Santosh Janardhan, Meta's head of infrastructure and co-head of engineering, announced he will step down from leading the infrastructure organization in early 2027, ending a 17-year tenure at the company. The departure was revealed in an internal memo obtained by Bloomberg News; the message did not specify reasons for the exit, citing only the need to "pass the baton." Janardhan will remain at Meta through the end of the first quarter of 2027 to support the transition, moving into an advisory role within Meta Compute.
The successor is already in place: Dave Brown, vice president of infrastructure who joined Meta earlier in 2026 after 19 years at Amazon Web Services. At AWS, Brown worked on compute, cloud, and AI systems — expertise that positions him directly to lead Meta's next phase of infrastructure operations. The transition comes at a pivotal moment as Meta scales massive data center investments and AI training capacity to compete with Google, Microsoft, and OpenAI.
Janardhan is one of Meta's longest-serving executives, having joined the company in 2009 when it was still called Facebook. Over his nearly two-decade tenure, he oversaw the evolution of infrastructure serving billions of users across Facebook, Instagram, WhatsApp, and Threads. He was the architect of the technical foundation that enabled Meta to grow into one of the world's largest internet platforms, managing an ecosystem that processes trillions of requests per day.
In recent years, Janardhan's responsibilities expanded well beyond traditional social media infrastructure. He took on leadership of Meta Compute, an initiative launched to manage and scale the company's data center strategy. The project represents a fundamental shift: Meta, which for decades used its data centers almost exclusively to serve content for its own users, now wants to sell AI compute capacity to external companies. That strategy depends on scale that has never been attempted at the company.
The company's stated ambition is to spend up to $27 billion on AI infrastructure in 2026 — a figure that has grown steadily since the race for foundation models and autonomous agents intensified. Janardhan was the central figure in executing that investment, and his departure — without an announced public reason — raises questions about the direction the new leader will give this massive operation.
The organizational context at Meta has been volatile. In early 2026, the company executed one of the largest restructuring efforts in its history, cutting 10% of its workforce — approximately 8,000 employees — and canceling 6,000 open positions as part of "Project OT" (Organization Transformation). The plan went further: it envisioned autonomous AI agents taking over routine tasks performed by thousands of employees, with smaller human teams supervising the systems. But the project encountered both internal resistance and AI technology that failed to perform as expected. In July, CEO Mark Zuckerberg admitted at an internal meeting that "the trajectory of agentic development over the last four months hasn't really accelerated in the way that we expected."
The infrastructure leadership change reflects Meta's central tension: the company needs massive AI operations but is still learning how to organize around them. Janardhan built infrastructure for social media at planetary scale; Brown brings AWS-specific expertise in cloud and AI compute — the exact profile Meta needs for the next phase.
What remains to be seen is how deeply the company's culture will shift with an Amazon-born executive taking the helm. Meta operates with an engineering structure distinct from Amazon's, making technical decisions rapidly with few hierarchical layers. AWS, by contrast, is an organization with decades of operational maturity in cloud computing. Whether the transition leads to greater adoption of Amazon practices at Meta — or whether Janardhan leaves behind a team prepared to maintain innovation velocity without an external manager — will define the next cycle of Meta's infrastructure. The answer likely arrives when the company announces Q4 results and reveals how far it is from selling compute to third parties — the real test for Meta Compute.
Sources: Bloomberg, CNBC, TechCrunch
✓ Independent sources cross-checked and verified before publishing
