Microsoft's Strategic Pivot to AI and the OpenAI Partnership

Key Insight

Microsoft possessed a long history of AI investment dating back to the 1990s, but faced considerable internal resistance to adopting deep learning techniques, despite early indications of their potential in speech recognition during the late 2000s. Many of the company's leading AI researchers expressed antagonism towards neural networks, with one prominent colleague publicly dismissing their potential in computer vision as a 'dead end' in 2012. Recognizing these internal challenges and the necessity for a different strategic approach, Satya Nadella appointed Kevin Scott as Chief Technology Officer in early 2017. Scott, observing the fragmented AI investments and the slow adoption of deep learning despite the high intellect and enthusiasm within the company, took the critical step of centralizing the company's Graphics Processing Unit (GPU) budget, mandating that these vital resources for training complex AI models be allocated only after demonstrating 'really, really strong, evidence-based conviction.'

Scott candidly acknowledged by 2019 that Microsoft was 'multiple years behind the competition in terms of machine learning.' He also recognized that competitors like Google and Facebook held significant advantages due to their richer data pools, accumulated from billions of daily search queries and vast troves of user data from popular services such as Instagram and WhatsApp. To rapidly bridge this competitive gap, Scott concluded that Microsoft required 'high-ambition partners.' His assessment led him to OpenAI, an organization originally established to counterbalance the growing influence of Big Tech. Scott viewed OpenAI as unequivocally the 'highest-ambition partner' in the field, noting that its blend of idealism and intensity mirrored the vibrant atmosphere of Google in its formative years.

The 'most unlikely of partnerships' between Microsoft, a colossal enterprise with over 200000 employees, and OpenAI, an organization founded to mitigate the power of large tech companies, was finalized around 2018. This strategic non-controlling investment granted Microsoft critical access to OpenAI's advanced technologies, enabling the rapid integration of cutting-edge AI into its products and accelerating its competitive catch-up. OpenAI, in turn, secured essential resources, receiving approximately half a billion dollars in Azure credits to train and execute its models within Microsoft's cloud infrastructure. By late 2019, Microsoft further committed hundreds of millions of dollars to develop a bespoke computing environment, integrating tens of thousands of GPUs, which became the foundational platform for training GPT-3. This innovative, externally focused strategy, which 'neither Bill nor Steve would have done' due to their preference for internal development, swiftly yielded results, exemplified by GPT-2's independent ability to code and the subsequent launch of GitHub Copilot, a programming tool generating over $100 million in revenues within a year. This resurgence in innovation helped Microsoft reclaim its position as the world's most valuable company and become one of the first two trillion-dollar companies.