Technology Sector

Google released three new AI models on Tuesday—Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber—designed to undercut rivals on price and speed. Think of it like a restaurant cutting portions and simplifying the menu to beat competitors on cost: 3.6 Flash uses 17% fewer computing tokens (the basic units AI charges by) than its predecessor, while Flash-Lite handles high-volume, routine tasks at rock-bottom prices. The move shows Google betting it can win through efficiency rather than raw power, but it also exposes a gap: the company's flagship model, Gemini 3.5 Pro, still hasn't shipped months after competitors OpenAI and Anthropic rolled out their top-tier versions, signaling internal struggles to meet performance targets (CNBC).

Nvidia unveiled detailed specs for Vera, its new server CPU (the brain of a data center), marking its first major chip outside graphics processors and GPU territory. Vera chips have already shipped to OpenAI, Anthropic, and SpaceX, and Nvidia claims 50% better performance for AI agents than chips from Intel and AMD. The pitch: AI agents—software that runs on its own in the background—need beefy CPUs to stay fed with data, so Nvidia is selling the whole stack (GPU plus CPU plus cooling) as one vertically integrated system rather than letting customers mix-and-match competitors' parts (CNBC).

Intel announced Tuesday that cybersecurity firm Fortinet will manufacture its next-gen security chip using Intel's fabrication services—a symbolic win for new CEO Lip-Bu Tan, who took over in March 2025. Think of a foundry as a factory that custom-builds chips for other companies; Intel has been desperate to land marquee clients to justify its massive spending on U.S. factory buildouts. Fortinet's chip will use Intel 4, older technology suited for specialized (not cutting-edge) processors, but it's tangible proof Intel can still convert customers despite years of manufacturing fumbles (CNBC).