Multiple chip giants showcase their AI ambitions.
On June 12th local time, at the Samsung wafer foundry forum 2024 held in Silicon Valley, USA, AI chips became the core topic of the forum. Samsung optimistically predicts that its AI chip foundry sales this year will be 1.8 times higher than last year, and the number of customers will double compared to last year; By 2028, the number of AI chip OEM customers will increase four times compared to 2023, and OEM sales will increase nine times compared to 2023. "We are currently continuing to receive orders for AI chips," said Cui Shiyong, President and Head of OEM Business at Samsung Electronics.
Samsung has the ability to provide logic, memory chips, and advanced packaging. The company has also released a roadmap for AI chip technology, including two new cutting-edge nodes, SF2Z (2nm process, expected to be mass-produced in 2027) and SF4U (4nm process, expected to be mass-produced in 2025); The preparation work for SF1.4 (1.4nm) is progressing smoothly, and the performance and yield targets are expected to be achieved in mass production by 2027; We plan to mass produce its second generation 3nm process (SF3) GAA processor (Gate All Around) in the second half of this year, and provide GAA processors on the upcoming 2nm process.
Samsung has also launched a Samsung artificial intelligence solution platform that integrates its advantages in OEM, memory, and advanced packaging (AVP) business.
On the same day Samsung outlined its AI blueprint, Broadcom raised its annual revenue forecast for AI chips by 10% and announced a stock split. The company expects revenue related to AI chips to reach $11 billion in 2024, higher than the previous forecast of $10 billion. The news stimulated the stock price of Broadcom to skyrocket by nearly 15% in late trading.
Broadcom is a leading global wired and wireless communication semiconductor company, with core products aimed at helping transmit large amounts of data used in AI applications such as ChatGPT. In the second quarter of this year (ending May 5th), the company's revenue from AI products reached a record breaking $3.1 billion. Botong CEO Hock Tan said, "Botong's performance in the second quarter was once again driven by AI demand and the acquisition of VMware."
The latest news is that Broadcom is customizing AI chips for Google and Meta.
"As the data center market shifts towards AI servers, Broadcom has extremely high potential for growth. In many ways, Broadcom will become the second largest beneficiary of this transformation, second only to Nvidia," said Ben Bajarin, analyst at Creative Strategies.
Samsung and Broadcom are both top global chip manufacturers, with the former focusing on contract manufacturing and storage, while the latter focuses on communication chips. The two companies coincidentally view AI as a performance transmitter, which reflects the tremendous driving force of AI on the chip industry.
The competition for AI chips is becoming increasingly fierce
At present, the layout of AI at home and abroad continues to advance, and the rapid iteration of AI big models will be conducive to the landing of end-to-end AI applications, thereby promoting the development and expansion of the end-to-end AI market. AI chips are expected to be increasingly applied to consumer devices such as smartphones and PCs.
For the chip industry, this means that the speed of product upgrades and iterations is further accelerating, and competition in the AI chip industry is intensifying.
On June 2nd, Nvidia CEO Huang Renxun announced at the ComputeX2024 conference that Nvidia had previously followed a slower two-year update cycle to release chips, and the subsequent product launch speed will be changed to a "one update per year" rhythm to release new AI chip models. Competitors AMD and Intel also unveiled a series of new AI chips at the conference.
Despite fierce competition in the field of AI chips, both AMD and Nvidia do not manufacture chips themselves, but outsource manufacturing to foundries in order to concentrate resources and energy on chip design and technological innovation.
In addition to AMD and Nvidia, OpenAI, Google, and others have also started their own research on AI chips. In the wave of the AI era, the chip industry will face more fierce competition, and sub sectors such as contract manufacturing are also expected to fully benefit.