Is OpenAI about to replace a bunch of startups? It's complicated
OpenAI has to build a go-to-market strategy for both an enterprise and consumer business. And its biggest potential achilles heel could be its lack of focus.
OpenAI’s developer day spent a lot of time trying to ease some of the frustrations and challenges developers had with GPT-4 by reducing its price and increasing rate limits. At the same time, it went much deeper into consumer products, with the launch of its agents-like marketplace that it’s calling GPTs.
That marketplace, along with ChatGPT proper, has basically reinforced OpenAI’s status a company that has both a consumer and enterprise business. And the reality is these are two very different growth problems, which poses an immense challenge for OpenAI. The former is more of a marketing and brand-ish problem, and the latter is, depending on the organization, a sales-ish (and marketing-ish) problem.
OpenAI has the formidable task of building go-to-market operations for both consumer and enterprise markets, as well as whole specific use cases as it tries to break into larger enterprises. And like any startup attempting to do both, it runs risk of doing-a-bit-of-everything-without-excelling-at-a-few by spreading itself so thin across a wide variety of products and types of customers.
That challenge also means that the opening for startups going after various parts of the AI stack—from modalities to infrastructure—could easily still be up for grabs, regardless of OpenAI’s influence. That’s left sources I’ve talked to plenty bullish on companies spanning different parts of the AI stack, from text-to-speech startups like ElevenLabs to training and inferencing startup Together AI. (The latter of which just announced an inferencing engine today.)
With the series of announcements OpenAI made at dev day, it feels like many were quick to point out how many startups and potential products OpenAI squashed with its new tools and APIs. But the reality is complicated. Here are a few examples:
Vector databases are doomed because OpenAI abstracts them out, never mind that enterprises still are skittish about using an API in the first place).
ElevenLabs and text-to-speech startups are doomed because OpenAI launched a cheap text-to-speech, never mind ElevenLabs’ voice cloning and exceptional latency and big head start).
Image-generation startups like Midjourney are doomed because of its DALL-E integration, never mind that Midjourney has created an absurdly large community and, from what I last heard in May, has blown past $110 million in annual recurring revenue. (The Information in September also indicated Midjourney should break $200 million in revenue this year).
You don’t need an agents/chaining tool like LangChain or LlamaIndex thanks to its assistants API, never mind the very large communities behind those tools.
“Replacing” every one of these startups all represent unique sales and marketing problems. Pitching a text-to-speech tool for a company that powers call centers is very different than a company that wants to search across legal documents. It’s one thing for large, sprawling companies to pull this off—but OpenAI is still less than ten years old.
OpenAI has to build a moat somehow, and for the time being it has no unbreakable moats to speak of. Its main enterprise advantage is that it’s the highest-performing products and its ability to crush the competition on pricing and ease of use. And its main consumer advantage is it’s largely the best user-accessible multi-modal tool.
OpenAI clearly wants to continue to grow in both areas. And, needless to say, this is a really challenging and complex go-to-market problem—and the team handling it has its work cut out for it.
A quick look at some of OpenAI’s go-to-market profiles
For the uninitiated in the mess that’s marketing and sales, go-to-market is a catch-all term that basically encompasses a bunch of teams that are around getting these into the hands of users. That generally includes account managers, sales development, business development, marketing operations, and sales engineers.
The easiest place to get started reviewing OpenAI’s cluster is LinkedIn, which is actually turning into a very vibrant (and apparently Cool) network following the Grand Social Media Blergh of 2023. I skimmed through a few dozen profiles to get a quick sense of some of what’s up with the team that’s now tasked with the complex problem of selling Artificial Intelligence.
(This is a small-ish cut of the search which spread out to twenty-something pages before I got throttled. LinkedIn wants to charge roughy $540 a year for a premium membership for a bootstrapped startup—though they’d love to get that up to around $1,000 with sales navigator! Maybe when I’m super successful and have a corporate card for something beyond subscriptions at SemiAnalysis and Platformer.)