Nvidia wants to turn old nuclear power plants into huge AI data centers — but says it draws the line at turning them back on
THE SO WHAT
Repurposing retired nuclear sites for AI data centers is a land, grid, and cooling play — not just a PR move about clean power. If you depend on large-scale training or inference, start mapping access to brownfield industrial sites with heavy grid interconnects as a strategic asset.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIClaudes new smart reports tell bosses when workers are off-topic
Usage-level reporting on Claude turns your AI assistant into a de facto productivity telemetry layer. Before you roll this out, decide explicitly whether it’s a coaching tool or a surveillance tool—because employees will assume the latter unless you prove otherwise.
Applied AITencent’s Cloud Boss Plays Long Game to Patch ‘Leaky’ AI Ship
Tencent publicly calling its AI effort a “leaky ship” and then reorganizing around it signals a multi-year rebuild, not a quick feature sprint. If you’re betting on Chinese AI platforms, expect more vertical focus and internal consolidation before you see aggressive global moves.
Applied AITencent Cloud Chief Says AI Demand Justifies Capex
Tencent shifting capex toward internal model training and enterprise AI over consumer agents mirrors a broader pattern—hyperscalers are prioritizing owned IP and B2B stickiness. For enterprise buyers in Asia, this likely means deeper vertical solutions but tighter coupling to Tencent’s stack.
Applied AIGoogle is rolling out Gemini 4 Carbon to selected cybersecurity defenders through Fairwind at $2/1M input and $10/1M output tokens, rising to $4 and $20 later
Frontier models are getting enterprise-style intro pricing and tight vertical gating—$2/$10 per million tokens for select cyber defenders now, doubling later, is a classic land-and-expand move. If you're a security leader, this is the window to negotiate access and pricing before these capabilities normalize at a higher cost basis.