The Meta Pivot in Full Throttle

Reports surfaced late Wednesday evening that Meta Platforms (NASDAQ: META) would be expanding its recent headcount reductions to include nearly 20% of its remaining global workforce, roughly 15,000 employees. Simultaneously, the company has significantly raised its capital expenditure guidance for 2025–2026, signalling that tens of billions of dollars will be channelled into data centres, custom AI chips and large-scale computing infrastructure.

This is a high-stakes strategic choice. By leveraging its own internal AI tools to automate roles previously held by thousands of engineers and support staff, Meta is attempting to prove that the "AI Flywheel" can generate internal efficiencies long before it produces external revenue.

The narrative is compelling: the company aims to reach a point by around 2026 where brands can upload a single product image and a budget, allowing AI systems to manage creative production, copywriting, targeting and optimisation across Facebook and Instagram. If true, this is a real moat. But it's also a bet that hasn't yet paid off.

Critical tension: "I don't think that Meta is going to be able to build a best-in-class generalist model, because from a resource standpoint, both with respect to GPUs [graphics processing units] and human talent, it's just not going to be easy to do," said Arnal Dayaratna, a research vice president at the tech consultancy IDC. "I think they're too far behind,"

My view: Meta is simultaneously correct and playing with fire. Correct because its advertising stack is genuinely well-positioned for AI optimization. Playing with fire because it's betting that internal AI talent can execute at the pace of OpenAI, Google, and Anthropic, while cutting the very engineering force needed to build and deploy world-class models.

Sources & References