Apple’s (AAPL) server plan offers new imaginative space for the AI narrative, with short-term sentiment leaning positive, but whether a 2029 product can carve out a gap in the inference market surrounded by rivals depends on whether Apple can shore up enterprise-grade capabilities, its biggest shortcoming. For investors, rather than chasing rumor-driven rallies, it is better to keep a close eye on three signals: whether data center capital expenditure is included on the balance sheet, whether anchor customers are secured, and the final choice of interconnect technology route.
According to people familiar with the matter cited by tech media outlet The Information, Apple is considering entering the enterprise server market, with plans to adopt in-house M-series Ultra chips and pair them with Nvidia (NVDA) networking equipment. The proposed servers will target AI developers, enterprises, and government clients, with the core focus on enabling customers to run local AI inference on private infrastructure. The product will come in two versions: a lower-spec version integrating two M8 Ultra chips still under development, and a higher-spec version carrying four M8 Ultra chips to form a compute cluster. To solve the problem of high-speed interconnect among multiple chips, Apple has held talks with Nvidia about adopting its NVLink Fusion interconnect solution.
If realized, the move would mark Apple’s return to the server market. From 2002 to 2011, Apple offered the Xserve server, but due to insufficient emphasis on the enterprise market and weak customer support, it was ultimately discontinued in early 2011. The market environment is now vastly different. In Apple’s fiscal 2026 third-quarter earnings report, Mac business sales surged 29% year over year to $10.3 billion, the fastest growth rate across all product lines. Large AI labs such as OpenAI have purchased tens of thousands of Mac minis and Mac Studios for reinforcement learning training, while Anthropic has also rented Mac minis from Amazon Web Services, with robust demand even causing some models to go out of stock.
However, the report noted that the new product will not be available until 2029, and the project may still be canceled. Apple’s new CEO John Ternus expressed support more than a year ago when the project was launched. Affected by the news, Apple shares rose slightly by 0.7% in the latter part of Wednesday morning trading, while Nvidia shares rose more than 2% at one point during the same day. In a commentary on Apple’s 3QFY26 earnings, CICC noted that quarterly revenue of $109.417 billion and net profit attributable to shareholders of $29.789 billion both exceeded expectations, with Mac revenue being the biggest contributor, and raised its target price by 10% to $340, citing “accelerating AI progress and a rising valuation center.”
Several institutions believe the relationship between “long-term water” and “immediate thirst” should be viewed calmly: the 2029 launch window means the project will contribute almost nothing substantive to Apple’s FY2026-27 earnings forecasts and is better seen as a “strategic call option.” The truly key question is whether Apple can sell differentiated private AI compute at a premium, rather than becoming just another buyer purchasing expensive compute.
If cooperation with Nvidia materializes, the market generally believes Nvidia is the more certain beneficiary. Networking equipment accounts for about 10%-15% of total AI data center hardware costs, and NVLink Fusion has already attracted partners such as MediaTek, Marvell, Fujitsu, and Qualcomm. If Apple joins, it would be the platform’s most significant customer to date. Intriguingly, Apple is also a board member of the UALink open interconnect alliance. If it ultimately leans toward Nvidia’s proprietary solution, the industry battle over interconnect standards may face a key turning point. On the risk side, memory chip shortages, gaps in the AI software ecosystem, and the loss of core talent are all unavoidable lessons.
