Meta Q3 2026 Earnings: Capex Floor Raised to $135 Billion as New Share Class Funds AI Spend

Meta Platforms reported third-quarter revenue of $64.60 billion on October 21, up 26% from a year earlier, and raised the floor of its 2026 capital-expenditure plan to $135 billion from $130 billion while holding the top of the range at $145 billion. The company also said its board approved a new class of non-voting common stock, a financing mechanism aimed squarely at paying for the AI buildout without touching founder voting control.

The capex number is the one that matters for the AI supply chain. Meta said it now expects 2026 capital expenditures, including principal payments on finance leases, in a range of $135-145 billion, narrowed from its prior outlook of $130-145 billion. The floor came up $5 billion in one quarter, and the spend is already running hot: Q3 capital expenditures were $34.91 billion, against $64.60 billion of quarterly revenue, and capital spending is starting to outweigh cash inflow. Meta generated $32.79 billion of operating cash flow in the quarter but free cash flow was a negative $2.12 billion, and headcount fell 15% year over year to 66,775 as the buildout absorbs the dollars. That is the pattern across the hyperscalers, and for Nvidia, Broadcom, TSMC and the power infrastructure names it is the demand signal every supplier is watching.

Related reading: Five AI Deals in One Day, and Not One Was About GPUs.

The share-class move is the structural story. Meta said its board established a new class of publicly listed, non-voting Class C common stock with an effective date of January 2027, following the unanimous recommendation of a special committee of disinterested, independent directors. The structure mirrors the mechanism Google used in 2014: issue a non-voting class to fund large investments and acquisitions without diluting the controlling founder's votes. For Meta it creates a war chest for AI capacity and, potentially, deals, while keeping voting power exactly where it sits today. The careful committee mechanics read as an attempt to head off the governance objections that followed Alphabet's version.

Profit took a hit from a settlement, not from the buildout. Q3 costs and expenses were $51.61 billion, up 68% year over year, and that includes a $9.89 billion charge tied to Meta's August agreement with 52 state attorneys general to resolve litigation over teen mental health and social media, with payments of up to roughly $18.59 billion. The charge pushed the full-year expense outlook to $175-179 billion and it is what cut the operating line: income from operations came in at $12.99 billion, and excluding the charge it would have increased by $9.89 billion to $22.88 billion, with diluted EPS up $3.57 at $7.40. Reported diluted EPS of $3.83 still climbed from $1.05 a year earlier, but the driver was the tax line, not the settlement: in the third quarter of 2025 the effective tax rate was 87% on an $18,954 million provision, versus $3,141 million this quarter. Reality Labs lost $4.13 billion on $443 million of revenue, and Meta still expects full-year Reality Labs losses in the $18-19 billion range, narrowing from 2025.

Related reading: AI Capex Hits $998 Billion in 2026, and Memory Is Why.

The constraint on this whole plan was never the money, and here is the part that now cuts both ways. On October 21 the president said new and large data centers should generate their own electricity and give or sell any excess back to the local community or grid. Only the states can actually mandate siting and power terms, and several already have: New York Governor Kathy Hochul signed a statewide moratorium on new hyperscale data centers, Ohio Governor Mike DeWine paused new data-center tax exemptions, and the Illinois Department of Commerce and Economic Opportunity stopped processing new data-center incentive applications, per the governors' announcements and Executive Order 62, as reported by AP News, Signal Ohio and CBS News. The real question for Meta is not whether it can raise the money but whether and where the states will let it be spent.

Self-generation on that scale is not cheap, and it is not fast. Building a co-located power plant, gas-fired or geothermal depending on the site, adds years to a project already on a multi-year permitting clock, and it changes the economics of a $135 billion plan. Meta has been a steady buyer of long-dated power contracts, and the open question for the next few quarters is whether hyperscalers simply pay more per megawatt or lean on new supply that can actually be delivered. That line item now moves the unit economics of AI capacity as much as the chip bill does.

Related reading: Meta Earnings Preview: A $130-145B Capex Guide Frames Q4.

The market read is straightforward. A raised capex floor is constructive for the semiconductor and power names that sell into the big four, and Meta's after-hours move was higher on the print: shares closed the regular session at $697.52 and first traded at $725.86 in extended hours. The offsetting question is whether the industry can physically deploy $135-145 billion of new capacity through a tightening power and permitting window. Watch next for how hyperscaler peers guide in the coming weeks, whether the Class C mechanism draws investor pushback before its January 2027 effective date, and Meta's fourth-quarter report early next year.

Related reading: Google Locks In 702 MW of Fuel-Cell Power From AEP for Its AI Data Centers.

This content is for informational purposes only and does not constitute financial advice. Past performance is not indicative of future results. Consult a qualified financial advisor before making investment decisions.