This article is part of our series that explores the business of artificial intelligence
When Microsoft decided to invest in OpenAI in 2019, there was no idea how it would recoup the investment. OpenAI was creating game-playing bots, robotic hands, and language models, an amusing-but-unstable technology that had no business prospects. Three years later, Microsoft has turned the partnership into a very successful business opportunity and might soon turn in in profits on the billions of dollars it has poured into OpenAI.
According to the transcript of Microsoft’s earnings call, it now has “more than 2,500 Azure OpenAI Service customers, up 10X quarter-over-quarter.” Launched in 2021, Azure OpenAI Service provides access to OpenAI’s GPT-3 and DALL-E models through Microsoft’s cloud platform. The service’s customers include big industry names such as Mercedes-Benz and Shell as well as tech companies such as Coursera and Grammarly.
Epic Systems, which has a large share of the market for healthcare software in the U.S., is using Azure OpenAI Service “to integrate the next generation of AI with its industry-leading EHR software,” and Siemens will be using “a Teams app integrated with Azure OpenAI Service to optimize factory workflows.”
Moreover, Microsoft is using OpenAI’s technology to maintain its edge over its competitors in different industries with initiatives such as the integration of GPT-4 into the clinical documentation applications developed by Nuance, a company it acquired in 2021.
GitHub Copilot, the AI code generation tool powered by OpenAI models, is now used by more than 10,000 organizations, including Coca-Cola and GM. Copilot’s business plan costs $20 per developer per month, which can amount to hundreds of millions of dollars in revenue per year.
Microsoft is also benefiting from OpenAI’s own API service. The customer base for OpenAI is growing, and now includes big names such as Shopify and Snap. OpenAI’s services are powered by the Azure cloud. ChatGPT is reportedly costing OpenAI $700,000 per day, and as the usage of OpenAI’s services grows, so does its Azure bill.
In the earnings call, Microsoft’s Chief Financial Officer Amy Hood said, “It’s easiest… to think about [OpenAI] as a customer of ours, like any other customer who would use the Azure infrastructure and our Azure AI Services in service of supporting their end customers. And so, when they do that, like any other customer who has a commercial relationship with us, we recognize revenue on that behalf.”
Therefore, in addition to receiving exclusive access to OpenAI’s cutting-edge technology, Microsoft is recouping much of the investment it makes into the AI lab as payment for Azure services.
Venture capitalist Tomasz Tunguz has estimated Microsoft’s ML services to have a $900 million run rate. And given its fast growth, it could become a multi-billion-dollar business by the end of the year. Microsoft might have already recouped its investment in OpenAI.
Microsoft’s partnership with OpenAI has also enabled it to revamp its Bing search engine and slowly chip away at Google’s dominant position in online search and browser market share. With AI assistants becoming much better at retrieving information, Microsoft might slowly unbundle some of the tasks that were previously tied to search engines.
Microsoft’s success in commercial AI can be attributed to a few key factors. First, it made the early investment in OpenAI when there was no roadmap for profitability. Second is its vast and exclusive distribution channels. Microsoft’s penetration in many industries through its office, collaboration, and productivity tools enabled it to market OpenAI’s technology to millions of users and thousands of organizations. And finally, an aggressive branding campaign pushed OpenAI and by extension Microsoft at the forefront of the new wave of generative AI models. While OpenAI is neither the pioneer nor the only company working in the field, its name has effectively become synonymous with large language models and text-to-image models. Many people have become aware of advances in AI through news stories about GPT-3, ChatGPT, and DALL-E.
But while Microsoft has the head start in commercial AI, the market is still going through major shifts, and other companies are leaving their mark. Amazon is challenging Microsoft’s position with new products, including its recently launched Bedrock, a platform for scalable, serverless access to a diverse set of LLMs and diffusion models. It is also giving individual developers free access to CodeWhisperer, its rival to GitHub Copilot.
Google is also putting more focus on rolling out AI products and has merged DeepMind and Google Brain to unify its AI research. And startups like Cohere, AI21 Labs, and Anthropic are creating rival services to OpenAI.
At the same time, there is a growing community effort to create open-source alternatives to the closed services provided by big tech. These efforts are meant to prevent a few wealthy companies from having too much power in the fast-growing market for generative AI.