How do you compete with the everything company?
Or, is OpenAI going to successfully become everything to everyone in AI?
Earlier this week, OpenAI hosted an event that sent chills down the spines of startup founders and VC’s everywhere: Dev Day. In a flash, OpenAI announced a cheaper & faster version of GPT-4 (with a larger context window), the ability for users to create “custom GPT’s” (read: GPT wrappers), their own version of an app store for developers to sell those custom GPT’s, and several new API’s that promise best-in-class performance to develop assistants, text-to-image, and text-to-speech applications.
In what is becoming a bit of a trend, initial market reactions were overwhelmingly positive, with several news outlets and “thought leaders” on Twitter (Does anyone call it X yet?) claiming that OpenAI had just wiped out any chance of success for an entire generation of AI startups, and would perhaps just “win everything”.
Could that possibly be true?
OpenAI is without a doubt, an incredibly fierce competitor. They ship products and updates exceptionally quickly. They are incredibly ambitious, and clearly have aspirations to target multiple markets / parts of the AI stack at the same time (not to mention the fact that they’ve shown a willingness to replicate their customers products, Amazon Basics style). Perhaps most impressively, they have shown an ability to marry market-leading AI technology with excellent products that are actually easy to use, a combination that has been an Achilles Heel for most other AI companies.
One could even make a convincing argument that OpenAI is the most successful “start-up” (if they can still be called that) of all time, having only been founded in 2015 and currently being valued somewhere in the neighborhood of $90B today. Having watched them closely over the last few years, I have both a deep admiration for the society-altering organization that OpenAI has become, as well as a deep fear of what it means to compete against them.
However, if history has taught us anything, it’s that markets rarely end up with a single monopolistic winner (the one notable exception being web search). Even in the social media era, the upstart emergence of TikTok struck back against the argument that Network Effects were insurmountable.
As businesses expand horizontally into multiple product lines and end markets, their ability to be everything to everyone becomes even more difficult. As OpenAI continues on their quest towards world domination, I continue to believe that there are still many legitimate strategies of competing with them, even as a brand new startup.
Sizing up the competition
First things first: just because OpenAI says that they’re going to do something, does not mean that they are automatically going to be able to do it better than anyone else. Earlier this year, OpenAI announced “Plugins” as their new “killer app” for ChatGPT. A little over six months later, and the idea has largely fizzled out, with OpenAI CEO Sam Altman even claiming that plugins “don’t have Product Market Fit” and are probably not coming to the API anytime soon.
At the time of the original plugins announcement, the was similar market hysteria, with some writers going as far as to claim the OpenAI had just developed the “best business model” ever. As plugins have fizzled and largely disappeared from the limelight, it’s clear that this didn’t play out (yes, the new GPT app store concept is clearly a reworked version of plugins and another bite at the apple).
The point being, actually executing on delivering revolutionary products is extremely hard, even for well funded incumbents (let us not forget Apple Maps!). While it’s important to respect OpenAI’s incredible capabilities, it’s also important to not over estimate them.
While I think it would be incredibly challenging to try to compete with OpenAI by building a foundation model / GPT-4 competitor from scratch today, there are many other use cases and layers of the AI stack in which I think startups can win.
Competing with the everything company
If you accept the hypothesis that OpenAI is not infallible across every single business line (as I do), then there are clearly ways to successfully compete with them.
Below I’ve listed a few hypotheses I have on successful methods of counter-positioning, but this list is clearly not exhaustive (as always, I’d love to hear feedback / alternative viewpoints):
Deep vertical focus: The most obvious way to compete with a generalist competitor is to focus on acting as a specialist. Companies that know their customers and users better than anyone else are much more likely to provide more value than any horizontal player can provide
An example here would be the OpenAI-backed Harvey AI, a vertical AI platform tailored to the specific needs of lawyers. Given legal needs / requirements are highly specific, law firms have been gravitating towards tools like Harvey and CaseText instead of the more general ChatGPT (OpenAI’s backing of Harvey likely indicates similar thinking internally)
Once you “win-out” in one vertical, you can begin to expand into adjacent end markets where you can build similar bespoke workflows and datasets (e.g., Hebbia’s expansion from financial services into the government end market)
Proprietary data moats: We have talked about this many times before, but data moats appear to be the most defensible form of moat in the AI era. For startups, getting access to proprietary datasets of your own is a critical form of differentiation against OpenAI, particularly for use cases in which the best model performance is absolutely crucial
This is increasingly proving to be true in the Healthcare space, where the datasets are hard to acquire and require unique security measures to manage, AND the difference between 95% and 99% accuracy is critically important
Focus on a cloud agnostic story: While Microsoft’s multi-billion dollar investment into OpenAI enabled them to plow cash into R&D and model training, it has also tied OpenAI’s offering closely to Microsoft Azure. For organizations where their largest expense item is their annual cloud bill, being locked-in to a single vendor is a huge headwind to integrating closely with OpenAI
AI companies that are compatible with any cloud environment, offer significantly more flexibility (and therefore Cloud bargaining power) for their end customers
Today, many customers are inherently distrusting of Big Tech, particularly when it comes to trusting them with proprietary data. Independent alternatives almost always represent less business / competitive risk
Prioritize Enterprise needs: While OpenAI does have an enterprise offering, the vast majority of their now 100M weekly active users are consumers and individual developers. It has always proven to be incredibly challenging to cater to consumers and enterprises at the same time. As OpenAI’s core business largely caters to individual consumers and developers, startups are more likely to find success tailoring their offerings to Enterprise-specific needs and use cases
Similar to the above point, Enterprise customers are increasingly wary of trusting OpenAI itself with their data. OpenAI has shown a willingness to compete with anyone, anywhere (often without warning). For customers who are wary of OpenAI eventually becoming a competitor, independent startups represent a safer vendor / partner
Leverage the Open Source community: We’ve discussed the many problems with Open Source AI in the past (particularly when it comes to open source models). Open Source software (e.g., Tooling), however, remains a valuable and mighty weapon when it comes to competing with OpenAI’s closed source products. By leveraging an active and thriving open source community, startups can often build much faster than closed source offerings, at a fraction of the cost
It’s worth noting here that OpenAI’s increasing push towards regulation has also generated a significant amount of backlash in the AI community. Last week, AI luminary Yann LeCun suggested that OpenAI was pushing to secure “regulatory capture of the AI industry”, and trumpeted the development of open source AI as an alternative. As Big Tech providers become more associated with monopolistic practices, I suspect more and more users will protest their usage
Are there strategies I missed, or do others believe that OpenAI really will become the “everything company” across the entire world of AI? If so, how big can one company possibly get?
Either way, I’d love to hear feedback in the comments below (unless it’s an infallible argument constructed by the now self-aware and autonomous GPT-5, in which case I think I’d prefer not to know).