Garden Research: Weekly Recap #12
A deep (deep...) dive on neuromorphic computing, plus a recap of the seed activity from last week.
Hey Everyone,
Happy Thursday. For those who are new here, every week I deliver a recap of the activity in the US seed ecosystem, a full list of every completed round, and a deep dive into a company I find interesting. If you’re interested in following along, click below.
Artificial intelligence has achieved incredible milestones over the last decade, largely driven by the exponential scaling of deep learning and neural networks. From self-driving cars to ChatGPT, these systems have reshaped how we interact with technology.
But beneath the surface, deep learning remains fundamentally constrained by inefficiencies.
Traditional neural networks are incredibly wasteful. They learn through brute-force approaches, requiring millions of training examples, vast amounts of power, and extensive retraining just to incorporate new knowledge. Neural networks process every pixel of every image independently, even when most contain no useful information. They can't adapt in real-time or learn without extensive retraining.
As AI moves beyond research labs into robotics, edge devices, and embedded systems, these limitations are becoming impossible to ignore.
Enter Neuromorphic Computing
Neuromorphic systems mimic the actual structure and function of biological intelligence. They use spiking neural networks that process information only when necessary, store knowledge within their network, and learn continuously without retraining.
When you stare at a still object, your brain doesn't waste energy continuously reprocessing its features. The same principle applies in a neuromorphic system—neurons only fire when the input signal changes in a meaningful way.
This approach enables AI to respond in real time, adapt to new inputs as they occur, and operate with a fraction of the energy used by deep learning models. A neuromorphic chip could power an intelligent system on milliwatts of energy, while a GPU-based AI model would drain an entire battery pack performing the same task.
A Fundamental Architectural Shift
Neuromorphic chips integrate memory and computation into a single system, eliminating the von Neumann bottleneck that plagues traditional computing architectures.
Intel's Loihi 2 chip packs 1 million artificial neurons and 120 million synapses, with each synapse storing its own weight value. This eliminates the need to shuttle data back and forth between memory and computation units, drastically cutting energy use and latency.
The most revolutionary aspect: neuromorphic chips can learn directly on the hardware itself. While traditional AI models remain frozen until retrained on massive datasets, neuromorphic systems adapt continuously through experience—working much more like our brains do.
If deep learning is the first draft of artificial intelligence, neuromorphic computing is the rewrite—a step closer to truly brain-like AI.
Read the full deep dive for a (very) deep dive and long list of startups building solutions in this young, but potentially transformative space.
Last week, 62 companies raised Seed capital in the U.S. 🌱
Quick Stats 📈
Total Funding: $356.8M
Highest Round: $33M
Lowest Round: $50k
Average Round: $6.1M
Top Raisers 💰
Achira: Pioneering a novel approach to drug discovery by integrating AI with physics-based simulations. Their platform develops atomistic foundational simulation models that combine geometric deep learning, quantum chemistry, and statistical mechanics to accurately simulate molecular interactions. Raised $33M in a round led by Compound, Nvidia, and Dimension.
Ahead Computing: Specializes in developing and licensing high-performance 64-bit RISC-V CPU intellectual property (IP) cores. Their processors are designed to enhance per-core performance, addressing challenges in AI computing related to bandwidth and data processing limitations. Raised $21.5M in a round led by Eclipse Venture.
Cygnus Finance: DeFi industry’s first Instagram-like app layeri, making decentralized finance more accessible and user-friendly. By bridging non-EVM blockchains like The Open Network (TON) with Ethereum-based ecosystems, it allows users to stake, earn rewards, and access real-world assets in a visually intuitive, social-driven interface. Raised $20M in a round led by OKX Ventures and Optimism.
Karman+: Focused on mining near-Earth asteroids to provide sustainable, carbon-neutral energy and resources for space and Earth applications. Their upcoming mission, 'High Frontier,' aims to extract asteroid regolith, which can be utilized for propulsion, manufacturing, and construction materials in orbit. Raised $20M in a round led by Hummingbird Ventures and Plural Platform.
Valar Atomics: Pioneering the development of small modular reactors (SMRs) to produce carbon-neutral synthetic fuels and provide off-grid energy solutions for high-demand industries. By constructing "gigasites" housing numerous high-temperature gas reactors, the company aims to generate affordable hydrogen and clean hydrocarbon fuels, as well as supply power directly to data centers and heavy industrial facilities. Raised $19M in a round led by Riot Ventures.
My Favorites 🕺
Subsense - Raised $17M in a round led by Golden Falcon Capital
Subsense is developing brain-computer interfaces (BCIs) that use optically powered, ultra-miniaturized implants to enable direct communication between the brain and external devices. By leveraging energy-efficient, light-based neural interfaces, their technology eliminates the need for bulky external components, making BCIs more practical for long-term use. This approach aims to revolutionize assistive neurotechnology and cognitive augmentation, offering a more seamless, low-intrusion pathway for controlling digital environments through thought alone.
Foundation Alloy - Raised $16.9M in a round led by Material Impact Fund and Engine Ventures
Foundation Alloy is pioneering advanced metal alloys designed for high-performance manufacturing, leveraging AI-driven material design and novel processing techniques to create stronger, more heat-resistant, and customizable metals. Their approach integrates additive manufacturing and advanced metallurgy, enabling faster production cycles, reduced waste, and enhanced performance across industries like aerospace, defense, and industrial applications. By optimizing material properties at the atomic level, Foundation Alloy is reshaping how high-performance metals are designed and produced, unlocking new possibilities for next-generation manufacturing.
Vema Hydrogen - Raised $13M in a round led by Propeller and Extantia Capital
Vema is pioneering Stimulated Geologic Hydrogen, a method that accelerates natural subsurface reactions to produce large-scale, carbon-free hydrogen at costs below $1 per kilogram. By injecting catalysts and stimulating specific geological formations, Vema aims to provide a sustainable and economically viable hydrogen source, potentially supplying energy needs for thousands of years. This approach offers a scalable solution to global decarbonization efforts, leveraging abundant natural resources to meet the growing demand for clean energy.
OpenInfer - Raised $8M from Cota Capital, Essence Venture Capital, and Operator Stack
OpenInfer develops an AI inference engine that enables high-performance, on-device processing of large models with minimal memory usage. Their technology offers seamless integration with popular frameworks like LangChain and LlamaIndex, allowing developers to enhance AI capabilities on edge devices without extensive code modifications. By focusing on efficient, local computation, OpenInfer aims to provide scalable, cost-effective, and private AI solutions across various hardware platforms.
Level Zero Health - Raised $6.95M in a round led by Redalpine
Level Zero Health is pioneering AI-driven solutions for hormonal health, focusing on early detection and personalized intervention for hormone-related conditions. Their platform analyzes real-time biomarkers and physiological data to predict hormonal imbalances before they lead to chronic health issues. By integrating AI with proactive monitoring, Level Zero Health aims to transform hormone management from reactive treatment to preventative, data-driven care.
Everstar - Raised $4M in a deal led by Third Prime
Everstar is revolutionizing the nuclear energy sector by integrating advanced artificial intelligence to streamline and accelerate regulatory compliance and licensing processes. Their flagship AI platform, Gordian, automates complex workflows, significantly reducing approval timelines and associated costs while maintaining rigorous safety standards. By partnering with industry leaders and engaging proactively with regulatory bodies, Everstar aims to make nuclear power deployment faster, safer, and more efficient, addressing the growing global demand for clean energy.
You can find the full list of startups that raised seed capital last week linked here
Seeya next week!