I started the new year fighting a cold, laid up in bed and catching up on news both general and tech-centric. The former was rife with truly awful stories I won’t rehash here, while the latter led me to this story on Daring Fireball about an OpenAI blog post published the Friday between Christmas and New Year’s.
The post, attributed only to “OpenAI,” outlines their Board’s plan to transform the company’s current for-profit arm into a Delaware Public Benefit Corporation (PBC). A PBC structure would allow OAI to, “raise the necessary capital with conventional terms like others in this space.” Said capital, the post goes on to say, would fuel the company’s non-profit arm, “result[ing] in one of the best resourced non-profits in history.”
OpenAI’s announcement wasn’t unexpected. The company’s relatively short history is full of structural changes, board-related news items, and ongoing tension within its unique corporate structure. Much-reported details about the company’s financial status — centering around ever-growing expenses and its relationship with Microsoft — made the post feel less like news and more like a confirmation of what we already knew.
What made the Daring Fireball post interesting to me was this bit from the lede, which followed the sufficiently attention-grabbing headline, “OpenAI’s Board, Paraphrased: ‘To Succeed, All We Need Is Unimaginable Sums of Money’:
From an un-bylined post from OpenAI’s board of directors on Friday:
The hundreds of billions of dollars that major companies are now investing into AI development show what it will really take for OpenAI to continue pursuing the mission. We once again need to raise more capital than we’d imagined. Investors want to back us but, at this scale of capital, need conventional equity and less structural bespokeness.
My take on OpenAI is that both of the following are true:
OpenAI currently offers, by far, the best product experience of any AI chatbot assistant.
There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble.
I won’t offer any analysis or opinion either on that first bullet point (I’m currently preferring Anthropic’s Claude and an ever-changing list of locally-hosted, open source LLMs, myself), or the fact that, as DF put it, “OpenAI’s board [is] now stating “We once again need to raise more capital than we’d imagined” less than three months after raising another $6.6 billion at a valuation of $157 billion.”
Let’s focus, instead, on that second bullet: “There is no technical moat in this field, and so OpenAI is the epicenter of an investment bubble.”
OpenAI’s out-of-the-gate lead on its competition in the world of AI models is now, for all intents and purposes, gone. ChatGPT blew everyone’s doors off when it launched in November of 2022. Twenty-five months later, the Net is flooded with LLMs, chatbots, and other Generative AI models available to most anyone with an internet connection — many of them free to use. That’s not even mentioning the myriad purpose-tuned applications built atop AI models, which leads me to the crux of DF’s argument:
OpenAI is to this decade’s generative-AI revolution what Netscape was to the 1990s’ internet revolution. The revolution is real, but it’s ultimately going to be a commodity technology layer, not the foundation of a defensible proprietary moat. In 1995, investors mistakenly thought investing in Netscape was a good way to bet on the future of the open internet and the World Wide Web in particular. Investing in OpenAI today is a bit like that — generative AI technology has a bright future and is transforming the world, but it’s wishful thinking that the breakthrough client implementation is going to form the basis of a lasting industry titan.
Marc Andreesen and James Clark founded Netscape on the back of the Mosaic browser, which Andreesen had led development of at the National Center for Supercomputing Applications (NCSA). I vividly recall using Mosaic for the first time in the campus computing center during my early/mid-90s undergrad years. It was a sea change from the VAX terminal experience that had defined all of my college Internet use up to that point. I’d never seen anything like it. Techies flocked to Netscape, and the software quickly accounted for more than 90% of all Web browsing.
A decade later, Netscape had decisively lost the browser wars to Microsoft and others, registering only a 1% market share by 2006.
Is DF’s John Gruber right? Will OpenAI’s legacy ultimately be Netscape-esque, remembered for introducing the masses to a transformative technology only to miss out on the real opportunity to build commercial applications atop the “commodity technology layer”? Just as Netscape quickly gained and lost market dominance, and eventually sold to then-industry giant AOL, does OpenAI’s future lie in supplying Microsoft with commodity tech they can infuse into future versions of Office, Teams, and the like?
If not, what’s OAI’s move? Sora caught the public’s imagination with its early video generation demos in February 2024. While not the first Gen AI video model, it quickly became the best known and most talked about, at least in mainstream circles. By the time OAI officially released Sora to users 10 months later, Google, Meta, and a host of smaller competitors were hawking their own video models. Open-source alternatives like Mora had started popping up, as well.
From where I’m sitting, we’re in the early-to-mid stages of Gen AI’s “applications layer” era. AI labs like OpenAI, Meta, and the like are still working on pushing the boundaries of what their models can do. But the current models seem to be capable of a lot more than what users have done with them so far. Big opportunities still exist for developers to design and build useful — and profitable — applications and business models atop the current generation of LLMs. Will investors seek opportunities backing AI app builders in favor of putting billions more behind OpenAI (and, for that matter, their model-building peers)?
I’m expecting 2025 to be quite interesting on this front. What’s your take?