While most companies are looking to trim costs, Meta is moving in the opposite direction. Analysts at Rosenblatt and other major firms have highlighted that Meta’s capital expenditures—which sat at roughly $72 billion in 2025—will surge to $108 billion in 2026.
This level of spending is almost unprecedented, representing nearly 100% of the company’s projected annual operating cash flow being funneled back into hardware and facilities.
Where the $108 Billion is Going
The capital is being deployed across three critical pillars of the AI ecosystem:
- The GPU Arms Race: Meta is continuing to buy hundreds of thousands of high-end chips (like Nvidia’s latest architectures) to train its Llama 5 and Llama 6 models.
- Domestic Data Center Build-out: At a recent dinner with President Donald Trump, Zuckerberg reportedly committed to a $600 billion US infrastructure plan through 2028, with 2026 serving as the “peak” year for construction.
- Energy Independence: To power these massive sites, Meta has entered into historic agreements for clean energy, including a 1.2 GW nuclear power campus in Ohio with Oklo Inc., set to begin pre-construction in 2026.
Meta Capex Growth: 2024–2026
| Year | Capital Expenditure (Capex) | Growth vs. Previous Year | Key Focus |
| 2024 | $35–40 Billion | — | Llama 3 & Initial H100 Fleet |
| 2025 | $72 Billion | ~80% | Llama 4 & Data Center Expansion |
| 2026 (Est.) | $108 Billion | ~50% | Superintelligence & Nuclear Energy |
The Strategic Pivot: AI Over the Metaverse
As Meta ramps up its AI spend, it is simultaneously tightening its belt elsewhere.
- Reality Labs Cuts: Meta is reportedly cutting spending on its Metaverse and VR divisions by as much as 30% as it enters 2026.
- Personnel Shifts: Engineering talent is being aggressively reallocated from virtual reality projects to the new internal “Superintelligence Unit.”
- The “Zero Labor” Ad Goal: By late 2026, Meta aims for an AI-native advertising system where a brand can upload a single image, and the AI handles all creative, targeting, and optimization autonomously.
Investor Concerns: The “Money Wall”
Despite the high revenue growth (projected at $235 billion for 2026), some investors are wary. Bank of America analysts warned that hyperscalers like Meta are hitting a “money wall” where the cost of AI infrastructure may soon exceed their ability to fund it purely from cash flow. To bridge the gap, Meta issued $30 billion in bonds in late 2025—the largest such deal in its history.
“We want to make sure we’re not underinvesting… if our strategy on achieving AGI is proven incorrect, we will simply pivot.” — Mark Zuckerberg, Meta CEO
