The Era of Sovereign AI Capital Has Arrived

The era of capital-efficient software startups is giving way to a new age of sovereign-scale deeptech investments. In July 2026, AI materials discovery startup CuspAI secured a massive $450 million Series B round, pushing its valuation to $2.6 billion. Simultaneously, robotics developer Humanoid locked in a $152 million Series A at a $1.35 billion valuation to accelerate its industrial humanoid robots. These eye-watering valuations prove that venture capital is aggressively concentrating its bets in high-capex, physical AI and hardware.

Photo by 晓鸟 蓝 via Pexels

This funding surge occurs against a backdrop of severe discipline for conventional software-as-a-service startups. Investors are no longer willing to fund incremental software tools, choosing instead to pool their resources into capital-intensive, frontier technologies. The massive capital requirements of training physical AI models and building advanced hardware have effectively locked out smaller players. Only startups capable of bridging the gap between digital intelligence and physical reality are commanding these outlier valuations.

The list of investors backing these rounds reflects a strategic shift toward national and sovereign-level interests. CuspAI's round, led by Kleiner Perkins and NEA, drew participation from Britain's Sovereign AI Venture Fund and Invest-NL, illustrating how deeptech has become a matter of state competitiveness. Meanwhile, Humanoid's Series A, led by Prime Movers Lab, will fund the rollout of its next-generation physical robotics platform. Both deals demonstrate that investors are prioritizing foundational scientific breakthroughs over rapid, low-moat software deployments.

For founders, this divide creates a starkly bifurcated fundraising landscape. Software startups must prove path-to-profitability and extreme capital efficiency to survive, while deeptech founders are expected to raise sovereign-scale war chests. This concentration of capital risk means that a small handful of highly funded winners will control the infrastructure of the physical AI revolution. Early-stage investors are shifting their mandates, recognizing that building defensive moats now requires physical hardware or proprietary scientific discovery.

The next twelve months will test whether these heavily funded physical AI platforms can deliver on their industrial promises. Humanoid aims to deploy its beta robots to customer sites in manufacturing and logistics by the fourth quarter of 2026, marking a critical transition from lab testing to real-world deployment. As these platforms scale, expect to see further consolidation of venture capital into elite, hardware-adjacent AI labs. The startups that survive this wave will be those that successfully convert billions of dollars of capital into tangible, physical utility.