Apple is investigating how generative artificial intelligence could make the development of its custom processors faster and less complex. The possibility was highlighted by hardware chief Johny Srouji, who described AI-assisted engineering as a potential productivity boost while discussing the company’s long-running silicon strategy.
The comments offer a glimpse into an area of AI that receives less public attention than chatbots or image generators. Modern processors contain billions of transistors and must satisfy demanding targets for performance, power consumption, physical size and manufacturing yield. Engineers already depend on highly specialised electronic design automation software to manage that complexity. Generative systems could help those tools explore more options, automate repetitive work and identify promising designs earlier.
Chip design is becoming harder to manage
Apple’s move into custom silicon began with processors for mobile devices and expanded across watches, tablets, computers and spatial-computing hardware. Each product introduces a different combination of constraints. A phone chip must deliver high performance without draining a small battery, while a computer processor can use more power but must support sustained workloads. A wearable has even tighter limits on heat and energy.
Designing these systems involves far more than arranging individual components. Hardware teams must model how a proposed architecture will behave, verify that it performs correctly and ensure that software can use its capabilities efficiently. A change in one part of the chip can affect timing, temperature, memory access and manufacturing elsewhere.
As processors become more specialised, the number of possible design choices grows. AI could help engineers search that space more efficiently. Instead of replacing expert judgement, a generative tool could propose layouts, prioritise tests or flag conflicts that would otherwise take teams longer to find.
Electronic design software is the practical entry point
Apple relies on electronic design automation platforms from specialist suppliers including Cadence and Synopsys. These tools support essential stages such as simulation, verification, physical layout and power analysis. Both suppliers have been adding AI capabilities to their products, creating a realistic route for Apple to introduce automation without rebuilding the entire design workflow.
The immediate gains are likely to come from repetitive and computationally expensive tasks. An AI system can compare large numbers of alternatives, learn from previous projects and recommend where engineers should focus their attention. Agent-based tools may also coordinate several design steps, prepare tests and track whether a proposed change creates problems elsewhere.
However, processor development leaves little room for unverified output. A mistake discovered after manufacturing begins can be enormously expensive. Any AI recommendation therefore needs strong validation, reproducible results and clear human oversight. Productivity improvements will matter only if they do not weaken the rigorous verification process that makes a chip reliable.
Custom silicon supports Apple’s integrated strategy
Apple’s interest in AI-assisted design fits its wider effort to control the technologies inside its products. Custom processors allow the company to coordinate hardware features with operating systems and applications rather than adapting every product around a general-purpose component.
That approach played a central role in the transition of Mac computers away from Intel processors. Moving an entire product line to Apple Silicon required coordinated hardware and software engineering, but it also gave Apple more control over performance and energy efficiency. The company can now design accelerators for tasks such as photography, media processing, security and machine learning around the needs of its own devices.
AI could shorten the path from an architectural idea to a verified design. Faster iteration would make it easier to test specialised components or adjust a processor for a particular product. It could also help hardware and software teams evaluate trade-offs earlier, before choices become expensive to reverse.
Server processors add another level of complexity
Apple’s chip ambitions are also extending beyond consumer devices. The company has been linked to work on server processors intended to support cloud-based AI features. These systems would handle requests that are too demanding for local hardware while remaining part of an infrastructure designed around Apple’s privacy requirements.
Server chips face different priorities from mobile processors. They must process many workloads efficiently, move large quantities of data and operate reliably at data-centre scale. Generative design tools could help engineers explore those requirements while balancing cost, power and performance.
The combination of local and cloud silicon is strategically important for Apple Intelligence. Simple or sensitive tasks can run on a user’s device, while larger models can use protected server capacity. Owning more of the hardware on both sides gives Apple greater influence over performance, security and operating costs.
Engineers remain responsible for the final decisions
Generative AI will not remove the need for semiconductor expertise. Chip design depends on physics, manufacturing knowledge, architecture and years of practical experience. The most valuable systems will be those that expand what engineers can evaluate rather than presenting opaque answers that cannot be trusted.
Apple’s exploration therefore represents an industrial use of AI rather than a consumer feature. If the technology reduces routine work and improves iteration, it could accelerate the development of processors across the company’s product range. The decisive measure will not be how much of a chip an AI can propose, but whether human teams can use those proposals to deliver better hardware with fewer delays.
