Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
With Sam Altman’s return as chief executive of OpenAI after five days of chaos, it might seem natural to expect things to go back to the way they were. That would be a mistake. OpenAI’s valuation had been quickly racing towards $100bn, making it Silicon Valley’s hottest start-up. But much about the company has now been thrown into question.
The first job for the new board is to sort out the governance mess they have inherited and find a way to reinforce the company’s core mission of making artificial intelligence safe for the world. But they also need to make sense of a business that was at risk of biting off too much more than it could chew, while also alienating some of its closest customers and partners.
Version 1.0 of OpenAI’s commercial business — the company that Altman had been building since ChatGPT took the tech world by storm a year ago — involved racing in all directions at once. Version 2.0 will need to be more focused if it wants to fulfil its potential as the most important new company of the artificial intelligence era and avoid the cracks that were already starting to appear.
Two recent examples show how Altman’s ambitious business plan has risked backfiring, and why a narrower set of objectives is needed.
The first was the announcement this month that the company would give developers the power to create customised chatbots and intelligent agents — called GPTs — and release them through a new OpenAI marketplace, like an app store for AIs.
This move stunned a whole class of companies that had sprung up to build services on top of OpenAI’s models. At one fell swoop, Altman had disintermediated these businesses, putting his own company in the middle of a new AI-enabled market, where it could take a slice of any revenue.
OpenAI’s competitors claim this has led to a spate of interest from AI start-ups that had been building on top of OpenAI and now wish to reduce their dependence.
The second example has been OpenAI’s attempt to become an enterprise software company, providing large customers with more of the tools needed to build on its service. This is something that put it on a collision course with its close partner, Microsoft. It also requires a very different culture, given the high level of sales and customer support required in the enterprise IT world.
The biggest buyers of technology look for one thing in their suppliers above all: long-term predictability. They want reassurance not just that these companies are financially stable and well-managed, but that they are on a clear technology path that justifies sinking money into their products. The chaos of recent days has put a severe dent in OpenAI’s ability to meet these expectations.
These two examples show how the legacy of Altman’s first stint as chief executive has left plenty of questions over the company’s many different directions. In fact, it hasn’t been clear exactly what kind of company he was trying to build. The launch of ChatGPT appeared to give it a shot at becoming the next great consumer internet company, its service the first serious rival to Google in years. But OpenAI’s executives insisted instead that they wanted it to be a platform company, working in the background to support many other businesses that draw on its AI models.
Along with new directions like enterprise software and the marketplace, it was getting hard to see where the company’s ambitions ended and where it would focus its attention. On top of that, Altman has been exploring ways of getting into hardware, making AI-enabled gadgets that might one day replace the need to carry a smartphone, in the process becoming one of the leading producers of semiconductors. It is as though he decided to take on a laundry list of the world’s most powerful tech companies, including Google, Apple, Nvidia and TSMC.
Altman recently told the FT’s Madhumita Murgia that there was a coherent strategy behind this, and that OpenAI’s “one single product” was “intelligence, magic intelligence in the sky”.
But OpenAI’s intelligence is more like an ingredient, an all-purpose input that can be used to create any number of products — and Altman seemed bent on trying to build many of them himself. All of this, meanwhile, was being done on the back of an entity whose primary purpose wasn’t even commercial, but was to conduct research into advanced AI.
Altman at least appears to have his feet back under the CEO’s desk. Whether the new board he answers to will double down on all his ambitions is another question.
richard.waters@ft.com
