Meta Targets 14 GW AI Capacity with Custom Chip to Gain Competitive Edge

Meta Platforms is accelerating preparations to produce its in‑house artificial‑intelligence chip Iris. The company plans to deploy 7 GW of compute power in 2026 and expand to 14 GW by 2027, partnering with Broadcom for design and TSMC for manufacturing.

Iris: Meta’s Custom Silicon Play

■Iris is designed to complement, not replace, Nvidia and AMD GPUs.

■The goal is cost control for content recommendation, ad tools, and AI features.

■Funding comes from Meta’s robust advertising engine, reducing investment risk.

Gigawatt Expansion: 2026‑27 Roadmap

■7 GW of capacity will kick off in 2026.

■Capacity will double to 14 GW in 2027.

■The aim is to lower energy consumption and inference costs across data centers.

Competitive Advantage and Economic Impact

■Meta’s massive user base and ad infrastructure could be strengthened if custom silicon cuts scale‑up costs.

■The chip may reduce dependency on external GPU suppliers, improving long‑term margins.

■Signals a similar strategy for other tech giants.

Risks and Investment Framework

■Ongoing infrastructure spending remains significant; sustainability is key.

■Chip performance may take time to hit expected benchmarks.

■Nvidia and AMD still hold price and performance advantages in the competitive landscape.

Meta’s advertising engine continues to fund chip and capacity investments, raising risk tolerance for large‑scale AI projects and potentially delivering high upside for investors.