Microsoft is the godfather to OpenAI. Let’s face it,without the former’s support, the Sam Altman-led company wouldn’t exist as we know it. It would’ve barely survived a day, as it’s NOT your ‘normal company’.
UK-based industry analyst firm CCS Insight recently predicted that within three years, OpenAI will find it so difficult to raise funds that it will need to sell to Microsoft. That could explain why OpenAI recently opened a new office near the Microsoft headquarters in Seattle.
Microsoft’s partnership with OpenAI began in 2019 with an initial investment of $1 billion. In the Netflix series What’s Next? Microsoft founder Bill Gates revealed that he initially thought his next big focus would be eradicating malaria, and not AI. However, when Greg Brockman gave him a demo of GPT-4, it completely ‘blew his mind’, changing his perspective on AI.
Fast forward to today, OpenAI has raised $6.6 billion at a valuation of $157 billion. This time, however, the startup is not solely dependent on Microsoft, as the funding was led by Thrive Capital. Other significant participants in the funding round included NVIDIA, SoftBank, Khosla Ventures, Altimeter Capital, Fidelity, and MGX.
Surprisingly, Microsoft has invested less, perhaps focusing on its in-house AI research. Earlier this year, the company appointed Mustafa Suleyman, former co-founder and head of DeepMind and Inflection AI, as the CEO of Microsoft AI.
Microsoft has been building edge LLMs and recently launched the new Phi-3.5 models, which outperform Google’s Gemini 1.5 Flash, Meta’s Llama 3.1, and OpenAI’s GPT-4o.
The funding for OpenAI came at the right time as the startup is currently operating at a loss and has yet to make a profit. The company’s forecasts indicate it won’t achieve profitability until 2029, with projected revenue reaching $100 billion. However, losses could escalate to $14 billion in 2026, nearly three times this year’s anticipated losses, according to a recent report.
Moreover, OpenAI anticipates a steep rise in computing costs for model training over the coming years, with expenses potentially reaching $9.5 billion annually by 2026.
Today, a major chunk of the training compute is taken care of by Microsoft, which provides Azure servers to train OpenAI’s models. Back in 2022, when OpenAI debuted ChatGPT, Altman credited Microsoft for that. “Microsoft, and particularly Azure, don’t get nearly enough credit for the stuff OpenAI launches. They do an amazing amount of work to make it happen; we are deeply grateful for the partnership. They have built by far the best AI infrastructure out there,” posted the OpenAI chief on X.
The love was reciprocated by the Microsoft CEO at Ignite 2023 where he showered unlimited affection, heaping praises on OpenAI. He went on to say that the company is grateful to have OpenAI as part of its journey, besides showcasing the readiness to give up everything it has built so far. This included rebranding its search engine, revamping its Azure cloud infrastructure and much more.
Nadella has, on many instances, expressed his feelings towards OpenAI. Even on DevDay 2023, he said, “We love you guys… You guys have built something magical.” The relationship between Microsoft and OpenAI is symbiotic, where neither can survive without the other. But today, both the AI startup and the tech giant are exploring ways to reduce their dependence on each other.
The tech giant now offers LLMs from NVIDIA, Mistral, Cohere, G42 and Meta, in addition to the frontier models available from OpenAI in the Azure AI Studio. However, Microsoft’s flagship products, such as Copilot 365, still rely on OpenAI’s models.
Recently, the company introduced Copilot agents within its Microsoft 365 Copilot platform, designed to help businesses customise AI to meet their specific needs. These Copilot agents connect to data sources like SharePoint and Dynamics 365, automating tasks and integrating with existing systems to streamline workflows.
Microsoft also recently updated Copilot, adding voice and vision features.
OpenAI x Microsoft
Today, with AI startups like Anthropic, Cohere, and xAI breathing down OpenAI’s neck, the company will likely struggle to survive without Microsoft’s support.
The models being developed by OpenAI, such as the video generation model Sora and its latest release o1, require significant computing capacity to create and make available to consumers worldwide. It’s likely that without Microsoft’s initial $10 billion investment, this would not have been possible.
Under the current arrangement, Microsoft claims 75% of OpenAI’s future profits until this initial investment is repaid. After repayment, Microsoft will be entitled to 49% of the profits, up to a theoretical cap. With OpenAI planning to shift its structure from non-profit to for-profit, it remains to be seen what the new deal between Microsoft and OpenAI will entail.
Today, OpenAI provides Microsoft with a 20% commission for sales made to other businesses accessing its AI models via an application programming interface. In turn, Microsoft pays OpenAI a 20% commission when it resells OpenAI’s models to Azure customers through a Microsoft API.
In his keynote at Microsoft Build 2024, Nadella said that OpenAI is Microsoft’s “most strategic and most important partner”.
Ironically, just recently, Sebastien Bubeck, a Microsoft AI researcher, announced that he is departing to join OpenAI. He played a key role in Microsoft’s work on generative AI models over the past two years.
At the event, Microsoft CTO Kevin Scott said that if the system that trained GPT-3 was a shark and GPT-4 an orca, the model being trained now by OpenaI is the size of a whale. “This whale-sized supercomputer is hard at work right now,” he added.
Microsoft has always supported OpenAI during its challenging times. Following Altman’s unexpected ouster last year, Nadella announced just two days later that Altman had been hired to lead a new AI research team at Microsoft. This move was seen as a way for Microsoft to safeguard its stock price and could strengthen its influence within OpenAI.
Slowly Drifting Away
OpenAI is reportedly unhappy with Microsoft because the latter was not quick enough to deliver the necessary computing power, prompting the company to seek other data center agreements, a process it had already initiated. Interestingly both of them recently received the much awaited NVIDIA Blackwell GPUs.
This year, OpenAI also attempted to establish closer ties with Apple in the hope of securing additional funding. However, Apple did not participate. For Apple, this could mean investing more in its internal R&D and architecture rather than pursuing external investments.
Meanwhile, OpenAI is collaborating with Oracle to secure additional computing resources. This partnership will allow it to further scale its operations and services. Altman said, “OCI will extend Azure’s platform and enable OpenAI to continue to scale.”
The company is currently in discussions to lease the entire Abilene data center site from Oracle, which could eventually expand to 2 gigawatts if Oracle secures additional power for the location, according to a recent report.
OpenAI’s attempt to look for more compute options besides Microsoft Azure is natural, considering that the startup has made a blueprint for the next few years, where it has shifted its focus from LLMs to AI agents and consequently ASI. However, there is very little possibility that Microsoft will leave OpenAI and it will continue to play a huge role in OpenAI’s journey.
The post OpenAI Would Have Barely Survived Without Microsoft appeared first on AIM.
