متا رسماً آستین‌ها رو زده بالا و می‌خواد تا سال ۲۰۲۷ با دو تا چیپ اختصاصی هوش مصنوعی به اسم‌های «آرک» و «استرید»، حسابی تو دیتاسنترهاش گردوخاک راه بندازه

$META TO ROLL OUT TWO NEW IN-HOUSE AI CHIPS IN 2027, PLANS 1GW+ OF CUSTOM SILICON CAPACITY Meta plans to begin deploying its new MTIA 450 chip, code-named Arke, across data centers in the first half of 2027, as it pushes to reduce the cost and energy required to run AI models. The next-generation MTIA 500, code-named Astrid, is expected to finish design work in roughly a month and enter data centers by the end of 2027. Meta expects Astrid to be deployed even more broadly than Arke. Meta is working with Broadcom $AVGO on chip design and TSMC on manufacturing. The company has already committed to more than 1 gigawatt of capacity using its custom chips over a 12-month period and expects deployment to accelerate after that. Meta received its first 12 Arke chips from TSMC on Sept. 1. Early testing showed performance within 2% to 3% of simulations, and the chips were able to run models from Meta, DeepSeek and Alibaba on the first day. The chips are designed primarily as general-purpose inference “workhorses,” rather than for the fastest latency-sensitive workloads. Meta says its close coordination between its AI teams and chip designers allows the hardware to deliver better performance per watt and per dollar for its own workloads than current NVIDIA systems. Meta also canceled a previously planned chip called Olympus, which was intended to handle both AI training and inference, after determining that such a design could cost roughly 30% more. The company is instead concentrating its custom silicon roadmap on lower-cost inference at gigawatt scale. After Astrid, Meta plans to focus future generations on higher speed and throughput, including using fiber-optic technologies to improve performance. Source: Bloomberg