The AI Platforms Are Real. The People Who Built Them Are Leaving.

Brad Lightcap, OpenAI's former chief operating officer, left the company this week to start something new, the latest in a series of leadership shake-ups at the artificial intelligence lab. His exit came days after a $7 billion tender offer that allowed OpenAI employees to sell their shares.

Six days earlier, four of Alphabet's (NASDAQ: GOOG) (NASDAQ: GOOGL) most important engineers departed. Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le left to found Discovery Loop, a start-up built to automate scientific research.

The same day, Demis Hassabis stepped back from running Google DeepMind, where he was replaced by veteran Koray Kavukcuoglu, the last Gemini co-lead still at the company. All eight authors of Alphabet's seminal research paper "Attention Is All You Need" -- which in 2017 introduced the "transformer" model that now underlies AI large language models -- have now left the company.

Image source: The Motley Fool.

These latest talent reshufflings are just a few chapters in a story that's playing out across Silicon Valley. Top researchers from Meta Platforms, Alphabet, and OpenAI are leaving those tech giants to build AI-powered applications at private start-ups, all while demand for AI is accelerating, especially at the enterprise level.

Cloud adoption is not the problem

Revenue from Microsoft's (NASDAQ: MSFT) Azure grew 43% last quarter as it topped $100 billion in annual revenue for the first time, while its commercial backlog grew 84% to $678 billion.

Amazon (NASDAQ: AMZN) Web Services (AWS) revenue climbed 37% in the second quarter, its quickest pace in nearly five years, and the annual run rate revenue for the company's AI business surpassed $25 billion while growing by triple-digit percentages.

Alphabet's Google Cloud revenue grew 82% last quarter, and it ended the period with a $514 billion backlog. Despite losing the coding war to Anthropic's Claude Code, Gemini Enterprise is still widely adopted across the Fortune 100.

The hyperscalers are building the infrastructure for the next era of computing. But there's still a debate surrounding who will profit most from the model layer and the applications that run on top of it.

Where the value lands next

So, where is the next wave of innovation going to come from in big tech?

Space Exploration Technologies (NASDAQ: SPCX), which went public in June, is a viable candidate. Better known as SpaceX, it has built reusable rocketry, the largest satellite constellation ever deployed, and a vertically integrated stack that now extends into the model layer. Grok 4.6, a frontier-level model built for long-running agents, arrived this week. Grok 4.7 is expected to be ready in a matter of weeks, and Grok 5, trained on SpaceX's engineering data, should follow before year's end.

Meanwhile, its $60 billion acquisition of Cursor, an AI start-up with a coding platform used by nearly two-thirds of the Fortune 500, could close within days. And a new Nvidia partnership could help SpaceX put AI data center hardware into orbit, with the first prototype Starmind satellites scheduled to launch next year.

What SpaceX does not yet have is an established cloud. AWS, Azure, and Google Cloud already serve millions of customers.

Selling inference could be a good business. But investors want to see value accrue on top of the compute, and the people best at building that layer are increasingly heading to private companies. The hyperscaler that figures out both will have a distinct advantage.

Bryan White has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Meta Platforms, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.