Apple broke free for silicon. Memory could be next
Once upon a time Apple was held hostage by processor manufacturers who seemed unable to give it the chips it needed to optimize its product range – then came Apple Silicon.
Today the company appears to be imprisoned by memory manufacturers, who have turned their back on the needs of their PC market clients to focus their production on the high-end, high-profit memory required by AI.
How could Apple change the situation?
One way might be to leverage its world-class processor development teams and build its own memory production facilities, potentially in the US. That may not be a realistic proposal, but were Apple to decide to get into the memory market, what could it improve, what innovation could it apply, and what might that look like in its products?
Improvement 1: Power and performance
Like so many off-the-shelf components, RAM is mass-produced for the lowest common denominator, which means power and performance ratios can be compromised. Apple Memory could do better, leaning into the company’s design of software and hardware, Apple could make a smart, Apple-specific memory equipped with intelligent power gating. That means it should be capable of managing its own performance, pushing to the limit for high-end tasks and easing-off for the simple stuff. This approach would optimize memory capabilities, deliver even more impressive performance and likely reduce power consumption and boost battery life.
Please subscribe to The Core Apple news Substack and support our work.
Improvement 2: Bandwidth
Apple’s Unified Memory Architecture is a fantastic invention, sharing available memory between the graphics, processor, and neural engine. This delivers drastically reduced latency and boosted power efficiency, even when built using third-party memory modules.
Now imagine what happens if Apple designs support for unified memory directly into the RAM chip itself. It is, after all, already moving in that direction:
Making these kinds of microprocessor level hardware enhancements are and will be precisely where future performance gains in computers will be found as we head toward the limits of die-size. At this level, you can anticipate even more latency reduction and even better power efficiency. That efficiency translates into fast apps, faster devices, and faster and more performance AI. All with better battery life.
Improvement 3: SoC
The way Apple designs its systems already places memory directly on the same silicon package as the CPU and GPU. This is what enables Unified Memory, and it is why CPU, GPU, and the Neural Engine can all share the same memory pool. Memory on Apple devices is not a separate slide in module and is built onto the SoC to unleash more performance, letting the RAM operate at incredible data speeds.
It’s difficult to think how Apple might improve this efficiency were it to make its own memory, but you can imagine higher native throughput or even a scenario in which the memory isn’t just packaged to the same silicon but assimilated within the processor itself.
Could Apple do this?
I think it has already proved it has the capacity to design processors, so the challenge of designing memory (which some argue doesn’t innovate at anything like the same rate as chip design) should be well within its grasp.
What I’m less certain of is if the engineering and raw material skills and supplies exist in sufficient quantities to enable such a plan. Plus the cost of creating a new fab runs into the tens of billions, which Apple can afford, but it’s still a very significant investment.
(I imagine there are plenty of skilled engineers at memory manufacturing companies who would quite like to see the word Apple on their business card, and a salary to match).
Will Apple do this?
Look, if I were sat in CEO Tim Cook’s chair, or if I were settled into the seat currently held by future CEO, John Ternus, I’d certainly be brushing off any existing memory manufacturing contingency plan to see if the creation of an independent, Apple-owned memory production facility would enable the company to build a competitive edge in a market in which pricing is now dictated by memory costs.
There is a strong financial motivation to think about it, given the huge and frankly unsustainable increase in memory costs. It may well be that the huge investment in production facilities and staff that would be required to build such a facility may seem justified, given those price hikes.
I’m not saying Apple will do it, but I reckon it could.
What would it mean for users?
It may mean the cost of Apple devices returns to last week’s normal, which would be nice, if unlikely.
It would certainly mean your Apple devices would become faster, more performant, offer better battery life, and – due to the improved throughput – even more capable of handling on-device AI tasks.
What would it mean for Apple?
Apple didn’t like being constrained by its processor providers. It certainly won’t like being limited by memory manufacturers. Taking control of memory design and production can only enhance its competitive edge. This was less of a priority before, when Apple’s orders were among the biggest in the room, but this has changed with AI.
Today’s reality means memory may well be the next major hardware component to benefit from the company’s acclaimed and almost always successful “whole widget” approach. It won’t solve the current supply problem in the short term, but could form part of a medium term solution.
Please join me on social media at BlueSky, LinkedIn, or Mastodon, even better, please subscribe to The Core for your daily collection of human-curated Apple News.
