Microsoft, Google and Meta Cloud Investments: Nvidia Will Get Paid, But...
Nvidia (NVDA) will take the lion’s share of revenue (see TSMC’s July numbers), as MSFT, GOOGL and META build out their respective Generative AI infrastructure. However, beyond the building of infrastructure, software developers have been slow to build applications and services that incorporate Gen AI capability. Similarly, users have been slow to adopt Gen AI applications. What is the probability that 5-10 years from now that this Gen AI infrastructure build will look like China’s ghost cities?
We pulled the figures below from each company’s cash flow statements (MSFT FY 2025 estimate, GOOGL and META CY 2024 estimates are TEK2day’s).
MSFT: Azure and more specifically Gen AI have driven much of MSFT’s spending increase. We expect that to be the case for FY 2025 (MSFT is on a June FYE) with an estimated $60 billion of property and equipment spend.
GOOGL: Google has been cloud native since day 1. Therefore, building data centers and making related investments is nothing new to the team in Mountain View. However, Google has some catching up to do (as does MSFT, AMZN and others) in the Gen AI space as Anthropic and OpenAI have more sophisticated Large Language Models. That said, we expect Google Gemini to close the gap, particularly with OpenAI over the next year as OpenAI’s culture has driven 9 of the company’s 11 co-founders out of the business.
META: Meta / Facebook has invested heavily in Gen AI (2022 and 2023), although it is taking an open source approach which we prefer to closed, proprietary models.
