698: And Do What?

What It Covers

ATP episode 698 covers Apple's silicon roadmap disruption — skipping M6 Pro/Max/Ultra chips entirely to fast-track M7 on Intel 18A-P fabrication — alongside the global RAM shortage driving price increases across Apple TV, Xbox, and Steam Machine hardware, and the debut of the redesigned Siri AI in iOS/macOS 26 betas.

Key Questions Answered

• •Apple Silicon Roadmap Disruption: Apple is skipping M6 Pro, Max, and Ultra chips entirely, jumping directly to M7 to accelerate on-device AI capabilities. The base M6 will still ship, but higher-end variants are cancelled. M7 Pro and Max are targeted for late 2027, with M7 Ultra arriving in 2028. Anyone needing a high-end Mac desktop today faces an extended M5 Ultra wait with no faster alternative on the near-term horizon.
• •Intel Foundry as TSMC Alternative: Credible reports indicate Apple has signed an agreement for M7 chips to be fabricated on Intel's 18A-P process node, a refined 1.8-nanometer-class technology combining RibbonFET transistors with backside power delivery. This delivers up to 25% higher performance or 36% lower power versus Intel 3. Apple's motivation is clear: TSMC now prioritizes Nvidia — its largest customer — leaving Apple with reduced allocation leverage and pricing power at the world's leading fab.
• •RAM Shortage Driving Cascading Price Increases: Console storage memory prices have risen over 2.5x and Microsoft projects another doubling by fall 2027. This explains Xbox Series S jumping to $500, Series X reaching $800, Apple TV price increases of 54–67%, and Steam Machine launching at $1,050–$1,428. Products with RAM as their primary cost component — like Apple TV — absorb the largest percentage increases because there are no other expensive components to dilute the impact.
• •Apple's Supplier Negotiation Position Has Weakened: For decades Apple dictated rock-bottom prices to RAM and component suppliers, reportedly driving Micron's gross profits negative during downturns. Now that Nvidia generates higher-margin AI memory demand, suppliers prioritize high-bandwidth memory production over LPDDR chips Apple needs. Apple's historical strength — dictating terms from a position of dominance — fails when it is no longer a supplier's most valuable customer, exposing a negotiation skill gap within Apple's supply chain leadership.
• •New Siri AI Performs at Competitor Parity, With Inconsistency: Testing on macOS 26 beta shows Siri AI can accurately analyze a Perl script's full logic and purpose when routed to server-side models, but returns only generic file-type descriptions for the identical prompt when routed to an on-device model. This inconsistency — with no user visibility into which model runs — makes reliable evaluation difficult. The core capability gap versus ChatGPT and Claude has closed, but output consistency remains unpredictable.
• •Memory Bandwidth as the Key M-Series Performance Metric: The M5 delivers 153 GB/s memory bandwidth, M6 targets 200 GB/s, and M7 is projected at 240 GB/s. This metric directly determines AI inference throughput for on-device processing. Skipping M6 Pro/Max means the next high-end Mac chips jump from M5 Ultra's bandwidth ceiling directly to M7's architecture, which is specifically redesigned around AI workloads. Buyers evaluating current M5 Ultra machines should benchmark against this upcoming bandwidth jump before purchasing.
• •Apple's Siri AI Iteration Cadence Is the Critical Unknown: Competitors like OpenAI and Google continuously update models between major releases. Apple's historical pattern — annual OS-tied updates — conflicts with the fast-moving AI landscape. The Google deal powering parts of Siri AI may lock Apple into a static model snapshot. Apple also maintains a proprietary world knowledge index that requires continuous updates for accuracy. Whether Apple can ship meaningful Siri improvements in point releases like 26.3 or 26.4 will determine competitive relevance by mid-2026.