Meta Platforms is moving forward with manufacturing its custom-designed Iris artificial intelligence chip starting in September 2026, according to an internal memo reviewed by Reuters. The processor represents the first generation of Meta's Training and Inference Accelerators (MTIA) program, a multi-year effort to reduce the company's dependence on third-party chip suppliers and lower its computing costs. Testing of the chip has been completed successfully with no major issues uncovered.

What Is Meta's Iris Chip, and Why Does It Matter?

Iris is designed to power artificial intelligence systems running on Facebook and Instagram, handling both the training and inference phases of AI models. Inference is the process of generating responses from already-trained AI models, while training involves teaching models on vast datasets. Meta's goal is to optimize these workloads using silicon tailored specifically to its own computing needs rather than relying exclusively on general-purpose graphics processing units (GPUs) from Nvidia and Advanced Micro Devices (AMD).

The timing is significant. Meta has increased its 2026 capital expenditure guidance to between $125 billion and $145 billion, up from a prior range of $115 billion to $135 billion. The company aims to boost its AI data center capacity to 14 gigawatts by 2027, with Iris chips playing a central role in that expansion. For context, one gigawatt of computing capacity is enough to power a large-scale AI infrastructure operation serving millions of users simultaneously.

How Is Broadcom Positioned to Benefit From Meta's Custom Chip Strategy?

In April 2026, Broadcom announced a multi-year, multi-generation strategic partnership with Meta to design custom silicon for AI data centers. The initial phase was expected to deploy 1 gigawatt of computing capacity. However, with Meta now ramping up its Iris manufacturing and committing to 14 gigawatts of total capacity by 2027, Broadcom could end up delivering significantly more AI computing infrastructure than originally planned.

Broadcom's involvement extends beyond chip design. Taiwan Semiconductor Manufacturing Company (TSMC) will manufacture the Iris processors, but Broadcom's role in architecting the silicon positions it to capture meaningful revenue as Meta scales production. This partnership could accelerate Broadcom's earnings growth at a critical moment. The company reported a 48 percent year-over-year increase in revenue to $22.2 billion in its fiscal second quarter, with fiscal third quarter guidance pointing to an 84 percent year-over-year increase.

Broadcom's stock has gained just 7 percent in 2026, significantly underperforming the broader semiconductor sector, which has jumped 58 percent. The company trades at 62 times trailing earnings, a valuation that reflects investor skepticism about near-term growth acceleration. However, the Meta partnership could change that narrative. If Broadcom's earnings per share reach $25.85 by fiscal 2028 and the stock trades at 30 times earnings, the price target could reach $775, representing a potential 107 percent gain from current levels.

Will Custom Chips Replace Nvidia and AMD?

Despite Meta's push into custom silicon, the company has explicitly stated that Iris is intended to augment, rather than replace, the large volumes of GPUs it continues to purchase from Nvidia and AMD. Custom chips have primarily been used for inference workloads to date, while GPU training remains the dominant approach for building frontier AI models.

Nvidia stock actually rose about 2.3 percent on the day Reuters reported Meta's Iris manufacturing plans, suggesting investors view custom chips as complementary rather than cannibalistic. Wall Street remains constructive on Nvidia's long-term prospects. Morgan Stanley reiterated an Overweight rating and $288 price target, noting that Nvidia conveyed confidence in an accelerating and increasingly diversified growth story. TD Cowen also reaffirmed a Buy rating and $275 price target, citing strong demand for AI computing infrastructure and constrained compute availability.

Key Factors Driving Meta's Custom Chip Strategy

• Cost Reduction: Custom silicon optimized for Meta's specific workloads can deliver better performance per dollar spent compared to general-purpose GPUs, directly lowering the company's infrastructure expenses.
• Supply Chain Independence: By designing and manufacturing its own chips, Meta reduces reliance on external suppliers and gains more control over its AI infrastructure roadmap and capacity planning.
• Workload Optimization: Iris is architected for the exact computational patterns used in Facebook and Instagram's AI systems, enabling more efficient inference and potentially faster model serving.
• Competitive Positioning: As AI becomes central to Meta's product strategy, owning the silicon layer gives the company a strategic advantage in deploying new AI features faster than competitors.

The broader trend reflects a shift across the technology industry. Major companies including Google, Amazon, and others are increasingly investing in custom silicon to optimize performance and reduce infrastructure costs. This represents a fundamental change in how hyperscalers approach their computing infrastructure, moving from pure reliance on third-party suppliers toward a hybrid model that combines custom chips with continued GPU purchases.

Meta's Iris initiative demonstrates that the future of AI infrastructure will likely involve a mix of specialized custom processors and general-purpose accelerators, rather than a winner-take-all outcome. For Broadcom, this partnership represents a rare opportunity to capture meaningful revenue from one of the world's largest technology companies during a period of explosive AI infrastructure investment. For investors, the question is whether Broadcom can convert this partnership into the earnings acceleration needed to justify its current valuation and unlock significant stock appreciation in the second half of 2026 and beyond.