Apple Buys PrismML, Pushes 3-Pillar AI Strategy as $600 Billion Revenue Gap Shadows Data Centers
Updated
Updated · Computerworld · Jul 11
Apple Buys PrismML, Pushes 3-Pillar AI Strategy as $600 Billion Revenue Gap Shadows Data Centers
1 articles · Updated · Computerworld · Jul 11
Summary
Apple is positioning AI around three tiers—on-device processing first, Private Cloud Compute second and third-party servers for the heaviest tasks—while adding PrismML to strengthen model compression.
PrismML recently shrank Alibaba’s 27-billion-parameter Qwen 3.6 model from 54GB to under 4GB and ran it on an iPhone 17 Pro without benchmark losses, underscoring Apple’s bet that more AI can move off data centers.
That shift targets an industry the report says is overbuilt and debt-heavy, with Sequoia’s David Cahn estimating AI infrastructure spending exceeds supporting revenue by about $600 billion a year.
Memory is a key pressure point: vendors are prioritizing high-bandwidth memory for AI servers over consumer electronics, but analysts cited in the report say sector capex growth could stall by mid-2027, risking weaker orders by late 2026.
Apple’s edge-focused approach also aligns with rising privacy, data-sovereignty and trusted-cloud demands, suggesting a more fragmented AI market that could erode today’s centralized incumbents.