Apple announced the iPhone 17e today with an upgraded 16-core Neural Engine built for on-device AI workloads. This is Apple continuing to push inference onto the phone itself rather than relying on cloud round-trips. The practical difference for most people will show up in things like real-time photo processing and Siri response times.
DeepSeek V4 is reportedly dropping this week. The new model is said to be multimodal with 1M+ token context, optimized specifically for Huawei chips. Nothing official yet, so take that with the appropriate grain of salt.
In policy news, the US government is banning Anthropic across federal agencies and switching to OpenAI. Treasury, State Department, and the Pentagon are all dropping Claude. Trump ordered the switch in late February following the AI guardrails controversy, where Anthropic refused to let their models be used for certain defense applications.
Nvidia is dumping $4 billion into photonics, splitting it evenly between Lumentum and Coherent at $2B each. The play here is using light-based interconnects to speed up data center AI chip communication. It’s a bet that the bottleneck for AI scaling is increasingly about moving data between chips, not the chips themselves.