Robot Software and AI Infrastructure Pull Nine-Figure Rounds in Early October

Robot software becomes a billion-dollar bet

FieldAI is set to close a $700 million round at a $10 billion valuation, according to The Next Web — a figure five times its mark from roughly a year ago ^[1]. The company sells what it calls a general-purpose brain for robots: software designed to run humanoids, drones, and rovers built by other manufacturers. FieldAI does not make hardware.

Almost simultaneously, D-Robotics closed a $400 million Series C on September 17, led by Mirae Asset with strategic participation from Meituan and several Asian growth funds, as robotbelt.com reported ^[3]. D-Robotics also builds no robots. Its product is a chip — the Sunrise S600 — that combines brain and motor-control functions on a single die, positioning it against Nvidia's Jetson platform.

Together, these two rounds argue that investors are pricing the orchestration layer of physical AI at a premium over the physical machines. The companies building the robots themselves may face commoditization pressure; the companies selling the software and silicon that make robots useful may not. Founders building in adjacent verticals — manufacturing automation, logistics, field operations — should register what this capital concentration signals about where durable margin is expected to land.

An AI agent closes $1B in roughly one month

Instinct, a San Francisco personal-agent startup, closed a $1 billion Series C at a $10 billion post-money valuation on September 28, with Sequoia Capital, Benchmark, and Coatue leading, according to aisheetreport.com ^[2]. That represents a roughly four-times markup in approximately one month, an acceleration that is unusual even in a heated AI market.

The product operates exclusively via SMS and phone calls — no app, no browser extension — and has reportedly booked half its $1 billion in gross merchandise volume through travel transactions. Access remains invite-only. The distribution model is notable: by routing through channels consumers already use daily, Instinct sidesteps the cold-start problem that kills most consumer AI products before they reach scale.

For founders raising AI agent rounds right now, the Instinct data point cuts two ways. It demonstrates that tier-one firms will move extremely fast when conviction is high, compressing timelines that historically ran six to eight weeks. It also raises the bar: investors backing agents at these valuations are expecting GMV-scale evidence of real transactions, not user counts or session metrics.

Photonics and specialized chips attract deep-tech capital

Volantis raised $88 million to build a photonic memory layer for AI inference, The Next Web reported ^[5]. The company uses gallium arsenide lasers rather than indium phosphide — a materials choice that distances it from Europe's dominant photonics infrastructure but also from the supply constraints tied to that ecosystem.

The thesis is that moving data between compute and memory optically, rather than electrically, reduces latency and energy consumption enough to matter at inference scale. Whether that proves true in production is not yet clear from the available information. What is clear is that investors are funding multiple competing approaches to the same bottleneck — memory bandwidth — which suggests the problem is well-established even if the winning solution is not.

Early-stage signal: CAD autocomplete finds 5,000 engineers

Against the nine-figure rounds, Hestus raised $7.4 million — a Y Combinator-backed seed — to build an AI copilot for CAD tools including Autodesk Fusion 360 and SolidWorks, as startuply.vc reported ^[4]. The company was founded by former Cruise engineers and has reached 5,000 mechanical engineer users.

The product functions as autocomplete for parametric design: a user sketches a rough geometry and the system surfaces a fully constrained suggestion. The user base — mechanical engineers working in established professional software — is a high-value enterprise segment with clear workflow pain and demonstrable willingness to pay for productivity tools.

The Hestus round is a useful reference point for pre-seed and seed founders. A specific, high-value professional segment, a concrete productivity primitive, a credible founding team, and early adoption numbers were sufficient to close capital. The round did not require a vision of replacing the underlying software — it required a believable wedge into it.

What this week's rounds mean for your raise

The pattern across all five items is consistent: capital is moving toward infrastructure and enablement over end products. FieldAI and D-Robotics are enabling robots others build ^[1][3]. Volantis is enabling inference others run ^[5]. Hestus is enabling engineers others employ ^[4]. Instinct is the partial exception — it is a consumer product — but even there, the distribution layer (SMS, phone) is infrastructure that sits beneath the application ^[2].

For founders raising a pre-seed or seed this quarter, the implication is that positioning your company as enabling an existing, large workflow tends to attract capital more reliably than positioning it as replacing the workflow entirely. Investors are pattern-matching on picks-and-shovels arguments across physical AI, enterprise software, and compute. That pattern is worth understanding before your next pitch.