Podcast Episode · Accidental Tech Podcast · 29 October · 2h 40m

Report claims Apple's is developing a new chip specifically designed for AI servers, with a little help from Nvidia.

Apple is reportedly exploring a return to the enterprise server market with an AI-focused system built around its own silicon and potentially connected using Nvidia technology. In mid-September, _The Information_ reported that Apple is considering servers based on future M8-series processors, with a potential launch around 2029. The project remains preliminary and could change or be canceled. The proposed system would mark Apple’s return to commercial servers, having discontinued the Xserve in 2011. Unlike Apple’s earlier server products, however, the new platform would be aimed primarily at AI inference, giving companies an option for running trained models on premises rather than relying entirely on cloud infrastructure. A key element under consideration is the inclusion of Nvidia’s NVLink Fusion, a combination of hardware and software designed to allow custom processors to connect to Nvidia’s high-speed data center infrastructure. This will give Apple a fast, scalable interconnect for connecting multiple M8 processors to Nvidia GPUs without requiring the company to develop an entirely proprietary high-bandwidth fabric. Apple’s M-series processors are very powerful—they top the best from Intel and AMD in some client benchmarks. They integrate CPU, GPU and other accelerators around a unified-memory architecture. The thing is, for right now anyway, M-series is confined to clients and not servers. Nvidia introduced NVLink Fusion as part of a broader effort to make its interconnect technology available to companies developing custom AI processors. The approach allows third-party silicon to be incorporated into Nvidia-oriented data center architectures, potentially extending Nvidia’s influence beyond its own GPUs. A similar strategy is already being pursued with inference-chip developer d-Matrix. Apple also already operates specialized servers for its Private Cloud Compute system, which handles Apple Intelligence workloads that require processing beyond the user’s device. Scaling that infrastructure, however, reportedly creates bandwidth, cost, and performance challenges. A commercial AI server would allow Apple to extend its silicon strategy into the data center while giving customers direct access to Apple processors rather than Apple’s own cloud services. Analysts doubt Apple will go into the general server business, despite its potential upside, and predict it will focus on the needs of its own Private Cloud Compute. “My guess is they’re only going to build servers for themselves because they need them for their Private Cloud Compute services so they don’t need to satisfy anyone else. That, to me, is what I expect and what makes sense,” said Bob O’Donnell, president and chief analyst with TECHnalysis Research. He added that he “really doubted” Apple would get into business selling servers “because they don’t even have an enterprise sales force to sell the stuff. Almost certainly just going to be for themselves.” And while research firm Gartner projects the server market to be close to $2 trillion by 2030—“an unimaginable number” according to Adrian O’Connell, senior director analyst at Gartner— O’Connell doesn’t expect to see Apple getting a big piece of that market. “If Apple were to introduce a server, then it’s obviously got its own requirements that it could use those for,” O’Connell said. Apple could potentially look to address some sub-segment of the market for providing external services or for selling physical devices, he added. O’Connell said he thinks Apple should be looking at “certain inference use cases that that play into the strength that Apple can offer with its own processors and memory architecture.”

2881ANRワンボタンの声 ダウンロードリンク ■9/26配信 アップル、家族のiPhoneやMacをまとめて補償する「AppleCare」新プランを発表、AppleがXserve以来のサーバー再参入か～M8 Ultra搭載で企業向け、『Mac Duo』 を構築しました iPhone Duo と同じ効果でMacBook 用です ソフトバンク、｢Apple Watch｣で5G通信のまま発着...

2881ANRワンボタンの声 ダウンロードリンク ■9/26配信 アップル、家族のiPhoneやMacをまとめて補償する「AppleCare」新プランを発表、AppleがXserve以来のサーバー再参入か～M8 Ultra搭載で企業向け、『Mac Duo』 を構築しました iPhone Duo と同じ効果でMacBook 用です ソフトバンク、｢Apple Watch｣で5G通信のまま発着...

The chip might matter less than the wiring between chips, and that is where Nvidia comes into play.

Find Computers, phones & IT in #Battersea #SW11, Apple Xserve 2,1 (early 2008) 2.8GHz quad core, 4GB RAM, 80GB HDD – Monterey+, Apple Xserve (Xserve2,1) 2.8GHz, 4GB RAM, 80GB HD – 2008 Rack mountable, 1u, Apple Xserve OS support This device is supported by MacOS X 10.5–10… £95.00

애플이 서버 시장에 다시 뛰어든다는 소식이 전해지면서 업계의 이목이 쏠리고 있습니다. 2011년 Xserve를 단종한 뒤 18년 만의 복귀인데, 단순한 과거 재탕이 아니라 AI 시대의 수요를 정면으로 겨냥한 전략이라는 점이 흥미롭습니다. 심지어 평소 사이가 좋지 않던 엔비디아

## It was just a matter of time. It seems as though Apple just can't get enough of the AI boom, as reports are coming in that the company is researching a reentrance into the server space. Instead of the legacy XServe lineup of servers, these new Apple servers would be centered around AI infrastructure. These would be enterprise-focused servers to build out data centers further. As with all of Apple's recent computers, these servers would be powered by Apple silicon, specifically around the M8 Ultra generation family of chips. According to The Information, Apple is weighing options of using two to four M8 Ultra chips to power the servers. This demand for AI infrastructure from Apple is not a new one. Ever since the rise in AI, the Mac mini and Mac Studio have been exceptionally popular for running local AI agents, containing vast amounts of power in a small, and relatively affordable package. According to The Information, OpenAI and other AI companies have bought tens of thousands of Mac Studios and Mac minis for testing, proving that they are capable of AI. When scaled up, they would definitely have the power that these AI companies are looking for. The most interesting part of these new reports is that Apple is also in talks with NVIDIA to utilize their NVLink Fusion technology to build these servers. This proven technology in the server space is vitally important to Apple's servers' success, as Apple's chip-scaling has not proven to work beyond 2 Max chips. Apple's M2 Ultra. Source: Apple Because the companies are still early in talks, there is not much about what specifically Apple will leverage NVIDIA's technology for. Still, we can infer that it may be about the scaling up of the chips, because of Apple's previous failed attempts at it with the Apple silicon Mac Pro. The Apple silicon Mac Pro was planned to have 2 M2 Ultra chips (4 M2 Max chips), but Apple was not able to get the performance or latency they needed, so this design was scrapped. The worry that may arise from this is the image of AI datacenters currently. Apple adding to the confusing mess that is AI with something so deep as the actual infrastructure that powers it is still a risky move. Hopefully these servers will also not just be locked down to AI and enterprise use, so they can have the sprawling community that the XServe line of Apple servers had. It is less likely that this will happen, but because it is so early, it is not definitive yet. It is also worth mentioning that the Apple server project may never see the light of day, and Apple may scrap the NVIDIA partnership. However, it is clear that Apple is charting a new course with John Ternus in charge. The company is trying things in a different way than before, and it will be interesting to see how things play out.

Curate Feeds | Make Collections | Customize Email Briefs

Apple reportedly to launch AI server with M8 Ultra in 2029, marking a return to the enterprise market since Xserve.

Curate Feeds | Make Collections | Customize Email Briefs

Curate Feeds | Make Collections | Customize Email Briefs

I joined Apple twenty-two years ago to support the Xserve and Xserve RAID - rack-mount servers and storage that Apple sold for a few years before abandoning the product line. What made servers not work for Apple then is still true today: - Apple refused to support older operating systems on the s…

A empresa aposentou sua linha Xserve em 2011, que ...