OpenAI is about to clear a major hurdle, freeing the startup to pursue an ambitious and expensive “full stack” strategy that could catapult it into the Big Tech big leagues or spiral the AI pioneer into a financial ball of flames.
The startup said late Thursday it signed a memorandum of understanding to end a contractual fight with Microsoft. This will create a new corporate structure that helps OpenAI issue traditional equity, making it easier to raise the mountain of cash needed to pursue its dreams.
“Beyond boring corporate governance things, let’s talk a little bit about the business,” OpenAI CFO Sarah Friar said at the Goldman Sachs tech conference this week, after explaining the startup’s complex restructuring. “In terms of where we’re going, think of OpenAI in terms of becoming very much a full stack company.”
By full stack, Friar means a tech company that owns and operates every part of its business. Indeed, the startup is already working on its own solutions for almost every ingredient needed for AI success. That includes designing its own chips and building its own data centers, all the way up the stack to a suite of AI-powered software applications beyond ChatGPT.
“The company, as you look forward, will become very much a full stack company, with a lot of places that we can build both moats and really interesting business models,” Friar explained.
Mike Segar/REUTERS
She didn’t say it, but the target is most likely Google. The search giant has spent 25 years developing the building blocks it needs to win in AI. That includes energy assets, specialized in-house chips, massive data centers, leading AI models, talent, developer platforms, wide distribution, and billions of consumer and enterprise touchpoints.
OpenAI is a long way from Google’s dominance, but the startup has several strategy beachheads from which to expand — both up and down the stack.
OpenAI did not respond to a request for comment.
Energy
At the bottom of the AI tech stack, there’s energy. GPUs, AI servers, and data centers consume unprecedented amounts of power. OpenAI doesn’t have a clear answer here yet; however, CEO Sam Altman has personally backed several energy startups, including Helion, which is trying to use nuclear fusion to generate electricity.
Chips
Then, there are AI chips that are essential for training and running AI models. OpenAI hired Richard Ho in late 2023 to lead its efforts here. He worked at Google for almost a decade, helping to develop the tech giant’s TPUs, which power its AI models, including Gemini.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
Data centers
Those AI chips go in servers that are stacked inside giant data centers. OpenAI is currently working with all the major cloud providers, renting the computing power needed to train and run its models and chatbot service. However, Friar, the CFO, said on Tuesday that OpenAI will build its own data centers in the future.
“I’ve rarely seen scaled technology companies that don’t own and operate their own infrastructure,” Goldman Sachs internet analyst Eric Sheridan told Business Insider this week. “This helps them control more of their own destinies over time.”
Models and developers
At the model layer, OpenAI has already established a beachhead, with a frontier offering, GPT-5, that is among the best. It also recently launched a powerful open-weight model.
On top of that, the startup has APIs that developers use to tap into these models. About 4 million developers have built on this OpenAI platform already.
Distribution and gadgets
Without broad distribution, no one will use your AI products. OpenAI is working on new ways to get ChatGPT and other services to consumers. The startup bought Jony Ive’s AI gadget startup for more than $6 billion a few months ago, and executives have hinted at form factors that could mean OpenAI’s chatbot follows people wherever they go. It also launched a web browser, another powerful distribution tool.
Software and applications
The startup is also trying to reach corporate customers, not just consumers. It already has roughly 5 million paid seats from ChatGPT business products. On Tuesday, Friar suggested that OpenAI might challenge established software companies at some point, or at least the startup’s AI technology could support such disruption.
This takes OpenAI into the realm of applications, the software that sits on top of tech stacks and forms direct relationships with end users. ChatGPT has 700 million weekly average users, but the startup is looking beyond that.
In May, OpenAI hired Fidji Simo as CEO of Applications. The former Instacart CEO and Meta executive reports directly to Altman. This month, the startup brought on Vijaye Raji as CTO of Applications and acquired his startup, Statsig, for just over $1 billion.
A few days later, Simo announced the OpenAI Jobs Platform, a new service that will match tech employers with qualified candidates. It looked eerily like LinkedIn, the professional network owned by its partner Microsoft.
Goldman’s Sheridan said the ultimate goal would probably look something like Google’s suite of apps, which include not just Search, but also Maps, YouTube, Gmail, Calendar, Chrome, Classroom, Translate, Earth, Drive, Meet, Photos, Sheets, Docs, and Slides.
To build this entire AI stack itself and realize its dreams, OpenAI will need a mountain of cash, talent, and luck.
Sorting out its tortured corporate structure should help on the fundraising side. Indeed, without an agreement with Microsoft, OpenAI could miss out on at least $10 billion in funding from SoftBank. That’s looking a little more likely now.
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at [email protected].