Apple plans to reorganize the computing power supply chain through chip mergers and acquisitions to break the bottleneck of AI server chip research and development
The company (Apple) is currently actively planning to acquire semiconductor chip companies in order to overcome its performance bottlenecks in the field of artificial intelligence (AI) server chip research and development and get rid of its competition with competitor Nvidia ( ) Deep dependence on computing power infrastructure. Sources revealed that Apple has been in frequent contact with investment banks and a number of semiconductor start-ups for several months to evaluate potential mergers and acquisitions. Apple has long been extremely restrained in its M&A strategy, preferring small-scale, technology-oriented acquisitions. However, facing the window period of AI computing power dividends, Apple is breaking its traditional investment practices. Behind this strategic adjustment, it reflects the severe technical constraints the company faces in its self-developed computing power supply chain. Currently, Apple’s self-developed M2 Ultra chip deployed within its data center has unsatisfactory performance when processing large, highly complex AI models. When upgrading its smart voice assistant Siri earlier this year, the self-developed server was unable to carry (Google) The large language model Gemini requires huge computing power. Apple was forced to compromise and choose to rent Google cloud services based on Nvidia chips. Since Apple's next-generation self-developed AI server chip, code-named "Baltra", which was originally scheduled to be shipped this year, has encountered development delays, its computing power shortcomings may be difficult to make up for through endogenous R&D in the short term. At the same time, Apple is undergoing major changes in its high-level personnel and financial policies, providing institutional guarantees for the implementation of large-scale mergers and acquisitions. In September this year, John Ternus, Apple's senior vice president of hardware engineering, will succeed Tim Cook as CEO; Johny Srouji, head of chip engineering, will oversee all hardware engineering, including semiconductors. At the financial level, Apple Chief Financial Officer Kevan Parekh made it clear during the second quarter earnings call that the company is gradually abandoning its long-term "net cash neutral" financial policy. Analysts pointed out that this change means that Apple will end its model of burning cash through large-scale repurchases and dividends, and instead retain sufficient cash reserves to provide financial support for multi-billion-dollar semiconductor mergers and acquisitions led by multiple new executives. Historically, Apple's chip design capabilities began with the acquisition of the start-up company PA Semi in 2008, and subsequently achieved great success in the mobile device chip market. However, in the face of the field of high-computing AI server chips that consume large amounts of power and have completely different technical architectures, Apple's existing mobile design experience has hit its ceiling. In order to reshape AI's competitive advantage, Apple has formulated a clear roadmap to catch up with computing power. It is reported that Apple is currently promoting server upgrades based on the M5 Ultra chip and plans to launch the M7 Ultra chip with improved performance in 2029. The memory bandwidth of the latter is expected to reach 1.5 terabytes (TB), which is directly equivalent to Nvidia's Blackwell chip in terms of technical indicators. In addition to seeking external mergers and acquisitions, Apple is also hedging R&D risks by deepening OEM cooperation with industry giants. According to securities regulatory documents submitted by Broadcom last week, Apple has extended its long-term technology partnership with Broadcom until 2031, and the two parties will continue to collaborate in depth on cutting-edge areas such as customized AI chips.