TSMC, the world’s leading semiconductor manufacturer, is projected to generate a staggering $31 billion in revenue from its advanced 3nm and 5nm process nodes in just nine months. This impressive feat is driven by the insatiable demand from major tech giants like Apple, AMD, NVIDIA, Intel, Qualcomm, and MediaTek, who are all vying for TSMC’s cutting-edge chip production capabilities.
NVIDIA has successfully run its new Blackwell AI GPUs in real-time within its data centers, dispelling rumors of design issues and confirming its production schedule. The company will reveal more details about the Blackwell architecture at Hot Chips next week, and it will be available to select customers in the fourth quarter of 2024.
Samsung Electronics is set to receive a state-of-the-art High-NA EUV lithography machine from ASML, marking a significant step in its pursuit of semiconductor leadership. The machine will be operational in the first half of 2025, enabling Samsung to compete with industry giants like TSMC and Intel in advanced chip manufacturing.
SoftBank, owner of Arm Holdings, has abandoned its plan to partner with Intel to build an AI chip, instead turning to TSMC. The move is a setback for Intel and the US CHIPS Act, as SoftBank was seeking US government funding for the project. The decision was driven by Intel’s inability to meet SoftBank’s demands for speed and volume.
Nvidia’s stock is trading higher after reports surfaced that SoftBank is turning to TSMC for AI chip production instead of Intel. This decision comes after Intel failed to meet SoftBank’s demands for volume and speed, potentially due to its financial struggles. While TSMC’s high demand may pose challenges, SoftBank remains dedicated to its AI ambitions and is exploring partnerships with Google and Meta for expertise and resources. Despite its recent selloff, Nvidia’s valuation remains attractive amidst the growing AI frenzy.
Google is ditching Samsung for its upcoming Tensor G5 processor, opting for TSMC’s advanced 3nm process node. This move aims to give Google an edge in the AI smartphone market by integrating hardware and software seamlessly, similar to Apple’s iPhone strategy. The Tensor G5 will be manufactured using TSMC’s cutting-edge 3nm process, placing it at the forefront of mobile processors alongside Apple’s A18 chip.
Intel’s next-gen desktop platform, Nova Lake, is set to revolutionize the CPU landscape with a groundbreaking architecture and TSMC’s cutting-edge 2nm process node. The upcoming Nova Lake-S CPUs will succeed the Arrow Lake series, offering significant performance improvements and a possible chance for Intel to regain its lead over AMD. However, the release of these powerful CPUs is still years away, with an estimated launch in 2027.
Intel is reportedly expanding its outsourcing of chip production to TSMC, including its new 3nm process node. The company is also outsourcing backend design and advanced packaging to Taiwanese companies like Egis, Shichixin, and KYEC, indicating a greater reliance on TSMC’s expertise as Intel struggles with its own advanced processes.
NVIDIA’s upcoming Blackwell AI GPU family is encountering significant production challenges, primarily due to issues with the Blackwell GPU architecture and TSMC’s new CoWoS-L packaging. The problems have led to delays in production, with the launch of the flagship Blackwell B200 GPU pushed back to late 2024 and a new, less advanced variant, the B200A, expected to arrive in 2025.
Intel is actively recruiting senior engineers from TSMC, even as it outsources its own next-generation processors to the Taiwanese giant. This move signals a fierce competition brewing in the semiconductor industry, with Intel aiming to challenge TSMC’s dominance.
