SAN FRANCISCO — Meta CEO Mark Zuckerberg said the company faces a critical trade-off over its AI infrastructure investment: whether to keep its compute capacity internal or commercialize a portion of it for outside customers.
Meta plans to spend approximately $60 billion on AI infrastructure by the end of 2026, a figure detailed in its Q1 2026 earnings report. That capital expenditure positions Meta among the largest buyers of high-performance GPUs globally, with Nvidia supplying a significant share. Nvidia shares traded at $190.01, down 3.6 percent on the day.
The internal model focuses on embedding AI across Meta's core products—improving ad targeting and personalization on Facebook and Instagram, tightening content moderation, and powering features for devices like the Ray-Ban Meta smart glasses. It also supports the development of Meta's Llama family of AI models.
Commercializing that capacity would mean offering compute power, large language model access, or other AI services to third-party developers and businesses—putting Meta in direct competition with Amazon Web Services, Microsoft Azure and Google Cloud, all of which already sell AI compute and model access to enterprise clients while using those same systems internally.
Meta has so far kept its most capable infrastructure as an internal advantage. It has released some Llama model versions as open-source, building a developer ecosystem, but has retained its most powerful models and the underlying compute for its own applications.
Meta shares traded at $585.61, down 1.3 percent. Investors are pressing for clarity on how the company intends to generate returns on its capital deployment, and the question is not abstract: the economics of AI compute are punishing. Nvidia's H100 and B200 GPUs are expensive, and the electricity and cooling costs of operating large data centers make utilization discipline critical.
Keeping compute internal strengthens Meta's product differentiation and may improve ad revenue and user engagement over time. Selling it externally could offer a more direct and measurable return on invested capital—but could also dilute the proprietary edge and pull the company into a resource competition with its own potential customers.
