Perplexity makes it easier to run AI workflows entirely on your own Mac

Perplexity now lets Pro, Max and Enterprise subscribers run its full Computer experience locally on recent Apple Silicon Macs with at least 36GB of RAM. Users can choose Qwen 3.6 or Perplexity’s fine-tuned PPLX Pro model, with the option to authorize access to more than 15 frontier models; local processing is positioned as a way to improve control over sensitive data and reduce credit spend.

Analysis

The direct AAPL read-through is modestly positive for Mac differentiation, but not yet an earnings catalyst: a 36GB-memory threshold confines the feature to a subset of the installed base, and many users may already own eligible machines. The more important second-order effect is strategic. Local execution makes Mac hardware a privacy and inference-cost asset, while Perplexity’s option to escalate selected tasks to cloud models preserves a role for frontier inference rather than eliminating it. That hybrid design could support subscription retention for Perplexity and reduce cloud usage on simple tasks, but the net effect on paid inference demand depends on how often workloads actually stay local.

In the next few days, the launch alone is unlikely to justify an AAPL position. Over 1–3 months, watch adoption, task quality, and whether users upgrade to higher-memory Macs; over 6–18 months, broader local-agent capability could improve Mac value proposition and pressure cloud providers’ unit economics at the margin. The contrarian risk is that privacy positioning overstates practical privacy if users authorize cloud escalation, while local models may be too weak for demanding workflows. Verify which data leaves the device, local task-completion rates, eligible-device penetration, and any evidence of incremental Mac upgrades. The thesis weakens if usage remains niche or Apple fails to translate local AI into hardware mix or retention improvements.

Key Decisions for Investors

• No standalone AAPL trade on this launch: treat it as a small ecosystem-positive signal, not a near-term estimate revision.
• Set an AAPL watch item for evidence that local AI features lift higher-memory Mac sales, upgrade rates, or retention; absent that, do not capitalize the feature as material hardware demand.
• Track Perplexity’s local-versus-cloud task split and escalation frequency. High local completion would imply marginal pressure on cloud inference consumption; frequent escalation would instead support a hybrid-demand thesis.
• Reassess the privacy angle if product documentation or independent testing shows sensitive content is transmitted during cloud escalation or tool execution.