Microsoft has been investing $13 billion in artificial intelligence startup OpenAI, the company said on April 10. These investments have valued OpenAI at nearly $30 billion, with a lot of potentials but also a lot of uncertainty.
When Microsoft invested $1 billion in OpenAI in 2019, the deal was no more eye-catching than a typical corporate venture capital investment. The startup market was hot at the time, and AI was just one of many areas attracting huge investments, including electric vehicles, advanced logistics and aerospace.
Three years later, the whole market looks very different.
Microsoft’s not-so-impressive investment in OpenAI is now a major topic of conversation among the venture capital community and investors. Many people are trying to figure out what this means for the value of their holdings. Microsoft’s cumulative investment in OpenAI has reportedly risen to $13 billion, and OpenAI is already valued at about $29 billion.
That’s because Microsoft is not only investing huge amounts of money directly into OpenAI but is also providing computing power exclusively for OpenAI’s product development and support for developer programming interfaces. Right now, many startups and large companies are racing to integrate their products with OpenAI, which means a lot of workloads are running on Microsoft cloud servers.
Microsoft itself is integrating OpenAI’s generative AI technology into its Bing search engine, marketing software, GitHub coding tools, Microsoft 365 suite of productivity tools and Azure cloud services. In total, these will add more than $30 billion a year to Microsoft’s revenue, about half of which will come from Azure, said Wells Fargo analyst Michael Turrin.
What does this mean for Microsoft’s investment and overall planning?
Bragging rights to capital
OpenAI was founded in 2015 and started out as a nonprofit, but that organizational structure changed in 2019. At the time, two executives published a blog post announcing the creation of a “profit-capped” entity called OpenAI LP within the original organization, with a limit that required the first investors to earn no more than 100 times the return on their investment, with later investors like Microsoft earning even lower returns.
An OpenAI spokeswoman said Microsoft will receive a percentage of OpenAI LP’s profits after the investment is recovered, up to an agreed cap, and the rest will go to a nonprofit organization affiliated with OpenAI LP. A Microsoft spokesman declined to comment.
Greg Brockman, the co-founder of OpenAI and co-author of the above blog post, said in 2019 that for investors, the system “feels comparable to what they would get if they invested in a fairly successful startup, but less than what they would get if they invested in the most successful startup of all time. The return”.
This is a rare pattern in Silicon Valley, where venture capitalists are used to maximizing returns on their investments. For Elon Musk, one of OpenAI’s founders and early backers, it doesn’t make much sense either. On several occasions this year, Musk has taken to his personal social media Twitter feed to express his concerns about OpenAI’s unconventional structure and its impact on the AI industry as a whole.
“OpenAI was started as a not-for-profit open source company (which is why it was named ‘Open’ AI) to rival Google, but now it has become a closed-source company that is effectively controlled by Microsoft and seeks maximum profits. ” Musk tweeted in February that “this was not my intention at all.”
If OpenAI succeeds, Brockman said, it will “create orders of magnitude more value than any company to date. As a major investor in OpenAI, Microsoft is bound to benefit from it.
In addition to the investment income, OpenAI may also help Microsoft to significantly reverse its decline in the field of artificial intelligence. Previously, Microsoft has suffered many setbacks in the field of artificial intelligence, and the company has not built any notable business. After a flurry of activity, Microsoft removed Clippy Assistant from Word, took Cortana off the Windows OS taskbar, and removed the Tay chatbot from Twitter.
Unlike areas such as advertising or security, Microsoft has not disclosed the size of its artificial intelligence business. The company’s chief executive Satya Nadella said only last October that Azure Machine Learning Services revenue had doubled for four consecutive quarters.
The partnership with OpenAI has at least given Nadella the capital to brag about it. At Microsoft’s annual shareholder meeting a month after ChatGPT went live, Nadella revealed that “ChatGPT, the most popular AI application today, was trained on Azure supercomputers.”
In February, Microsoft held a press conference at its headquarters in Redmond, Washington, to announce the integration of new artificial intelligence technology for the Bing search engine and Edge browser.
While Bing’s integrated chatbot generated some incorrect answers at launch and subsequently had a number of problems interacting with users, Google’s AI chatbot, Bard, also stumbled and didn’t work well for Microsoft, leading Google employees themselves to describe the Bard project as too “rushed” and “screwed up, “screwed up.
Despite some early problems, the tech industry as a whole is enthusiastic about this new technology, which is based on a large language model.
At the heart of OpenAI’s chatbots is a large language model called GPT-4, which learns to automatically compose fluent text after being trained on a large number of online sources, said an OpenAI spokesperson, adding that Microsoft has an exclusive license to GPT-4 and all other OpenAI artificial intelligence models.
In addition to OpenAI’s large language models, there are a number of other similar products on the market.
Last month, Google said it had started letting some developers try out a large language model called PaLM.
Startups AI21 Labs, Aleph Alpha and Coherence all have large language models, and Google-backed Anthropic also offers such products and has chosen Google as its “preferred” cloud provider. Like Altman and Musk, Anthropic co-founder Dario Amodei, who previously served as vice president of research at OpenAI, has expressed concern about the limitless capabilities of artificial intelligence.
In an email, an Anthropic spokesperson said, “We were and remain focused on developing innovative architectures that provide incentives for the secure development and deployment of AI systems, and we will share more information about this in the future.”
Looking across the AI industry, one thing is clear: It’s just getting started.
Quinn Slack, CEO of code search startup Sourcegraph, said that while OpenAI is a top provider of large language models, he personally has yet to see a significant advantage to Microsoft from working with OpenAI.
“I don’t think people should feel that Microsoft has a complete lock on OpenAI, or that OpenAI is doing what Microsoft wants it to do,” Slack said. “I believe that OpenAI employees are motivated to develop amazing technology and make it as widespread as possible. They all see Microsoft as a great customer, not as wanting to control everywhere. That’s great, and I hope it stays that way.”
Skeptics about OpenAI abound, too. Late last month, the nonprofit Center for Artificial Intelligence and Digital Policy called on the Federal Trade Commission (FTC) to block OpenAI from releasing a new commercial version of GPT-4, saying the technology is “biased, deceptive and would have a negative impact on privacy and security. The FTC has called on the FTC to block OpenAI from releasing a new commercial version of GPT-4, calling the technology “biased, deceptive and a threat to privacy and public safety.