Meta preps rack-scale ASICs with expectations of beating NVIDIA's next-gen Rubin AI GPUs

Meta's first foray into a full-fledged ASIC is coming according to new reports, where it will reportedly compete (and beat) the specifications of NVIDIA's next-gen Rubin AI GPU.

In a new research report by Nomura Securities analyst Anne Lee and her team picked up by insider @Jukanrosleve on X, we're hearing that Meta's ambitions in the AI server business are "rapidly escalating". Meta's new proprietary ASIC server project codenamed MTIA, is expected to hit a significant breakthrough in 2026, where it has the potential to challenge NVIDIA's long-standing market dominance.

Meta is reportedly gearing up to debut with millions of high-performance AI ASICs (1 million to 1.5 million units) between late 2025 and 2026, with cloud service providers like Google and AWS also boosting the deployment of their in-house ASICs.

NVIDIA currently dominates with over 80% of the AI server market, with ASIC AI servers accounting for only 8-11% of the server market, but there are some big changes on teh way. Google's in-house TPU shipments are projected to hit 1.5 million to 2 million units, while Amazon's in-house AWS Trainium 2 ASICs to reach 1.4 million to 1.5 million units, up against the 5-6 million from NVIDIA's vast AI GPU supply.

Once we have Meta kicking off its large-scale deployment of in-house ASIC solutions in 2026, and Microsoft in 2027, we are to expect total ASIC shipments to surpass NVIDIA AI GPU shipments sometime in 2026.

Meta's new in-house MTIA T-V1.5 is expected in mid-2026 with "further upgrades" with the chip area doubling, and exceeding the specifications of NVIDIA's next-generation Rubin AI GPU, and its computing density will "directly approach" NVIDIA's current-gen GB200 system. The next-gen MTIA T-V2 is expected in 2027, bringing larger-scale CoWoS advanced packaging, and higher-power (170KW) rack designs.