Anthropic, Akamai Strike $12 Billion AI Computing Deal
THE SO WHAT
A $11.6B, seven-year compute commit to Akamai is another data point that frontier labs are locking in vertically integrated, non-hyperscaler capacity. If you’re planning large-scale training or inference, treat bandwidth, colo, and edge footprint as strategic levers, not just AWS line items.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIMicrosoft is quietly killing its Copilot+ PC brand
AI PC as a marketing label is fading—AI as a baseline hardware capability remains. For IT and OEMs, this means less time chasing badges and more time standardizing around NPU/RAM/storage thresholds that matter for local inference performance.
Applied AIOpenAI hack on Australian government reveals anxiety at heart of global artificial intelligence dilemma
The reported OpenAI-linked Medicare incident underscores that AI vendors are now entangled with core state infrastructure—breach fallout is no longer just a vendor risk, it's a national systems risk. Operators handling health, ID, or payments data should assume AI providers will be pulled into the same regulatory and diplomatic blast radius as traditional critical infrastructure.
Applied AINSW crackdown on ‘AI-doctored’ real estate listings hits mural-covered wall
When a real mural gets flagged as AI fakery, you see the enforcement problem—governments can ban deceptive AI images, but they can't cheaply verify them at scale. Real estate, travel, and marketplaces should expect more friction and documentation requirements around imagery, and build that into listing workflows now.
Applied AITechCrunch Disrupt 2026: Ricursive Intelligence’s Anna Goldie and Azalia Mirhoseini on when AI starts designing its own hardware
AI-in-the-loop chip design compresses the iteration cycle between models and silicon—whoever closes that loop fastest gets compounding performance and cost advantages. If you're building on custom hardware, start asking vendors how much of their design stack is already AI-assisted and how quickly they can retarget for your workloads.