Meta Platforms is reportedly preparing to begin production of its latest in-house artificial intelligence chip, codenamed “Iris,” in September 2026, marking a significant milestone in its effort to reduce dependence on third-party chip suppliers such as Nvidia and AMD. The custom silicon is part of Meta’s long-term Meta Training and Inference Accelerator (MTIA) program, which aims to improve AI performance while lowering infrastructure costs across the company’s products and services.
According to an internal memo cited by Reuters, Iris has successfully completed initial testing and is expected to enter production in partnership with Broadcom, which assisted with the chip’s design, while TSMC will manufacture the processor. Rather than replacing Nvidia GPUs entirely, the new chip will complement Meta’s existing AI infrastructure by handling selected AI workloads more efficiently.
Meta Moves Deeper Into Custom AI Silicon
The Iris chip is designed specifically for Meta’s AI infrastructure and is expected to power recommendation systems, generative AI services, and other machine learning applications across the company’s platforms.
| Chip Details | Information |
|---|---|
| Chip codename | Iris |
| Production start | September 2026 |
| Program | Meta Training and Inference Accelerator (MTIA) |
| Manufacturing partner | TSMC |
| Design partner | Broadcom |
Meta plans to introduce a new generation of AI chips approximately every six months through 2027 as it rapidly expands its computing capabilities.
Why Meta Is Building Its Own AI Chips
Developing custom processors allows Meta to optimize hardware specifically for its AI workloads instead of relying exclusively on general-purpose GPUs.
The company expects the strategy to:
- Lower long-term AI infrastructure costs.
- Improve power efficiency.
- Optimize performance for Meta’s applications.
- Reduce dependence on external chip suppliers.
- Gain greater control over hardware development.
Custom silicon has become an increasingly important competitive advantage for hyperscale technology companies operating massive AI infrastructure.
Nvidia and AMD Remain Key Partners
Despite the launch of Iris, Meta is expected to continue purchasing large volumes of AI GPUs.
The company has indicated the chip will complement rather than replace processors supplied by:
- Nvidia.
- AMD.
High-end GPUs will continue handling many of Meta’s largest AI training workloads, while Iris is expected to improve efficiency for selected inference and production tasks.
Massive AI Infrastructure Expansion
The new chip is part of Meta’s broader plan to dramatically expand its computing capacity.
| Infrastructure Goal | Details |
|---|---|
| Computing capacity target | 14 GW by 2027 |
| Capacity deployed in H1 2026 | 1 GW |
| Expected capacity by end-2026 | 7 GW |
| 2026 AI infrastructure spending | Up to $145 billion |
The company is investing aggressively in data centers, networking equipment, and AI hardware to support future AI products and services.
Industry Trend Toward Custom Silicon
Meta joins a growing list of technology companies developing proprietary AI processors.
Major players investing in custom chips include:
- Google (TPUs).
- Amazon (Trainium and Inferentia).
- Microsoft.
- Apple.
The trend reflects rising demand for AI computing and the increasing cost of relying solely on merchant GPU suppliers.
Challenges Ahead
Building competitive AI processors remains one of the semiconductor industry’s most difficult engineering challenges.
Key hurdles include:
- High development costs.
- Advanced chip manufacturing.
- Software ecosystem development.
- Supply chain complexity.
- Keeping pace with rapidly evolving AI models.
While Iris could improve efficiency for Meta’s internal workloads, Nvidia’s hardware and software ecosystem remains the industry standard for many AI applications.
Outlook
Production of Iris in September would represent a major milestone in Meta’s long-term AI strategy. As the company continues expanding its computing infrastructure, custom silicon is expected to play an increasingly important role in lowering operating costs and optimizing AI performance across its platforms.
Although Meta will continue relying heavily on Nvidia and AMD GPUs, Iris demonstrates the company’s ambition to build a more vertically integrated AI infrastructure that combines proprietary hardware, software, and data center technologies.
What It Means for the AI Industry
Meta’s decision to begin production of its Iris AI chip underscores a broader shift among major technology companies toward designing custom processors for AI workloads. As demand for AI computing continues to surge, companies are seeking greater control over performance, power efficiency, and infrastructure costs by developing chips tailored to their own software ecosystems rather than relying exclusively on third-party hardware.
For the semiconductor industry, Iris signals intensifying competition in AI hardware. While Nvidia is expected to remain the dominant supplier of high-performance GPUs, the growing adoption of in-house accelerators by hyperscalers such as Meta could reshape the AI chip landscape by creating specialized processors optimized for specific workloads and large-scale cloud infrastructure.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
