This week
AI infrastructure is eating everything. Between Groq's $350M Series A, Etched's $700M Series D, and Fractile's $600M Series B, investors are betting billions that the next decade belongs to whoever builds the picks and shovels for AI workloads. Meanwhile, China's Moonshot AI closed a $3.5B Series F and Pragmatik Labs raised $2B — both building on top of that infrastructure layer.
AI Hardware
Groq raised $350M Series A from Disruptive for faster, cheaper LLM inference chips. Etched closed a $700M Series D led by Jane Street to build TSMC-manufactured chips competing directly with Nvidia. Fractile pulled $600M Series B from Lightspeed and Redpoint for edge-optimized inference silicon. If you're building AI infrastructure, the message is clear: specialized hardware for specific workloads beats general-purpose GPUs, and VCs are writing nine-figure checks to prove it.
AI Applications
Moonshot AI's $3.5B Series F (for Kimi Operator, an LLM-powered assistant) and Pragmatik Labs' $2B round (for AI agents that execute long-horizon tasks) show that application-layer companies with real product-market fit can still command mega-rounds. Wispr raised $280M Series B from Menlo for speech-to-text, and Callosum grabbed $100M seed from Atomico to route AI workloads across providers. For founders: the window for horizontal AI tooling is narrowing, but vertical solutions with clear ROI are still getting funded at every stage.
Robotics & Physical AI
Atoms raised $1.7B from a16z to build a universal robotics platform for heavy industry and logistics. This is the biggest robotics round of the year and signals that investors believe the embodied AI moment is here — not in five years, now. If you're building robots that move atoms (not just bits), this is your category's breakout week.