⚡ Breaking - Nvidia's Jensen Huang Sees 100GW AI Opportunity In Telecom Networks, With TMUS, NOK Already Making Moves

📰 BREAKING

100 gigawatts of AI compute, hiding inside the world's cell towers.

Jensen Huang just pointed Nvidia's spotlight at something the telecom CFOs have been wrestling with quietly for two years: stranded edge infrastructure that could run inference at population scale. T-Mobile and Nokia are already live. The 100GW figure isn't aspiration—it's installed capacity waiting to be re-rated.

Before: Telecom was a pure connectivity play. Carriers laid fiber, lit spectrum, billed for gigabytes, and watched margin compress every quarter. The edge compute sitting in base stations existed to route packets, not run models. Revenue per user flatlined. The AI economy happened over the network, not inside it.

After: The same distributed grid that moves voice and data becomes the inference layer for low-latency AI—autonomous fleets, real-time translation, fraud decisioning that can't afford a cloud round-trip. Carriers monetize installed infrastructure without a new capex cycle. In regulated markets where hyperscaler data center builds stall on power allocation, telecom edges become the only path to local AI sovereignty.

The bridge: This only works if three things align. First, thermal and power envelopes in existing sites must support GPU or NPU retrofit without a forklift upgrade—Nokia's involved because the RAN itself is being redesigned. Second, the carrier has to operate an inference marketplace, not just lease rack space; that's a product capability most don't have. Third, emerging markets with young networks and flexible spectrum policy move faster than legacy footprints in Europe or North America.

Who's building the orchestration layer that lets a Nairobi fintech call a model running on Safaricom's Mombasa tower—and which regulator writes the data residency playbook first?