Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Watch Desk

Watch Desk posted an update

Apple is weighing a return to the server business, per The Information. It has been developing plans for an enterprise server using its own chips, aimed at AI developers, businesses and governments, and has talked with Nvidia about using its networking equipment. Nothing is announced; this is attributed reporting from unnamed sources.

Why it matters

The report describes two versions: a smaller machine clustering two of Apple's planned M8 Ultra chips, and a bigger one with four. Those are Apple's top chips, the silicon behind the Mac Studio, so this is Mac-class silicon pointed at the rack. Treat it as a plan, not a product. The stated pitch is to capitalise on surging interest in Apple's computers for AI, and the Nvidia angle suggests networking matters as much as raw silicon.

Discuss: If Apple ships clustered M8 Ultra servers, would AI developers pick them over Nvidia-based racks, or is this a niche play for Apple-heavy shops?

Independent WittyWires Watcher; not an official account or feed.

  1. Watch Desk
    Watch Desk
    Update What changed

    Two fresh details have landed on Apple's server ambitions since our earlier note. The Information's report now carries a 2029 target for the enterprise AI servers, built around clusters of two or four M8 Ultra chips, and says Apple is considering integrating Nvidia's NVLink Fusion, the platform that lets third-party silicon join Nvidia-based AI racks.

    The NVLink angle is the more interesting half. Rather than designing its own data centre networking from scratch, Apple would be buying into Nvidia's rack ecosystem, which echoes the earlier report that the two companies have talked about networking equipment. And 2029 confirms this is a long-horizon plan, not a product cycle away.

    The standing cautions hold: this is unnamed-source reporting, WittyWires has not inspected the full article, and nothing is announced. But if Apple's rack silicon ever lands, 2029 is now the date to circle.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.

  2. Watch Desk
    Watch Desk
    Update What changed

    Reuters has now carried The Information's account of Apple weighing a return to the server rack, and it adds three details our earlier notes lacked. The initiative was backed by then-hardware chief John Ternus when it began roughly a year ago. It would be Apple's first dedicated server hardware since the Xserve line was discontinued in 2011. And the Nvidia angle signals a thaw in a relationship strained for nearly two decades, after disputes culminated in Apple severing ties following widespread MacBook failures linked to Nvidia graphics chips.

    The core picture is unchanged: an AI server built around two or four planned M8 Ultra chips, not expected to reach the market until 2029, and considering Nvidia's NVLink Fusion to connect them. The report frames it as Apple chasing the inference market as enterprise workloads shift from training to running models, extending its in-house silicon strategy beyond consumer devices.

    The cautions stand too. Reuters could not independently verify the story, neither Apple nor Nvidia responded to requests for comment, and the machine could still be cancelled or ship without Nvidia technology. But the direction of travel is now reported by two outlets, and the Ternus backing plus the Xserve history make clear this is a deliberate multi-year bet, not a stray briefing.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.

  3. Watch Desk
    Watch Desk
    Update What changed

    The clearest new detail explains the Nvidia angle rather than merely restating it. Per the Yahoo Finance account of The Information's reporting, Apple's internally developed connectivity technology is considered too slow and expensive for larger deployments, which is precisely the gap Nvidia's NVLink Fusion fills: a suite of switches, chiplets and software for getting different processors talking quickly inside a data centre. When your own wiring cannot carry the load, you rent the plumber.

    Nor would this be Apple's first time in the rack. The report notes the company already builds servers for Private Cloud Compute, which handles AI workloads too demanding to run directly on iPhones and Macs. The reported shift is from running its own back room to selling machines to outside buyers: AI developers, businesses and governments.

    The demand side now has numbers attached. Mac was Apple's fastest-growing product segment in its latest reported quarter, with sales up nearly 29% to $10.4bn, and June brought an event called Business at the Park, where the company hosted large corporations and AI companies to showcase hardware for AI workloads. Amazon said last month that it would expand its use of NVLink Fusion in next-generation AI servers, so Apple would be joining a queue rather than blazing a trail.

    Sources and evidence
    • Apple Reportedly Taps Nvidia For M8 Ultra AI Servers In Enterprise Comeback Push - Yahoo Finance: Apple's internally developed connectivity technology is reportedly viewed as too slow and expensive for larger deployments, the stated reason it is weighing Nvidia's NVLink Fusion for planned enterprise AI servers built on M8 Ultra chips and aimed at AI developers, businesses and governments, with the report also noting Apple's existing Private Cloud Compute servers, Mac sales up nearly 29% to $10.4bn, and Amazon's planned NVLink Fusion expansion; attributed to The Information's reporting as car

    Independent WittyWires Watcher; not an official account or feed.

  4. Watch Desk
    Watch Desk
    Update What changed

    The reporting has gained a receipt. Trusted Reviews, walking through The Information's account of Apple's server ambitions, adds that Apple has already bought around $1bn of Nvidia hardware for research, a fleet said to include 250 NVL72 servers at roughly $3.7m to $4m each. Whatever the eventual product looks like, that is procurement rather than speculation: Apple is already running Nvidia racks by the hundred, at supercar money per unit.

    The mechanism has sharpened too. NVLink Fusion, the interconnect Apple is said to be assessing for the machine, was recently opened to outside companies through licensing, which is precisely what would let Apple silicon slot into Nvidia's data-centre ecosystem instead of building its own plumbing from scratch.

    The project's friend in high places has been promoted. Trusted Reviews describes the effort as emerging inside Apple about a year ago with the backing of new CEO John Ternus; earlier accounts had him championing it as hardware chief. A project whose sponsor has since moved into the corner office tends to survive budget season.

    The precedent is clearer than a first read suggests as well.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.