iPhone (2007): Touch interface, GPS, camera, accelerometer → Instagram, Uber, Angry Birds

Steam (2004): Powerful PCs with broadband → complex multiplayer games

Gaming consoles: Custom chips → exclusive AAA games that drove platform adoption

VR headsets: Spatial tracking → immersive experiences impossible on phones

Smart TVs: Living room + remote → streaming apps optimized for 10-foot UI

Current AI "apps" are mostly glorified chat interfaces because they're constrained by cloud API limitations. You get text-in, text-out because that's what works over HTTP requests.

Companies are already recognizing this constraint:

Apple: New Macs ship with 16GB+ RAM standard, M-series chips with NPUs, explicit "AI PC" positioning

OpenAI: Just acquired io ($6.5B) - Jony Ive's AI hardware startup - the largest "acquihire" ever

Microsoft: Heavy investment in "AI PCs" with dedicated NPU requirements

Google: Pushing Gemini Nano for on-device processing

NVIDIA: Massive push into edge AI chips (Jetson, etc.)

But nobody has executed the full platform play yet: Hardware + killer first-party apps + developer ecosystem.

The pattern suggests AI needs local processing hardware to unlock the next generation of startups:

Real-time multimodal experiences (voice + vision + context)

Privacy-preserving personal AI that learns from your data

Instant response times (not 200ms+ cloud round trips)

Rich interactive experiences beyond conversation

Counterarguments:

"Web apps don't need special hardware" → But the most successful app stores do have hardware differentiation

"Current AI apps are making billions" → From early adopters; mass market adoption requires different UX

"Edge AI chips are shipping" → In laptops/enterprise, but no consumer platform has nailed the ecosystem play

The opportunity: The first startups/companies to ship consumer AI hardware with compelling pre-installed experiences, then open to developers.

Think: What would iPhone's app store have looked like if iOS shipped with only Safari?

Hardware investments suggest this isn't profound. The question is who executes the hardware strategy best.