Why build AI data centers in a cornfield when Saudi Arabia has cheap oil and cheaper power?

Meta shares jumped >7% after a Bloomberg report said it is building a business to sell excess AI compute to outside customers, potentially competing with AWS/Azure/Google Cloud. However, MNTN CEO Mark Douglas argues the economics of U.S. AI data centers are deteriorating and that Gulf sovereign-wealth-backed capacity (e.g., Saudi PIF) could offer lower prices, making U.S. hyperscaling less attractive. He is skeptical the cloud/computing pivot fits Meta’s core ad-led business, while remaining bullish long-term on Meta’s AI/open-source models (Llama) for improved ad targeting and efficiencies.

Analysis

The market is paying for option value, not a durable new earnings stream. A compute-rental pivot at META would be a capital-intensive side business with weak differentiation versus the hyperscalers, so any revenue that shows up is likely to be low-margin and cyclical unless management can prove contracted load and very high utilization. The first-order pop is therefore more likely to fade once investors focus on depreciation, power, and customer concentration rather than the narrative of "new TAM."

The more important second-order effect is on infrastructure economics: sovereign-backed Gulf capacity can undercut U.S. buildouts on power and permitting, which matters most for colocation, power developers, and land-bank owners rather than for the large cloud platforms themselves. But the moat still sits with latency-sensitive, regulated, and integrated workloads; those are not easily shipped offshore, so the real displacement risk is in bursty AI training and commodity compute, not enterprise cloud cores.

Contrarian take: the bull case on META remains ad efficiency from better models, not cloud adjacency. If AI lowers customer acquisition costs and improves targeting, META can still expand ad margins even if the compute business never becomes material. The thesis breaks if management turns the tactical move into a capex escalation without visible external backlog or ad ROI; that would be a 1-2 earnings-cycle problem, not a 3-5 year story.

Key Decisions for Investors

• Fade META strength on a 1-3 month horizon: use a defined-risk put spread or short into any further 5% relative outperformance vs. QQQ. Risk/reward is attractive because the market is pricing narrative optionality before any evidence of contracted demand.
• Do not chase AMZN/GOOGL/MSFT on this headline alone; their franchises remain better positioned if compute commoditizes. Use them as quality longs only on broader AI/cloud pullbacks, not as immediate beneficiaries of Meta's entry.
• Pair trade idea: long MSFT / short DLR or EQIX for a 3-6 month view that lower-cost offshore capacity and weaker U.S. land/power economics compress colocation returns faster than they pressure hyperscaler software-like margins.
• Watch META's next earnings for capex guidance and any disclosed customer backlog; if capex rises more than ~10% without external revenue detail, treat the cloud story as a headline trade and add to the short.
• Speculative alert only: MNTN may get incremental sentiment lift if AI-driven ad targeting improves across the ecosystem, but wait for actual customer spend acceleration before taking risk.