Home » Meta’s $145B AI Spend Doubles CapEx to Fund Inference Shift

Meta’s $145B AI Spend Doubles CapEx to Fund Inference Shift

By

–

Meta's $145B AI Spend Doubles CapEx to Fund Inference Shift

Meta Platforms has tightened its 2026 capital‑expenditure outlook to a $135‑$145 billion range, essentially doubling the $72 billion it spent in 2025, with the bulk earmarked for AI‑driven inference infrastructure.

Why the Shift to Inference Matters

The company says its AI work is already boosting ad impressions by 10% year‑over‑year and lifting the average price per ad by 14% in Q3 2026, while its family of apps logged 3.64 billion daily active users. By funneling cash into data centers, chips, networking gear and the systems that train Meta’s Llama models, Meta is moving from the costly, research‑heavy training phase to a production‑focused inference stage that powers real‑time ad targeting and content recommendations.

Capital Allocation Breakdown

Meta’s 2026 spending plan covers the full AI stack: new server farms, custom AI chips, high‑speed networking products and the hardware needed to run Llama models at scale. Capital outlays in the third quarter of 2026 rose to $34.91 billion from $19.37 billion a year earlier, reflecting the rapid build‑out of this infrastructure.

Financial Trade‑offs

Despite generating $32.79 billion in operating cash flow, Meta posted a negative free‑cash‑flow of $2.12 billion in Q3 2026, underscoring the cash‑intensive nature of the AI push. The firm also carries $454.52 billion in non‑cancelable contractual commitments and $355.73 billion in leases tied to cloud capacity, data‑center power and energy needs.

Shareholder and Legal Context

Meta had planned to issue a new non‑voting Class C share to raise up to $95 billion for AI financing, but a Delaware court‑ordered freeze on the share issuance remains in place while a shareholder lawsuit proceeds. The proposed shares would have let the company fund AI spend without diluting Mark Zuckerberg’s voting control through his 10‑vote Class B shares.

What This Means for the AI Landscape

By allocating roughly $145 billion to inference hardware, Meta is betting that faster, cheaper on‑device and cloud‑based AI will translate into higher‑priced ad inventory and deeper user engagement. The move positions Meta’s ad engine to compete more aggressively with rivals that are also pouring billions into AI, while its forward 12‑month P/E of 18.78 × sits below the Zacks Internet Software industry average of 26.90 ×, suggesting the market is pricing in the near‑term cash strain.

Meta’s next milestone will be the Q4 2026 earnings release, slated for early February 2027.

Sources

Photo: Anthony Quintano from Honolulu, HI, United States (CC BY 2.0)

Get breaking stories first on our Telegram channel.

More stories by

Most read

AI Tools

Artificial Intelligence

Business

World News