My earlier reply was to the title of this thread. I had not read the longer thread from which this topic had been branched off. Now realizing that the point was to ask how Apple core competence can apply to AI, I can comment more directly.
As others have noted, Apple keeps very quiet about new technology nearing release.
One area that has not been so secret is Apple’s push to debug macOS, including the “soon to be discarded” (I don’t think so) variant that runs on Intel processors.
A related push has been to debug functions associated with the “secure enclave,” initially provided by the T1 and T2 chips. These complex operations have recently been at least partially figured out by Howard Oakley. Security features provided by the T2 far beyond the obvious face or fingerprint recognition include, among others, validation of code before it is allowed to run, and of RAM before it is allowed to be used. The complexity of validating security is why so many functions have been constrained to the startup disk. Tightening of T2 security explains some of the many problems users have experienced with “external” storage devices, including mounted on PCIe cards as well as USB attached. Other problems have been due to plain old negligent Quality Assurance in debugging, a management issue which also seems much improved. A recent benefit is improved security in the just-released A19 and A19 Pro chips.
Howard says “the T2 has … a Public Key Accelerator for signing and encryption/decryption using RSA and ECC methods.” This suggests variants of the Secure Enclave will apply ECC methods to memory, perhaps beyond the memory on a single M-series chip.
Apple’s in-house experience with the need to use ECC in complex calculations goes back to 2003 and the Big Mac, a surprisingly powerful and inexpensive supercomputer produced at Virginia Tech by a cluster of PowerMac G5s. The calculations came out wrong until the PowerMacs were replaced with Xserve G5 servers, which used ECC RAM.
Apple’s core competence by which it dominates is developing unique enabling technologies, incorporating the technology into products which disrupt markets, then efficiently manufacturing and selling the product. Using unique enabling technologies means Apple marketing competes by the fact that Apple products are different. Technology enabled by M5, M6 and later M series Apple Silicon likely will have the potential to disrupt the AI market.
Apple Silicon processors provide technology example where expanding innovations have enabled a long series of products and market disruptions. Assembling recent bits and pieces of news suggests Apple will continue pursuing this core competence. Hearing rumors some time ago, but nothing recently, may suggest an M-series server is immanent. Maybe not, but it would fit the pattern of Apple arriving late, then leaping to the forefront.
What have we heard?
The M series includes an increasingly powerful Security Enclave. This seems likely to enable more secure use of more RAM.
Increasing complexity in successive M generations leads to lower manufacturing yields, as well as increasing difficulty in design improvement. Within each M series generation, processing power has been increased by making the chip larger and more complex. This development, manufacturing, and marketing combination results in the top of the line of one series level reaching market at about the same time as the lowest level of the next M series. Buyers must choose between last year’s design with all the bells and whistles, or the new state of the art, just out of the door and far from fully enabled. Manufacturing which has been ramping up and refining last year’s M series, now starts over with the new M series.
Meanwhile, a different approach to processors has been developed (by AMD and others?), in which separately produced “chiplets” are assembled. Adopting this design and manufacturing approach could significant affect availability of M5, and later, Apple Silicon.
Greatly increased rate of production of Apple Silicon would be needed to meet the potential high demand for a market disrupting M5 (or M6?) Apple Silicon server. Matching manufacturing and product availability to demand for a market disrupting new product is an Apple core competence. News we have seen possibly related to preparations is Apple increasing its manufacturing capacity.
An Apple Computer Data Center (ACDC) server cluster would be distinguished from existing products by its much lower energy consumption. Lower consumption of energy and other resources are Apple core competencies. Mother Nature is watching, and she has not been very happy with a lot of recent server operations. Neither have utility customers, who are balking at rate increases and resource consumption for the benefit of data center owners. ACDC alternatives would be favored by both environmental and business communities.
Apple is focused on its core competencies. What I see holding Apple back is the difficulty of expanding manufacturing capacity sufficiently to meet future demand.
Wouldn’t it be great if somehow political support increased for expanded manufacturing capacity? 