After Google, Meta's demand explodes, will ASIC surpass NVIDIA's GPU next year?

The rise of ASIC servers: Can they shake NVIDIA's dominance?

According to news from the Chasing Wind Trading Desk, a recent research report released by Nomura Securities analyst Anne Lee and her team shows that Meta's ambitions in the AI server field are rapidly heating up. The self-developed ASIC (Application-Specific Integrated Circuit) server project MTIA is expected to achieve a critical breakthrough in 2026, potentially challenging NVIDIA's long-standing market dominance.

The report cites the latest supply chain news stating that Meta plans to launch several high-spec AI ASIC chips between the end of 2025 and 2026, with a total quantity expected to reach 1 to 1.5 million units. Meanwhile, cloud service giants like Google and AWS are also accelerating their self-developed ASIC deployments, and the competitive landscape of the AI server market is quietly changing.

ASIC chips are set to explode, with shipments expected to surpass NVIDIA next year

The report indicates that NVIDIA currently holds over 80% of the value share in the AI server market, while the value share of ASIC AI servers is only 8-11%.

However, from the perspective of shipment volume, the landscape is changing. In 2025, Google's TPU shipment is expected to reach 1.5 to 2 million units, Amazon's AWS Trainium 2 ASIC around 1.4 to 1.5 million units, while NVIDIA's AI GPU supply is over 5 to 6 million units.

Supply chain research shows that the combined shipment volume of Google's and AWS's AI TPU/ASIC has reached 40-60% of NVIDIA's AI GPU shipment volume.

As Meta begins large-scale deployment of self-developed ASIC solutions in 2026 and Microsoft in 2027, the total shipment volume of ASICs is expected to surpass NVIDIA's GPUs at some point in 2026.

Meta's MTIA ambition: Surpassing NVIDIA's Rubin specifications

Meta's MTIA project is one of the most noteworthy cases in the current ASIC wave.

Supply chain data indicates that Meta will launch its first ASIC chip, MTIA T-V1, in the fourth quarter of 2025, designed by Broadcom, featuring a complex motherboard architecture (36-layer high-spec PCB), and utilizing a hybrid cooling technology of liquid and air cooling, with assembly handled by manufacturers including Celestica and Quanta.

By mid-2026, the MTIA T-V1.5 will be further upgraded, with the chip area doubling, exceeding the specifications of NVIDIA's next-generation GPU Rubin, and the computing density closely approaching NVIDIA's GB200 system. The MTIA T-V2 in 2027 may bring larger-scale CoWoS packaging and high-power (170KW) rack design.

However, Meta's ambitions are not without risks.

The report shows that, according to supply chain estimates, Meta hopes to achieve shipments of 1 to 1.5 million ASICs by the end of 2025 to 2026, but the current CoWoS wafer allocation can only support the production of 300,000 to 400,000 units, and capacity bottlenecks may hinder the plan. Not to mention the technical challenges of large-size CoWoS packaging and the time required for system debugging (similar systems from NVIDIA take 6 to 9 months for debugging) If Meta and CSPs like AWS accelerate deployment simultaneously, the high-end materials and components required for AI servers may face shortages, further driving up costs.

NVIDIA's Technological Moat Remains Strong

NVIDIA is clearly not going to sit idly by.

At the COMPUTEX conference in 2025, NVIDIA launched NVLink Fusion technology, opening its proprietary interconnect protocol to allow third-party CPUs or xPUs to seamlessly connect with its AI GPUs. This semi-custom architecture appears to be a compromise, but in reality, it is NVIDIA's strategy to consolidate its market share in cloud AI computing.

The report indicates that data shows NVIDIA still leads in chip computing density (computing power per unit area) and interconnect technology (NVLink), and ASICs are unlikely to catch up with its performance in the short term. Furthermore, NVIDIA's CUDA ecosystem remains the preferred choice for enterprise AI solutions, which is a barrier that ASICs find difficult to replicate.

For investors, NVIDIA's technological moat remains deep, but whether its high-profit model will be forced to adjust under the cost pressures from other CSPs (cloud service providers) is worth ongoing attention