With AI models becoming commoditized, the race is now about execution, and Microsoft risks being left behind
-
By Parmy Olson / Bloomberg Opinion
-
-
In an industry numb to eye-watering artificial intelligence (AI) bets, it takes a lot to make a chief executive hesitate. So Nvidia chief executive officer Jensen Huang (黃仁勳) blinking at one such commitment to OpenAI is worthy of notice. Huang has been telling industry associates that his US$100 billion investment in the ChatGPT maker announced last year was nonbinding, the Wall Street Journal reported. He also reportedly privately criticized the company’s lack of business discipline.
It looks as though Huang’s planned investment, originally tied to an infrastructure build-out, would manifest as a smaller bet in the tens of billions as part of OpenAI’s fundraising efforts ahead of a potential initial public offering. OpenAI is said to be in discussions with Nvidia, Microsoft Corp and Amazon.com to raise about US$100 billion in capital, separate from the proposed infrastructure deal with Nvidia.
Huang denies he is unhappy with OpenAI.
“We will invest a great deal of money,” he told reporters on Jan. 31, but he is right to hedge his bets.
STRUGGLES
For all the outward charm, OpenAI CEO Sam Altman’s management of the company has been unsettling. There was his dramatic firing in late 2023, followed by a stream of complex, eye-popping deals that put his company on the hook for US$1.4 trillion in computing commitments, 100 times more than OpenAI’s projected revenue last year. The company’s product rollouts have been frantic: For example, efforts to build a developer marketplace with the GPT Store and custom GPTs fizzled after lacking a clear strategy.
However, OpenAI is not alone in struggling to execute. so too is Microsoft. The stable software giant, whose stock has nearly doubled since the launch of ChatGPT, got a steal with its early, US$13 billion bet on OpenAI, which now translates to a 27 percent stake valued at US$135 billion — or more than 10 times its original investment. Thanks to a restructuring deal announced in October last year, Microsoft has exclusive access to OpenAI’s intellectual property and models through 2032, cleaner agreements than before.
However, Microsoft is failing to capitalize on that coveted access. For all of Altman’s questionable business acumen, he is still managing to churn out some of the world’s most powerful artificial intelligence models. So why does Microsoft’s flagship AI tool Copilot, which has the seeming advantage of being underpinned by OpenAI’s technology, lag the competition?
Last month, Anthropic launched Claude Cowork, an app it built in 10 days with its own AI coding tool. It is remarkable. With permission, it could operate my personal computer, organize files, generate PowerPoint decks and Excel spreadsheets from my documents and reply to LinkedIn messages.
Microsoft Copilot cannot do any of this, despite Microsoft owning Windows, Office and LinkedIn. Users have instead complained that Copilot is confusing, constrained and hard to use, underscoring a perplexing gap between the high quality of OpenAI’s models and Microsoft’s seeming inability to turn them into useful products.
Something is wrong with Microsoft’s internal research efforts. The company has invested heavily in building its own AI models to broaden its strategy beyond OpenAI. However, that effort looks misguided and some industry watchers are now waiting to see what happens to the leadership of Microsoft’s AI program, says David Rainville, lead manager of Sycomore Sustainable Tech, which has invested in the software giant and bought into the company’s share rout on Jan. 29.
AI RACE
If the company does not release an equivalent to Claude Cowork in the next six months, heads would have to roll, Rainville said.
“There’s definitely been a disconnect between the quality of models and what Microsoft has been able to execute on,” he added.
Some blame for that can also be flicked back at OpenAI, which has resisted sharing full technical details of its models — to the frustration of Microsoft executives.
Rainville said that Copilot is already “drastically better” since Microsoft in August last year plugged it into OpenAI’s latest GPT-5 model — but it needs to be able to carry out tasks on a computer to keep up with competitors such as Anthropic and OpenClaw.
The next few years would likely put Microsoft at a crossroads. Should OpenAI’s financing efforts falter, it could be absorbed into a larger entity — perhaps Microsoft. To make the most of that possibility and the access it has today, the software giant should capitalize on its partnership, however tense it might be. Huang’s hesitation should be taken as a warning: Microsoft needs to turn its privileged access into better products. With AI models becoming increasingly commoditized, the race is now about execution, and Microsoft risks being left behind.
Parmy Olson is a Bloomberg Opinion columnist covering technology. A former reporter for the Wall Street Journal and Forbes, she is author of Supremacy: AI, ChatGPT and the Race That Will Change the World. This column reflects the personal views of the author and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
