Multiple Applications Drive Expanding Fab Investment, STM32 Devices to Surpass 20nm

1.SEMI: A number of applications to promote the expansion of the scale of investment in 300 mm wafer fabs

According to the report released by the International Semiconductor Industry Association (SEMI), due to the recovery of the memory market and the strong demand for high-performance computing and automotive applications, the global front-end facilities of the 300 mm wafer fab equipment spending will grow by 20% in 2025 to reach 116.5 billion U.S. dollars in 2026 to grow by 12% to reach 130.5 billion U.S. dollars in 2027, will hit a new record high of 137 billion dollars.

By region, China will invest $30 billion a year for the next four years, continuing to lead the fab equipment spending. Taiwan, South Korea and the United States followed. By area, wafer manufacturing to 2027, the compound growth rate will reach 7.6% to reach $79.1 billion, followed by memory, analog, optoelectronics and discrete devices.

2.ST released ultra-low-power controller STM32U0 series

STMicroelectronics (ST) in the 19th release of ultra-low-power microcontroller (MCU) STM32U0 series, the new product will be combined with ST's most advanced design technology and advanced manufacturing processes to achieve a huge leap in energy efficiency. The MCUs can reduce energy consumption by up to 50 percent compared to previous generations.

The STM32U0 features a number of analog peripherals such as ADCs and DACs) operational amplifiers and comparators. There is also an on-chip system oscillator, which helps to reduce material costs and save cost and PCB space. The product is suitable for use in industrial, medical, smart metering and consumer applications. the STM32U0 series is in mass production and starts at $0.68 for 1,000 pieces.

3.ST to introduce sub-20nm process for next-generation MCUs, boosting several key performances

STMicroelectronics (ST) announced on the 19th, the first STM32 microcontroller (MCU) based on the new technology will be sampled to selected customers in the second half of 2024. 18nm FD-SOI process with embedded phase-change memory (ePCM) will realize a leap in performance and power consumption.

This new process technology, co-developed by ST and Samsung Foundry, brings a quantum leap in performance and power consumption for embedded processing applications, while allowing larger memory sizes and higher levels of analog and digital peripheral integration. The first next-generation STM32 microcontrollers based on the new technology will begin sampling to select customers in the second half of 2024, with production scheduled for the second half of 2025.

4. Mature process foundry warming up, order visibility of 2-3 months

According to China Taiwan Business Times reported, mature foundry orders demand visibility of about two to three months, the market legal person and the industry is optimistic that the first quarter of this year will be the bottom of the valley, in which the market is optimistic about the world's advanced in the first quarter can be expected to recognize a long about the injection of revenues, the first quarter of the operation is expected to be better than expected, the second quarter of the revenues and gross profit margins will be more than the first quarter of the slight warming.

Supply chain news pointed out that in February, mature process factory revenue is better than expected, mainly due to consumer demand warming up, cell phone related demand is better than originally expected is the center of gravity. Overall, mature process wafers in the first half of this year, demand is still relatively flat, the market legal estimates, UMC capacity utilization rate in the first quarter of the low 60% -65%, need to wait until the second half of this year, the peak season of consumer products and the total economic environment recovery, can be seen to rise to 80% or so of the level.

5. Micron: this year's HBM capacity has been sold out, the vast majority of next year's reservation

According to TechCrunch Daily 21, Micron Technology CEO Sanjay Mehrotra said on an earnings call Wednesday that AI server demand is driving rapid growth for HBM, DDR5 and data center SSDs. Among them, HBM products are expected to generate hundreds of millions of dollars in turnover in fiscal year 2024, and HBM revenue is expected to contribute positively to Micron's DRAM business as well as overall gross margins from the third quarter of fiscal year 2024.

Mehrotra said Micron's HBM capacity this year has been sold out, and most of the capacity in 2025 has been booked. In addition, it is expected that the supply of DRAM and NAND flash memory industry will be lower than the demand in 2024.

6. Jen-Hsun Huang confirmed that NVIDIA is testing Samsung HBM memory

According to IT home news, this Tuesday in San Jose, California, held a media briefing, NVIDIA CEO Jen-Hsun Huang said, HBM memory is not only difficult to produce, and the cost is very high, in the HBM can be said to be a throwback.

Before this briefing, there is a lot of news on the network that NVIDIA will purchase HBM3 or HBM3E and other memory from Samsung, and in this briefing, Jen-Hsun Huang positively said: "Samsung is a very, very good company". Jen-Hsun Huang said that he has already begun to verify Samsung's HBM memory and is considering placing an order for procurement in the future.

7.Intel: AI PC to promote the popularity of more than 32GB memory

According to fast technology news, 2024 China Flash Memory Market Summit was held in Shenzhen recently, the media reported that Intel China Technology Department General Manager Gao Yu said at the summit, with the AI application scenarios becoming more and more extensive, the demand is getting higher and higher, the memory capacity of the AI PC will definitely be higher and higher, and the popular 16GB is just a starter, and 32GB, 48GB, and even 64GB will be gradually popularized.

According to the latest report from market organization Canalys, global AI PC shipments in 2024 are expected to be 48 million units, accounting for 18% of total PC shipments. It is expected that by 2025, global AI PC shipments will exceed 100 million units, accounting for 40% of total PC shipments; by 2028, it will reach 205 million units.

8. TSMC and Synopsys to use NVIDIA computational lithography platforms

TSMC and Synopsys will use the NVIDIA computational lithography platform in production in order to accelerate the manufacturing of next-generation advanced semiconductor chips and overcome physical constraints, NVIDIA announced at its annual GTC conference in San Jose, Calif.

TSMC, the world's leading foundry, and Synopsys, a leader in silicon-to-system design solutions, have integrated NVIDIA cuLitho into their software, manufacturing processes and systems to accelerate the speed of chip fabrication while speeding up support for the latest generation of future GPUs based on the NVIDIA Blackwell architecture.

Currently, many changes to fab processes require OPC modifications, which increase computation and create bottlenecks in fab development cycles. With the accelerated computation and generative AI provided by cuLitho, these costs and bottlenecks can be mitigated, enabling fabs to design more novel solutions in the development of new technologies at 2nm and beyond.

9. NXP and NVIDIA Partner to Deploy AI Models to Edge Device Portfolios

NXP has announced the ability to enable NVIDIA's trained AI models to be deployed on NXP's broad portfolio of edge processing devices through the eIQ Machine Learning Development Environment, gaining the ability to accelerate development by integrating NVIDIA TAO Toolkit functionality into NXP's eIQ Machine Learning Development Environment. nXP is the first semiconductor vendor to integrate the NVIDIA TAO API directly into an semiconductor vendor into an AI-enabled product, making it easier for developers to deploy trained AI models at the edge.

The NXP eIQ machine learning software development environment allows the use of AI algorithms across NXP's broad portfolio of microcontrollers and microprocessors. It is fully integrated into NXP's MCUXpresso SDK and Yocto Project Linux development environments, making it easy for developers to develop complete system-level applications.