Meta Platforms Is Moving Closer to Controlling Its AI Supply Chain. Here’s What the Company Is Planning.

Meta Platforms (META) is moving another step closer to building its AI future on its own technology. According to a recent Reuters report, the company plans to begin production of its custom AI chip named “Iris” beginning in September. The chip is expected to support Meta’s growing AI ambitions as it reportedly aims to double its computing capacity from seven gigawatts this year to 14 gigawatts next year. The company has partnered with Broadcom (AVGO) for its chip design, while Taiwan Semiconductor (TSM) is set to manufacture it. Despite the optimistic goals, the stock fell 3.5% in premarket trading. The decline came after Meta revealed it could spend as much as $145 billion on AI infrastructure this year to double its computing power. One reason for the negative reaction could be the confusion among market participants regarding the company's ambitions. Many have been left confused by the company's decision to sell access capacity while at the same time continuing to increase the same capacity.

The new chip is part of Meta’s in-house Meta Training and Inference Accelerator (MTIA) project. The company launched these MTIA chips in 2023 to gradually rely less on outside suppliers such as Nvidia (NVDA) and Advanced Micro Devices (AMD). The chip’s internal testing took roughly six weeks and did not reveal any major issues. At the same time, Meta has also been strengthening the rest of its AI supply chain, among other initiatives. The company has signed long-term agreements with Samsung for memory chips, Sandisk (SNDK) for flash storage, and Sumitomo Electric for fibre optic components.