Hardware

Quantum computing enters large-scale Series A funding regime

Oratomic, founded in 2026, raised a $300M Series A to run Shor's algorithm at practical scale — a round that would have been considered Series C territory just two years ago. A concurrent $300M Series A backed by Arch Venture Partners, Spark Capital, Khosla Ventures, Bezos Expeditions, General Catalyst, and Bain Capital further confirms that deep-tech hardware is attracting growth-stage capital at early-stage valuations. This compression of the funding ladder — seed-scale conviction at nine-figure check sizes — represents a structural shift in how capital markets are pricing cryptographically relevant quantum timelines. The implications extend beyond hardware: if Shor's algorithm becomes practically executable, encryption infrastructure across finance and defense faces an acute reset.

Despite losing roughly $1 trillion in market cap in under two months (signal [10]), Nvidia remains the most active hardware investor with 28 deals in the last 28 days — co-leading a $2.5B Series C alongside Sequoia, Lightspeed, JPMorgan, and B Capital at a $27.5B valuation (signal [32]), and participating in an $800M Series C with General Catalyst and Vista Equity (signal [45]), and a $200M Seed alongside Kleiner Perkins and a16z (signal [47]). Nvidia's Isaac Gym simulation platform and RTX 5090 GPUs are simultaneously becoming the de facto compute backbone for physical AI training, running 62,000 parallel environments (signal [9]). Trading at ~40x trailing earnings versus Cisco's 150-180x at the telecom bubble peak (signal [5]), Nvidia's valuation appears structurally supported by both its chip sales and its expanding portfolio strategy.

Why it matters · Nvidia's dual role as chipmaker and lead investor means portfolio companies get both capital and preferential compute access, creating a moat that pure-play rivals cannot easily replicate.

Broadcom's simultaneous $30B chip deal with Apple and custom inference chip partnership with OpenAI (signals [19, 25]) illustrate how hyperscalers are systematically reducing Nvidia dependence through bespoke silicon — while acknowledging that manufacturing concentration at TSMC and peers creates a different chokepoint (signal [24]). Etched and Fractile are building inference chips with weights baked directly into hardware for maximum speed, while Tenstorrent — having raised over $1.8B — is now in takeover talks with both Intel and Qualcomm. The vertical integration imperative cited by a16z ('people who love software need to build their own hardware,' signal [27]) is driving not just chip design but full-stack ownership from model to metal.

Why it matters · Teams that control their own silicon can dramatically undercut cloud inference costs, but the barriers to advanced-node manufacturing mean only the largest players can fully close the loop.

Diode and Quilter are both attacking the PCB design bottleneck from different angles — Diode using Claude as an AI electrical engineer to compress hardware timelines from months to weeks, while Quilter applies reinforcement learning trained on real-world physics to automate full PCB layout. Schematik, backed by $4.6M pre-seed from Lightspeed, extends this to natural-language hardware project generation for non-experts. Sipeed's NanoKVM-Go (107 Product Hunt votes, signal [26]) adds another layer: giving AI agents hardware-level screen visibility and input control via open MCP server architecture, closing the loop between AI software agents and physical device control.

Why it matters · If hardware design timelines compress from months to days, the barrier to building custom silicon and embedded products collapses, unleashing a long tail of specialized hardware startups.

The a16z Show's framing of humanoid robotics as 'Tesla versus China, mirroring EV market dynamics' (signal [0]) is substantiated by Generalist AI's body-agnostic pre-training paradigm pushing task success rates from 50% to 90% (signal [1]), and by BYD's vertical integration playbook being explicitly invoked as the model for Chinese robotics challengers (signal [2]). Tesla's vertical integration, however, is identified as a structural constraint: no third-party OEM will buy robotics technology from a direct automotive competitor (signal [7]), potentially ceding the supplier ecosystem to Chinese specialists. Standard Botsand Neuro Robotics represent the Western challenger tier, while NVIDIA's Isaac Gym (signal [9]) and RTX 5090 training infrastructure underpin both sides.

Why it matters · The robotics supply-chain bifurcation mirrors the EV dynamic: Western incumbents risk losing the high-volume manufacturing tier to Chinese competitors who can leverage vertical integration at scale.