Meta ramps up AI spend as it pushes advanced models

Meta is continuing to invest aggressively to meet its technology infrastructure requirements, involving datacentre expansion and supply chain deals to secure components for future capacity. The company’s latest quarterly earnings filing shows Meta has embarked on a strategy to sign up for multi-year cloud contracts driving $107bn in contractual commitments for Q1 2026.

For the quarter that ended in March 2026, Meta posted revenue of $56.3bn, a 33% increase from the same quarter in 2025.

The company has forecast that its capital expenditures, including principal payments on finance leases, has increased by $10bn due to component price increases and additional datacentre costs, putting CapEx in the range of $125bn to $145bn.

Chief financial officer Susan Li said: “Our investments will support our training needs for future models, and most importantly, provide us with the inference capacity necessary to deliver personal and business agents to billions of people around the world, along with several other AI product experiences we’re developing.”

Responding to a question during the earnings call about balancing model training versus product launches and the potential impact on Meta’s 2027 capital expenditure, CEO Mark Zuckerberg said the company is moving towards greater capabilities and scaling of AI models. “We have the research team, which is focused on scaling increasingly intelligent models with capabilities for the specific things that we’re focused on, which are business and personal agents,” he said.

Beyond model development, Zuckerberg said: “We have our next set of more advanced models in training now. And that work will continue. I don’t think we’re going to be done with that anytime soon.”

He emphasised the significance of Meta AI models in product development. “The product team is really unlocked to be able to build things on top of our models because we now have a very strong model,” said Zuckerberg.

When Li was asked about how the company uses large language models in its ad business to direct adverts to users, she said: “The size and complexity would make them too cost-prohibitive.”

Instead, Li said the way Meta uses large language models is to transfer knowledge to smaller, more lightweight models. “The inference models are bound by strict latency requirements since they need to find the right ad within milliseconds, and that has, again, historically prevented us from meaningfully sizing up – to scale up their size and complexity,” she added.

Li said Meta plans to tackle this scaling issue with the introduction of an adaptive ranking model later this year, using the model complexity of a trillion parameters. “We made advances in the model architecture and co-design the system with the underlying silicon, so it maintains the sub-second speed that is required to serve ads at scale,” she added.

Commenting on Meta’s strategy, Forrester vice-president research director Mike Proulx said: “Meta’s future‑facing AI ambitions are being underwritten almost entirely by the company’s legacy business: advertising inside social media apps. There’s no material AI revenue yet.

“The question is whether Meta’s core can continue to act as a cash cow while the company reduces headcount and diverts focus toward AI,” he said. “If Meta’s ad engine slows, the market’s margin for patience shrinks fast. Meta’s slight dip in daily active users is already beginning to raise eyebrows. Q2 will tell us if it’s really just a blip or the start of a trend.”