OpenAI’s Turn

… at the Amazon Flywheel

Yesterday, I outlined how Jeff Bezos harnessed some sound strategy advice at the depths of Amazon’s history in 2001, and adopted the now famous Amazon Flywheel to drive its businesses for the next two plus decades.

And it paid off in spades:

“Then adapted it to several new businesses, and collectively turned all of that into the Four key businesses that drive Amazon’s $1.2 trillion plus market cap and business today.

Those businesses of course are the Amazon Online Store, Amazon Prime, Amazon 3rd Party Retail, and of course, AWS, their world leading Cloud services business.“

I then underlined the reason to go down memory lane:

“The reason I’m going through all this history, is that the latest AI technologies make it possible to build another set of world changing Flywheel loops.

So many are going to have a shot at building both what I call “AI Native” and “AI Add-On businesses, and their own Flywheel Loops.

Chief among them of course is OpenAI, the ‘overnight success’ story behind ChatGPT. Working in Herculean ways to dent the universe with AI. In as good a way as possible.

And Founder/CEO Sam Altman has his potential Amazon Flywheel Loop to build.

I’ve taken the liberty of sketching one out in the chart below.”

Today, I’d like to go through the above chart in more detail.

But some differences before starting out:

1. OpenAI, has a corporate structure and deal with Microsoft, that sets it on a different footing vs. Amazon in 2001. OpenAI is an ‘AI Native’ company with a 49% partnership in its commercial business, with ‘AI Add-on’ company, Microsoft.
2. World now knows what Amazon did with Flywheel loops. Competitors and investors now understand the exponential compounding possibilities, and have high expectations.
3. OpenAI, and the AI Industry, face far greater Regulatory headwinds and customer fears than Amazon and e-commerce did in their day.

With that out of the way, we can start with no.1 for both. Bezos at Amazon was famous for his singular Customer focus from the beginning.

“The most important single thing is to focus obsessively on the customer. Our goal is to be earth’s most customer-centric company.”

So he started the Flywheel there. As Author Jim Collins, who gave the Flywheel inspiration to Bezos and team, put it:

“Let’s diagram how our flywheel works.” If we can create the compounding momentum of our flywheel … so that flywheel essentially was, we offer lower prices on more stuff. If we do that — now notice there’s an inevitable momentum and logic — if we do that we’re going to increase customer visits. If we do that, we can’t help but get more third-party sellers. And if we do that, then we can’t help but expand the store and extend distribution. If we do that, we’re going to raise revenues per fixed costs, and if we do that, we can lower prices on more stuff, and around it goes. And if you stand back and you think about the Amazon story, what is it? It’s that, for a big chunk of it, it’s that flywheel turning and building and compounding, momenting, extending, all the artificial intelligence, machine learning, technology accelerators.

So Steps 2 through 6 are all driven by Step 1, focusing on the customer experience. It came through in how Amazon from the beginning in 1995 innovated around that, with its literally patented ‘1-Click Buy’ button being a key ‘game-changer’ early example.

OpenAI had started similarly with its startling customer experience, both with it’s innovations around ChatGPT, that saw the most rapid ramp to a 100 million plus users in the shortest amount of time.

And it continues to drive the customer-facing innovations directly, and indirectly, with Microsoft, that are driving Steps 2 through 6 for them.

The Data and Models Step 4, is a key differentiator and advantage for the company with GPT4 and beyond, and keeps them in the lead against the likes of Google, Meta, Amazon, and a plethora of private competitors. There’s a lot of innovation to come here, and I’ll have a lot more to say here in future posts.

Steps 4 through 6 are also critical to bringing down the aggregate cost of ‘Compute’ down for the company at scale, over time. Much more to discuss here too in later posts.

And of course Step 5 points to some of the key revenue sources at scale, transactions fees to the company via their APIs, and other software on-Ramps like apps, plugins, and functions.

The company already has a reportedly 9 figure business with subscriptions for its ChatGPT Plus offering. And the third leg of the online revenue stool, Advertising, is likely not far behind, with Microsoft Bing as a partner.

A critical piece for them and competitors is Step 6, the ‘Reinforcement training and learning loops, which are based on user sourced queries into the system. These loops are the key ingredient to making LLM AI models more reliable and far more useful over time. And as we will discuss in future posts, a critical differentiator vs past major Tech waves like the PC and the Internet. Put a Pin here. We’ve barely scratched the surface of what’s possible. And a lot more to say here in future posts.

OpenAI so far is relentless in its innovations in all of the six steps, and continues to positively surprise their end customers. And it has less than 500 employees thus far in 2023. Microsoft has over 220,000.

Importantly, they’re delighting Developers, who use their software infrastructure to create new uses for the Foundation LLM AI models (Steps 3 to 6). Note how giddy this Developer is with ‘Functions’, OpenAI’s latest feature for its Foundation LLM AI MODELS. Oh, and as that article shows, they’re also passing on lower costs to developers, increasing their Flywheel loops and momentum.

OpenAI is well on its way to running its Flywheel hard. And it has the potential to keep them ahead of competitors like Google, Meta, and many others who are spinning their own. Even their Partner Microsoft is hard at work building and spinning their version.

The AI industry innovations have just begun. The Flywheels have a lot of spinning to do and there are steep challenges ahead. But OpenAI has a strong start indeed.

(You can also subscribe to my Newsletter “AI: A Reset to Zero” for free on Substack for more content like this.)