that the bet can have the right economic structure

Market

OpenAI has a big market opportunity today, which will likely expand in the future.

Today, its market can be compared to Google / Alphabet which has a $2T market cap [1] and more than half its revenue from search [2]. ChatGPT is already replacing many Google searches for power users like software engineers, and OpenAI’s API is becoming a critical data processing infrastructure for many companies’ offerings, similar to Google’s other revenue sources.

In the future, we can expect the market to expand. More value will be unlocked as we learn how to better apply the technology. We’re only in the early innings of understanding how to best apply today’s fundamental technology. People are just starting to experiment with LLMs in fields ranging from drug discovery to music generation [3]. Further value will be unlocked as the fundamental technology incrementally improves. That leaves a bigger market than Google’s even if we fall short of the market changing AGI breakthrough that both OpenAI and Anthropic CEO’s predict [4].

Ability to Win

OpenAI has the best offering, execution, and team today, despite some concerns about talent retention. 

OpenAI has the best technology (LLM) and generalized application (ChatGPT). They continue to ship the fastest. They’re quickly expanding their services and distribution to enable faster enterprise adoption [5]. And they’ve attracted more capital and talent than any other startups in the space.

The closest competitor is Anthropic. Anthropic is neck and neck with OpenAI in fundamental technology and shipping speed. However, they are a half step behind in their go-to-market infrastructure and organizational maturity [6]. Incumbents like Google, Meta, or Amazon don’t have the focus, agility, or shipping velocity of OpenAI or Anthropic. Other upstarts are too far behind in scale and resources to keep up with the fundamental technology.

While recent executive departures have raised concerns about OpenAI’s ability to retain talent, I believe that these concerns are manageable. The company has survived through similar challenges, including losing the founding team of Anthropic, in its nine-year history. Smart, competent people I know across functions continue to join the company. From one anecdote: a friend on their pre-training team remarked that a recent Stanford CS professor hire “has the brain power of Illya, just not the fame.”

For the purposes of our argument, we’ll take for granted that if OpenAI can continue to execute on the above, then it will eventually figure out a way to make a Google-sized, successful business with working unit economics.

Biggest Risk

OpenAI’s biggest risk, governance, appears to be mitigated.

Last year’s board turmoil put the company into crisis. It distracted the team, drained talent, and put the entire enterprise at risk. The event put into question: (1) the stability of the company, [8] (2) CEO Sam Altman’s leadership, and (3) the moral direction of the company with respect to AI safety. However, these risks have been acceptably addressed:

The new board has stabilized with experienced members who publicly appear to be working effectively alongside Sam. 

Despite public criticism, Sam continues to have trust with the people he leads. 97% of OpenAI employees called him back from his board ouster [9]. In my own, personal experience, Sam is a thoughtful, consistent manager, who looked out for his teaching team over the two years I TA’d for him. Of the three famous entrepreneurs I TA’d classes for at Stanford, Sam was the first to follow through on his promise to return my call for advice on fundraising when I took him up on it nearly a decade later.

With regard to AI safety, I begrudgingly agree with Sam Altman over the old board that AI progress at OpenAI is probably best sped up, not slowed down. I wish we lived in an idealized world where we could slow down AI progress and build the safest AI possible. I also consider myself a citizen of the world (and am half Chinese). However, in a globe that is increasingly polarized, I have more trust in Sam Altman running an American company, than alternatives running a company in a less-aligned nation, leading the cutting edge of technology with major potential ramifications for the world.

The Economic Bet

For the structure of the bet, assume a $20M investment from a $250M fund. This makes sense as a high conviction bet that (1) is not catastrophic in failure, (2) returns the fund in a big outcome, and (3) insures against a longtail outcome:

Failure

A total loss of 8% is painful, but not devastating. The bet size requires high conviction, but there are still 12.5x the investment dollars for other, more diversified bets.

Big Outcome

The main bet is that in the next 10 years OpenAI follows a similar trajectory that Google did in the last 20 years. A $2T Google valuation would return the fund with a ~11x-13x multiple and 27-29% IRR over 10 years depending on subsequent dilution [10].

Google looks like this today:

OpenAI in 2024 looks roughly analogous to Google in 2005:

Given that today’s internet population is 5x [13] the size of 2005’s internet population and significantly more connected today than 20 years ago, it’s plausible that OpenAI can do in the next 10 years what Google did in the last 20:

Longtail Insurance

The kicker for this investment is hedging against the massive success and potential that we might not yet be able to grasp. LLMs feel like magic. The pace of releases from OpenAI are reminiscent of the “Mother of All Demos,” [14] but with a commercially driven leader at the helm. We’re only at the cusp of understanding how best to apply today’s models. What is the potential when this technology is 2x, 5x, or 10x better?

Even if OpenAI and the industry fall short of achieving their vision of AGI, proverbially landing on the moon instead of in the stars could still transform how society operates. A small part of this bet is insurance so that you are participating if that does happen.

Endnotes

[1] ~$2T Alphabet market cap: finance.yahoo.com

[2] ~57% of Alphabet revenue is from search: sankeyart.com

[3] Example of beautiful AI-generated music: suno.com

[4] Intelligence Age: ia.samaltman.com and Machines of Loving Grace: darioamodei.com/machines-of-loving-grace

[5] OpenAI expanding enterprise engagement: openai.com

[6] From LinkedIn reviews, external marketing, and employee anecdotes.

[7] Other risks: nonprofit-to-for-profit restructure, Microsoft’s stake, unexpected breakthroughs, commoditization via open source, and regulation. Anthropic may also be a strong alternative investment.

[8] "Sam Altman Mythmaking" - The Atlantic

[9] "OpenAI reinstates Sam Altman" - NPR

[10] Valuation dilution assumptions: openai.com

[11] Yahoo Finance: GOOG and DemandSage: Google Statistics

[12] Google revenue & search use statistics: statista.com and pewresearch.org

[13] Internet user statistics: ~1B in 2005 vs. ~4.9B today globalpolicy.org;~5.35B current internet users datareportal.com

[14] "The Mother of All Demos" - Wikipedia