Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Leaked Images Give First Look at Apple's AI Servers
- Thread starter MacRumors
- Start date
- Sort by reaction score
However, dead humans (other than by incineration) can't generate electrical energy. Sending the matter into living humans helps to generate heat and electrical energy via the body functioning.
IIGS User
macrumors 68000
Happy_John
Suspended
There are more effective ways to generate energy from organic matter than feeding it into humans.
It would have made more sense to grow mushrooms or bacteria, if the aim was to use organic material as an energy source.
The original storyline made more sense - keeping humans alive wasn't about energy - it was about the network their brains made. The matrix simulation wasn't simply a way to keep human docile while their were used as an energy source - the matrix simulation, and th network it ran on, was itself of vital importance to the machines.
But this is VERY off topic...
Happy_John
Suspended
Matrix, definitely. No need for Terminators if you can keep the humans docile by keeping them amused by memes. Skynet didn't factor in the need for billions of consumers.
Yeah same, even when I saw it in theaters when it first came out and I was (obviously) a lot younger, I was like "wait, for energy? really?"
I mean maybe one could say that they feed humans biological matter and use it to make the humans make energy but...
Cold fusion seems to exist...so...
Happy_John
Suspended
Yeah. There’s no way you can get more energy out of a human than you put in, no matter how you limit the calorie burn. Even if you could, there’s a load more ways to generate and store energy far more effectively.
It was the least clever Mac guff in a clever film. I loved the whole using phone lines in the simulation to send data in and out of the simulation…
Last edited:
DonutHands
macrumors 6502
FIRST LOOK! hahah, silly Macrumors.Simply not true; even at this macrumors site. The look of the Private Cloud Compute servers was already covered.
March 6th 2026
Photos of Apple's own new in-house AI servers
Sources: WSJ Reuters https://www.reuters.com/business/apple-begins-shipping-ai-servers-houston-factory-2025-10-23/ No, these are not available for sale to anyone, but with the baseline obsession over every notch in every bezel Apple makes, I'm flummoxed how these week-old photos seem to have...forums.macrumors.com
October 25th 2025
https://www.macrumors.com/2025/10/24/apple-starts-shipping-made-in-america-ai-servers/
What is quite odd about those pictures is that appears to literally be a plain Mn ( M4 , M5 , M6) doesn't particularly matter. There are only two RAM packages attached to a plain M5 SoC. The max RAM capacity is relatively quite limited to be an "AI cloud server". Maybe compared to an iPhone, but every M4+ Mini Pro (and up) and many MBP Pro/Max systems will all have more RAM than this.
If the vast bulk of the traffic is shuffling "Siri AI" requests that are too big for an iPhone off into the cloud this makes some sense. But this seems odd even for that, because will not want to do 1-to-1 client server assignments. Multiple queries need to be handled by a 'server'.
The 32 daughter boards isn't really new. ( four rows of eight appear in the Houston factory tour given to journalists by Apple earlier ).
Same deeply limited external networking to the box exhibited as before also. ( there is a diagram of the lid ).
Limited local memory isn't particularly surprising since PCC OS dumps data after use. Doesn't need a big 'scrach work' area. The only two RAM packages are more substantive.
Which is entirely not surprising if read Apple's published blog on how PCC OS works. Random factory workers cannot mess with server before shipped.
tzm41
macrumors 6502
Apple keeps closing the gap on Nvidia, and they keep doing it while using a heck of a lot less energy. And then you have stuff like https://github.com/drumih/turbo-fieldfare that takes advantage of Apple Silicon's unified memory and on SoC SSD controller to take a 26 Billion parameter MoE AI model and run it in only 2Gb of memory, or roughtly 1/6th to 1/8th of what it would normally need. It is something Apple theorized about publicly, and now people are doing it open source.
What happens if M7 comes, matching Nvidia performance using half the energy and even only 1/4 of the the RAM?
What happens if M7 comes, matching Nvidia performance using half the energy and even only 1/4 of the the RAM?
Yeah I think that was a big part of its success, it was right at the very end of the glory days of 90s computing.
With a 'plain' M5 SoC and just x4 PCI-e lanes for slots? No. That won't fly.
Those aren't really modules. Those cards are entire systems on a board. Each one runs an independent PCC OS instance (which doesn't run a GUI user interface for a single person).
The propriotary edge connectors on those for (power and networking) probably won't 'upgrade' very well. Those cards have zero display or I/O port out. Not going very far as a Mac Pro with no display in most use cases.
Part of the old rumors were that Apple was going to use the Mac stuff as a stopgap unitl then.
December 2024
" ... Apple Intelligence servers are currently powered by the M2 Ultra chip, and they are expected to start using M4 series chips next year..."
https://www.macrumors.com/2024/12/11/apple-intelligence-servers-new-chip-report/
That report is looking pretty bad right now. This card pictured is no where near big enough to hold a M2 Ultra. Apple could have a different set of servers made and deployed besides these being shown to the press ( and now further leaked). That isn't an Ultra.
Apple could have deployed two different server designs. The 'old' one never got any big press coverage.
What they now doesn't look like Ultra capable ( power and cooling).
And there was no Ultra 4. The linked to M4 didn't explictly say M4 Ultra but walking down to M4 Max/Pro would be slightly odd ( unless the expectation was that the model deployed would be getting smaller. For Apple that wouldn't be too surprising, but a good expectation for them to have. ).
"... while Broadcom is said to be providing some networking technology for it. TSMC is expected to begin mass production of the chip in 2026, using its third-generation 3nm process, known as N3P. ..."
If Ultra 3 sank and slowed it would be easy for that '2026' to slide into '2027'. The 'buzz' around skipping M6 Pro/Max to get to much better AI focused M7 generation would be a pretty good candidate to coupled to the "PCC compute server" roll out.
More recently July 2026
" ... Apple's own AI server chip, internally code-named Baltra, had been expected to ship this year but has been delayed, according to people said to be familiar with the project. ..."
Apple Reportedly Looking to Acquire AI Chip Companies
Apple is looking to acquire AI chip companies as part of an effort to reduce its dependence on Nvidia for demanding AI workloads, according to The Information. Apple currently handles some AI processing in its own data centers using its own chips, but relies on Nvidia hardware housed within...
www.macrumors.com
later in that article .
" ... That limitation became apparent during development of the revamped Siri, when engineers reportedly attempted to run Google's Gemini models on Apple's own server infrastructure but found that the chips, designed with Mac workloads in mind, could not handle a model of that scale. As a result, Apple was required to process portions of the new Siri's workload using Nvidia chips within Google's cloud infrastructure. ..."
If there are no Ultra class nodes and these have mainly been plain Mn powered server boards., then missing running a medium-large model wouldn't be hard. There was a big mismatch between how much RAM was needed and how big of server cards they were making.
Maybe getting these M5 nodes because the projected M4 Ultra/Max didn't work.
P.S. No M4 Ultra ( March 2025 ... just 3-4 months after the Dec 11 report. ).
Here's Why Apple is Unlikely to Release an M4 Ultra Chip for Macs
The new Mac Studio is available with mismatched M4 Max and M3 Ultra chip options. Will there be an M4 Ultra chip in the future, though? It looks unlikely. Here are three reasons why Apple might never release an M4 Ultra chip. First, among Apple's line of Mac chips, the highest-end Ultra...
forums.macrumors.com
Is the chip packaging M5 Max sized or M5 Ultra sized? (likely it is neither chip, but some derivative)
Bluestrike2
macrumors member
That's only because you've failed to factor in the transcendental power of the Narrative. The math on human batteries didn't work, but it needed to so, so the Narrative ensured it did. Give John Scalzi's *Redshirts* a read and you'll come to understand the Narrative's abilities.
Happy_John
Suspended
You're assuming we all don't get that already. The whole "you'll come to understand.." is more than a bit patronising.
It's not that stories don't need Mcguffins, it's about doing so while maintaining "suspension of disbelief".
michaeljk
macrumors regular
I don't think I'd spend that much, but I'd probably be willing to fork over $10,000 for a usable AI behind my own firewall.
Each of those nodes is likely a 512gb M5 Ultra at 15k a pop. An entire system will have 16tb of LPDDR5 and if it is ever sold by Apple, likely going to cost $499,000.
Hajj.david
macrumors 6502a
That was the actual original story of Matrix, humans were used as "cpu cores" so to speak.