Microelectronics self-developed AI chips meet bottlenecks! Recommending "Transfer chip" should compete with NVIDIA

According to foreign media The Information, due to problems with self-developed chip design, Microsoft was forced to postpone its original plan and will launch a transitional product in order to meet the fierce competition with NVIDIA.

Amazon and Google have reduced their dependence on NVIDIA chips through their self-developed AI chips. Microsoft is also focusing on developing self-developed chips, but because their business properties are mainly software-based, they face greater challenges. According to reports, the goal of microsoft is to grow on the same side and develop an alternative to NVIDIA, but now it has to adjust the plan to address the problem of design delays.

It is reported that the self-developed AI chip "Braga" of Microelectronics was originally designed to be mass-produced in 2025, but it has been postponed to 2026. Braga is a post-sequence product of the Maia chip released by Microelectrome in April 2024, using a tyre-electrome 5-nanometer process.

According to the latest news, Braga's delay has triggered market concerns about the subsequent version of Braga-R and Clea, worrying that the product competition may not be as good as NVIDIA's latest chips. In this regard, Microsoft will choose a compromise solution to launch a new chip between Braga and Braga-R in 2027, named "Maia 280".

The

report quoted two people familiar with the matter as saying that the Maia 280 design is to connect two Braga chips in series to improve performance. Like Amazon, Microelectronics hopes to challenge NVIDIA in performance-per-watt. According to Microsoft, the performance of this new chip is expected to increase by 30% compared to the chips expected to be launched by NVIDIA in 2027.

It is reported that the purpose of microscopy is to produce hundreds of thousands of self-developed AI chips every year. In addition, with the rising demand for self-developed chips, the orders of chip design companies such as Broadcom and Marvell have increased significantly. Marvell shares closed at 88% in 2024, but have fallen 33% this year, mainly due to the failure to realize the market's high expectations for AI.

Although both Amazon and Google have invested in self-developed AI chips, NVIDIA is still the market overlord, which is difficult to shake. Executive Director Huang Renjun said that the frustration of each self-developed AI chip project is because NVIDIA has unsubstitutable competitive advantages. Although stock prices were once weak, NVIDIA benefited from the explosion of future AI demand and GPUs playing a key role in global computing infrastructure. It has recently rebounded and once again ranked among the top companies with the highest market value.

Extended reading: The United States has lifted restrictions on exports of EDA software to China, and Ximen resumes supply Affiliate to AI "lucky number"? From 1 to 50, guess the numbers at random, and the answers are all the same