Meta (META) plans to start making its Iris AI chip in September, Reuters reports

Meta Platforms is laying out a timeline to begin manufacturing its in-house artificial intelligence chip, code-named Iris, starting in September, according to a report published July 9 by Reuters and carried by Yahoo Finance.

The report frames Iris as part of Meta’s broader effort to build and expand the computing infrastructure used for AI workloads across its platforms. Meta has long relied on a mix of third-party and custom hardware for data center operations, but the move to schedule manufacturing for an in-house chip highlights how central chip design and supply are becoming to its AI strategy.

Meta’s AI infrastructure push is also described as involving a significant scale-up in computing, though the reported item does not provide specific figures in the material available here about how many additional chips, racks, or data center capacity Meta plans to add. The focus, at least in the coverage provided, is on execution timing and on Meta’s willingness to invest in custom hardware to support its AI models.

Iris is one of several custom initiatives Meta has used in recent years to improve performance and efficiency for AI training and inference, the two main phases of AI compute. Training is the intensive process of teaching models from large datasets, while inference is the ongoing step of using those trained models to power features that respond to users. Building chips internally can give a company more control over performance targets and cost structure as compute demand grows.

The Reuters-based report also indicates Meta expects manufacturing to begin in September, suggesting the company is moving beyond the research and development stage and into production planning. For investors and business watchers, that shift is important because it can affect when Meta can translate chip availability into real capacity for AI workloads and product experimentation.

Meta did not provide additional technical specifications or manufacturing partner details in the material available with this prompt. In particular, the reporting excerpt does not specify process node information (such as the semiconductor fabrication technology), expected production volumes, or how Iris compares on paper to any prior-generation chips Meta has deployed.

For context, the artificial intelligence arms race among large technology companies has increasingly become a supply chain and infrastructure competition, not just a software one. Companies that can align chip roadmaps, data center capacity, and model demand may be better positioned to run large AI workloads at lower cost per compute unit and with more predictable performance as they scale.

What to watch next is whether Meta provides more detail on Iris after this initial manufacturing timeline, such as production targets, deployment plans across its data centers, and any milestones tying the chip’s arrival to measurable AI capacity upgrades. Additional disclosure could come through Meta’s investor communications around technology and infrastructure, or through follow-on reporting that clarifies what “begin manufacturing” means operationally for the company’s AI schedule.