Meta stock has over 50% upside, but investors may be missing this catalyst

Evercore ISI says Meta could monetize its AI infrastructure investment by selling excess computing capacity externally. Meta is targeting 14 GW of total compute capacity by 2027 as it expands data centers and builds proprietary AI infrastructure, which could create a new revenue stream if implemented.

Analysis

The market should frame this less as a near-term revenue line and more as optionality on Meta’s balance sheet: if external compute sales become real, capex starts to look like a partially monetizable asset base rather than pure reinvestment. That can support a higher long-duration multiple, but only after investors see evidence of utilization, pricing discipline, and customer stickiness; until then, the stock is still hostage to depreciation and capex optics.

Second-order winners are the accelerator ecosystem and any customer that needs scarce compute capacity: NVDA and AMD benefit if Meta keeps pulling through GPUs regardless of eventual end-market use, while enterprise AI buyers gain an additional supply source that could ease pricing pressure. The more interesting loser is the “scarcity premium” cloud trade — if Meta ever commercializes excess capacity at scale, it could compress the premium embedded in AI-cloud names such as ORCL and, at the margin, reduce the urgency premium for managed-infrastructure vendors.

The key risk is execution, not demand. Externalizing compute requires enterprise-grade security, SLAs, networking, and support; that is a different business from training models for internal use. Near term, the catalyst is not revenue but commentary on capex intensity and utilization over the next 1-3 quarters; over 6-18 months, the thesis only works if Meta proves it can turn incremental depreciation into high-margin third-party revenue without slowing product innovation.

Contrarian view: consensus may be overestimating how transferable Meta’s infrastructure is to outside customers. The more likely base case is that most of this capacity remains a strategic reserve, with only a small portion monetized opportunistically. If so, the upside to earnings is real but delayed, while the immediate market reaction could be too optimistic relative to the actual revenue bridge.

Key Decisions for Investors

• Buy META on any post-capex pullback; treat this as a 6-12 month optionality trade, and reduce if management raises capex faster than it can show utilization or third-party revenue proof points.
• Pair trade: long META / short ORCL over the next 1-3 months if the market starts capitalizing Meta’s compute as a new supply source; thesis breaks if Oracle keeps winning large AI-cloud commitments despite incremental supply.
• Stay long NVDA on the view that Meta’s buildout preserves accelerator demand even if some capacity is later externalized; reassess if Meta signals slower GPU procurement or a lower 2027 capacity target.
• Set a watch item, not a trade, on DLR/EQIX and other infrastructure landlords: if Meta meaningfully commercializes capacity, those names could see multiple pressure from a lower scarcity premium, but the signal is too early for a hard short.