Hardware

Nvidia is the default capital partner for hardware

Nvidia appears in 52 deals, far ahead of Blackstone and Samsung (9 each), AMD Ventures and Jane Street (8 each). It co-invested in a $220M Series A at a $1B valuation alongside Lightspeed, Bessemer and Dell Technologies Capital. It was also the sole investor named on a $250M Series E at a $2.3B valuation, and it backed a $250M growth round with Greenoaks, Iconiq and IVP. It even appears in a space-hardware round (Reflex Aerospace). Strategic capital from the compute supplier is now setting the pace of hardware financing, and it reaches well beyond chips.

Inference chips are being pitched as faster and cheaper than training GPUs. Intel launched its GPU D-series for agentic inference without HBM, and argues that agent workloads push the CPU-to-GPU ratio toward 1:1. Tenstorrent, a RISC-V chipmaker, was courted by Intel and Qualcomm but stayed independent. Etched ($10.3B valuation) and a $1.5B Series E at a $6.2B valuation show how big the cheques are. The Series E backers include Trace Capital and Innovation X.

Why it matters · The market is splitting from GPU-centric training toward specialised inference and CPU-heavy agent stacks, which opens room for challengers and puts pressure on incumbents.

Microsoft pulled out of a UK data center deal with Nvidia-backed Nscale because of electricity grid delays, showing power as a real bottleneck. Capital is also flowing to enabling hardware such as Heron Power's solid-state transformers (50GW of orders) and liquid cooling from ZutaCore, backed by Mitsubishi Electric, Carrier and Samsung. Much of the money sits in the late stage, where series_d_plus rounds total $15.8B over 90 days.

Why it matters · Cooling, power electronics and grid equipment are becoming gating factors for AI buildout and attract strategic industrial investors.

A cluster of AI-native tools aims to cut hardware development from weeks to hours. Confetti lets users build gadgets in plain English, and ESP-Mosaico is a board built for agents. Schematik, Diode and Quilter automate firmware and PCB design, and Diode partnered with Anthropic. Own 1, a local-AI personal computer, shows agents also driving new device categories.

Why it matters · Lower design barriers should multiply hardware startups and shift value toward tooling and manufacturing access.

Seed and Series A rounds total about $5.8B over 90 days, with a $17M seed and a $125M Series B backed by Intel Capital and Oxford Science Enterprises. Neural interface interest persists, with a Spark Capital-backed Series A in non-invasive BCI. Physical AI research keeps leaning on Nvidia's GR00T and Cosmos models, which ties robotics demand to Nvidia's stack.

Why it matters · Early investors are underwriting physical AI and sensing hardware at larger cheque sizes, with Nvidia's ecosystem as the common dependency.