TLDR
- Meta stock rose 2% Tuesday before paring gains to 0.75% after Bloomberg reported plans to deploy in-house AI chips in data centers in early 2027.
- The chip, called MTIA 450 or Arke, is Meta’s third-generation custom processor being developed with Broadcom and TSMC.
- A fourth-generation chip, dubbed Astrid, will finish design work in about a month and enter data centers by end of 2027.
- The move aims to reduce reliance on Nvidia and cut energy and infrastructure costs.
- Analysts hold a Strong Buy on META with an average 12-month price target of $758.03, representing 13.5% upside.
Meta Platforms (META) stock moved higher Tuesday morning, rising as much as 2% after Bloomberg reported the company plans to roll out its own custom AI chips in data centers during the first half of next year. The stock later pulled back, trading up around 0.75%.
The chip at the center of the news is the MTIA 450, also known as Arke. It is the third generation in Meta’s custom silicon program, which the company first announced in 2023.
Meta is currently testing the Arke chip. A fourth-generation version, code-named Astrid, is expected to wrap up design work within a month and reach data centers by the end of 2027.
Yee Jiun Song, Meta’s vice president of engineering and head of the custom chip program, told Bloomberg that each new chip generation delivers better performance per watt of energy and per dollar spent compared to off-the-shelf options.
Song also said the chips can run AI models more efficiently than current Nvidia processors, particularly for inference tasks, which is the stage where AI models generate responses in real time.
Reducing Nvidia Dependence
Meta developed the chips in partnership with Broadcom on design and Taiwan Semiconductor Manufacturing (TSMC) on production. The goal is to cut back on the company’s reliance on Nvidia hardware for running AI workloads.
Meta has committed to more than a gigawatt’s worth of its custom chips over a 12-month period. Song pointed out that at that scale, a 30% cost increase is simply not an option, which is part of why the homegrown approach matters.
The first batch of 12 test chips arrived from TSMC on September 1. Performance came in within 2% to 3% of design simulations. On day one, engineers ran Meta’s own AI models as well as tools from DeepSeek and Alibaba.
Inference Over Training
Meta had previously been developing a chip called Olympus, planned for 2028 or 2029, which was designed to handle both model training and inference. The company canceled that project due to cost concerns.
The current chip strategy focuses entirely on inference, keeping costs lower while still meeting Meta’s needs for running AI tools at scale.
Meta Superintelligence Labs also plays a role, sharing upcoming model requirements with the engineering team so chips can be designed around actual use cases.
Tuesday’s gain builds on recent momentum following Meta’s launch of the Muse AI agent.
According to TipRanks, META currently holds a Strong Buy consensus rating based on 44 analyst reviews over the past three months. The breakdown shows 38 Buy ratings and six Hold ratings.
The average 12-month price target stands at $758.03, implying 13.5% upside from current levels.
Stop guessing and start investing with confidence. KnockoutStocks gives you the AI insights, market intelligence, and stock research you need to spot opportunities, cut through the noise, and make smarter investment decisions — all in one powerful platform.
Sign up today and get 50% OFF full access to our premium stock picks.
Simply use coupon code SPECIAL50 at checkout to claim your exclusive discount.
