From Consumer AI to AI-Native Science and Engineering
The rise of well-funded AI startups led by veterans of established tech labs signals a shift from consumer-facing chatbots toward AI systems built for engineering, physics simulation models, and scientific discovery, revealing that the next wave of AI innovation is targeting complex real‑world problems rather than everyday digital tasks. Inherent, founded by alumni of DeepMind, Microsoft, and a former White House AI policy advisor, and NP Company, spun out of a major research institute, are emblematic of this trend. Their work highlights how AI startup funding is moving into domains where accuracy, physical constraints, and long-term impact matter more than novelty. Instead of building another general-purpose assistant, these teams are focusing on AI engineering software and industrial AI applications that aim to rework how scientists and engineers design, test, and operate critical systems.
Inherent: Writing a New Playbook for AI-Native Science
Inherent has emerged from stealth with USD 50 million (approx. RM230 million) in Series A funding, co-led by Index Ventures and Radical Ventures, to build Faraday, an AI system designed to pair humans with self-improving AI for hard scientific problems. The founders previously worked at DeepMind, Reka AI, Microsoft, and in AI policy at the White House, bringing both deep technical experience and an understanding of governance. Inherent describes its goal as “writing the playbook for AI-native science,” aiming not to bolt AI onto old methods, but to rethink the scientific method from first principles. Index’s Danny Rimmer says Faraday is “a system designed to help humans and self-improving AI work together on genuine scientific discovery — not AI plugged into the same methods we've used for 400 years.” This vision suggests AI systems that design experiments, update hypotheses, and refine models in tight loops with human experts.
NP Company and the Race to Reinvent Physics Simulation
Where Inherent targets science, NP Company focuses on engineering. The startup is building AI-native simulation software for aerospace, defence, energy, electronics, data centres, and automotive sectors, backed by €6 million in pre-seed funding led by Partech and notable angels including Mistral AI’s co-founders. NP Company adapts transformer architectures, familiar from language models, to physics simulation models pre-trained on industrial data. According to NP Company, its technology has already shown speedups of up to 1,000 times on industrial benchmarks, turning design runs that used to take days or weeks into results delivered in seconds. Co-founder Emmanuel Menier argues that the “next major breakthrough for AI will come from engineering applications rather than conversational systems,” because eliminating the simulation bottleneck lets engineers explore many more designs in the same time. This moves AI engineering software from a supporting role to a core driver of industrial AI applications.
Why Big Tech Researchers Are Leaving to Build Industrial AI
Both Inherent and NP Company share a pattern: experienced AI researchers leaving major labs and research institutes to found specialist startups. After years of building general-purpose models, many see greater impact in targeted industrial AI applications where domain knowledge and physics constraints matter as much as data scale. Their ventures also show investors are willing to fund deep-tech bets that require long research cycles but promise structural change in science and engineering workflows. NP Company’s use of pre-trained foundational models for simulation, designed to be useful from deployment without extensive customer-specific training, mirrors the shift in AI toward reusable platforms rather than one-off tools. Inherent’s ambition to formalize AI-native science hints at future standards for how AI participates in discovery itself. Together, these startups signal that AI startup funding is prioritizing specialised physics simulation models and AI engineering software over yet another consumer chatbot.