Meta unveils four brand-new AI processors, intensifying its rivalry with Nvidia and AMD

Meta Unveils New AI Processors

Meta (META) has introduced four advanced AI chips, expanding its Meta Training and Inference Accelerator lineup. This move is part of Meta’s strategy to leverage both commercial GPUs from Nvidia (NVDA) and AMD (AMD), alongside its own hardware, to address growing AI needs and reduce dependence on any single supplier.

Overview of Meta's New Chips

• MTIA 300
• MTIA 400
• MTIA 450
• MTIA 500

Each chip is tailored for specific AI workloads within Meta, ranging from ranking and recommendation models to high-performance inference tasks.

MTIA 400: Generative AI and Scalability

The MTIA 400 is optimized for generative AI and recommendation systems. Meta notes that up to 72 of these chips can be integrated into a single server rack, mirroring the design of Nvidia’s NVL72 and AMD’s Helios racks.

Meta highlights the MTIA 400 as its first processor to deliver both cost efficiency and performance that rivals top commercial products, though it does not specify which competitors are included—likely Nvidia and AMD.

Recently, Meta secured multiyear agreements with both Nvidia and AMD for their chips.

Nvidia CEO Jensen Huang displays a Rubin GPU during a keynote at CES 2026 in Las Vegas. (REUTERS/Steve Marcus)

MTIA 450 and MTIA 500: Enhanced Memory and Speed

The MTIA 450 offers improved high-bandwidth memory compared to its predecessor, while the MTIA 500 further increases memory capacity and speed.

Meta has already started deploying some of these chips, with plans to roll out the rest in 2026 and 2027. All processors share a unified infrastructure, allowing Meta to upgrade seamlessly as needed.

Industry Trends: Custom AI Chips

Meta is not alone in developing proprietary processors for AI. Google (GOOG, GOOGL) and Amazon (AMZN) have long used their own chips for AI training and inference, and Microsoft (MSFT) recently launched its Maia 200 processor.

Google and Amazon also provide chip access to Anthropic for running AI models. According to The Information, Google and Meta have entered a multibillion-dollar agreement for Meta to utilize Google’s processors.

Impact on Nvidia and AMD

This shift toward custom chips could pose challenges for Nvidia and AMD. Nvidia’s CFO, Colette Kress, recently stated that over half of the company’s data center revenue comes from hyperscale clients.

Despite this, Nvidia continues to see revenue growth from other customers.

Major tech companies are not slowing their investments in AI infrastructure. In 2026, Amazon, Google, Meta, and Microsoft are projected to spend a combined $650 billion on capital expenditures, with most funds allocated to AI development.

Stay Updated

Subscribe to Yahoo Finance's Week in Tech newsletter for the latest updates.

See the latest technology news affecting the stock market

Read current financial and business news from Yahoo Finance