Apple is reportedly skipping its M6 Pro, Max, and Ultra chips entirely and fast-tracking an AI-focused M7 line (codenamed Andros) for 2027 — a silicon roadmap pivot that collides head-on with Nvidia’s Kyber rack delay. Two of the biggest companies in technology are solving the same problem — the AI compute bottleneck — from opposite ends, and converging on the same timeline: 2027-2028.
🔍 THE BOTTOM LINE
Let’s be clear up front: we’re reading the same publicly reported rumor stacks you are, and speculating on what they mean together. We have no inside knowledge of Apple’s roadmap, Nvidia’s manufacturing timelines, or anything else. What follows is connecting two public reports — one from SemiAnalysis about Nvidia’s Kyber delay, and one about Apple’s alleged M6-to-M7 pivot — and asking what they might mean if both are roughly accurate. Also, Apple is currently suing a tech YouTuber for allegedly leaking unreleased software via a FaceTime screen recording, so the company’s appetite for people speculating about its unannounced products is, let’s say, legally vigorous. With that disclaimer firmly in place, let’s connect some dots.
The Rumoured M7 Pivot
According to reported roadmap details circulating in tech media: Apple will release only a base-level M6 chip for entry-level Macs, skipping the M6 Pro, Max, and Ultra variants that were expected to power its premium computers. Instead, the company is fast-tracking the M7 line (codenamed Andros) to debut in 2027.
The M7 architecture reportedly focuses heavily on local AI processing, boosting memory bandwidth to roughly 240 GB/s (compared to an estimated 200 GB/s in the M6) and upgrading neural accelerators. The reported timeline: base M7 chips in the first half of 2027, Pro and Max versions later in 2027, and the M7 Ultra in 2028.
High-end Mac users would reportedly rely on the upcoming M5 Ultra (expected as soon as this year) as a stopgap until the M7 Pro and Max arrive.
We cannot independently verify any of these roadmap details. They are rumours and reports, not confirmed Apple announcements. Apple has not publicly commented on the M7 roadmap or the alleged M6 skip.
What Connects to the Kyber Story
The connection to Nvidia’s Kyber delay is the timeline. Both stories converge on 2027-2028:
- Nvidia’s Kyber NVL144 — the rack-scale architecture for its 2027 Rubin Ultra chips — has been delayed to 2028 because a single PCB midplane can’t be manufactured at scale, according to SemiAnalysis. The backup plan was cancelled after hyperscalers rejected it.
- Apple’s M7 Ultra — the top-end chip in the rumoured new lineup — is reportedly pegged for 2028.
If both reports are roughly accurate, 2028 is the year where two fundamentally different approaches to AI compute arrive at the same finish line — one from the data centre, one from your desk.
The editorial tension is real: Nvidia is trying to build bigger. Apple is trying to build smarter. Nvidia’s problem is that the manufacturing can’t keep up with the design. Apple’s reported decision to skip an entire chip generation suggests someone looked at the AI landscape and decided the incremental M6 path wasn’t worth shipping — that the only way to catch up in the AI arms race was a purpose-built architecture.
Scale-Up vs Scale-Down
Nvidia’s entire 2027 thesis depends on rack-scale architecture — 144 GPUs in one cabinet, connected through a midplane, functioning as a single machine. When that midplane can’t be manufactured, the thesis stalls. The scale-up story breaks.
Apple’s reported M7 thesis is the inverse: push 240 GB/s memory bandwidth and upgraded neural engines into a chip that sits on your desk or in your lap. No rack. No midplane. No hyperscaler procurement contract. The AI workload runs locally, on-device, without depending on a cloud provider’s rack roadmap staying on schedule.
These aren’t competing products — Apple isn’t selling rack-scale compute to hyperscalers. But they are competing philosophies about where AI compute should happen. And if the rumoured timelines hold, the year Nvidia’s rack-scale story is supposed to arrive (2028) is the same year Apple’s top-end AI-focused chip (M7 Ultra) is expected to land.
The Hardware Boss Angle
The reported M6 skip is the kind of decision that comes from someone who understands silicon roadmaps at a manufacturing level. Skipping a whole chip generation is expensive — it leaves money on the table in the M6 Pro/Max/Ultra tier and bets the premium Mac lineup on an unproven M7 architecture arriving on time.
This is a manufacturing-first call. It’s the kind of decision that makes more sense if the person making it looked at the AI arms race, saw how far behind Apple is in on-device AI capability, and concluded that incremental improvements to the M6 line wouldn’t close the gap — that only a purpose-built AI chip architecture would.
We’re speculating here. We don’t know who inside Apple pushed for this pivot or what the internal debate looked like. But the pattern — skip the safe incremental generation, jump to the purpose-built one — is a hardware person’s bet, not a software person’s bet.
What About the Lawsuit?
In a moment of timing that approaches self-parody, Apple is reportedly suing tech YouTuber Jon Prosser and Michael Ramacciotti over the alleged theft of trade secrets related to unreleased iOS 26 features. Apple claims the defendants coordinated to unlawfully access an Apple software engineer’s development iPhone by recording its screen during a FaceTime call. Prosser reportedly admitted seeing and recording the unreleased software during the call. The allegations have not been proven in court.
So here we are, writing an article speculating about Apple’s unannounced chip roadmap based on publicly reported rumours, while Apple is actively suing someone for allegedly leaking unannounced products. To be clear: everything in this article comes from publicly available reports and rumour aggregators, not from any internal Apple source. We are not Jon Prosser. We don’t have a FaceTime call. We’re just reading the same news and connecting dots.
If Apple’s legal team is reading this: hi. The dots are public. We’re just drawing lines between them.
NZ Angle
New Zealand has no domestic chip manufacturing and no Apple silicon design centre. But every Mac, iPad, and iPhone in the country runs on Apple silicon, and the reported M7 pivot — if real — means the next generation of NZ creative professionals, developers, and small businesses running local AI workflows will be doing so on chips explicitly designed for on-device AI. That could matter more here than in markets with cheaper cloud compute: NZ’s distance from major data centres already means higher latency and higher cloud GPU costs. A chip that runs a capable AI model locally, without phoning home to a Virginia data centre, is quietly an infrastructure story for a country at the bottom of the Pacific.
❓ FAQ
Is the M7 confirmed? No. Everything in this article about the M7 roadmap comes from publicly reported rumours and tech media speculation. Apple has not announced or confirmed the M7, the M6 skip, or the Andros codename.
Would an M7 chip compete with Nvidia’s GPUs? Not directly. Apple’s chips are designed for consumer and professional devices (Macs, iPads). Nvidia’s Rubin Ultra and Kyber racks are designed for data centre AI training at hyperscaler scale. The overlap is philosophical — both are responses to the AI compute bottleneck — not a product-level competition.
Why would Apple skip a whole chip generation? We’re speculating, but the reported reasoning is that the AI arms race made incremental M6 improvements insufficient. If on-device AI capability is the new battleground, a purpose-built AI chip architecture may have been deemed worth the risk of skipping the M6 Pro/Max/Ultra tier.
Should I wait for an M7 Mac or buy an M5 Ultra now? That depends on your workload. If you need maximum performance this year, the reported M5 Ultra stopgap would be the play. If you can wait until 2027-2028 and on-device AI matters to you, the M7 line is the rumoured bet. But remember — these are rumours, not confirmed products. Apple’s actual roadmap may differ from what’s been reported.
🔍 THE BOTTOM LINE
Two public reports, one about Nvidia’s manufacturing wall and one about Apple’s alleged silicon pivot, point at the same 2028 horizon from opposite directions. Nvidia can’t build the rack. Apple — if the rumours are right — is betting you won’t need one. Whether either bet pays off is a 2028 question. For now, both companies are responding to the same pressure: the AI compute bottleneck is real, and the old playbook — incremental improvements on an annual cadence — isn’t fast enough anymore.
📰 Sources
- SemiAnalysis (Nvidia Kyber delay reporting)
- Reported Apple M7 roadmap details (tech media speculation)
- Apple v. Prosser lawsuit (reported, allegations unproven)
