Garden Research: Weekly Recap #11
An introduction to Neuromorphic computing, plus a recap of US seed activity last week
A couple of weeks ago, I posed a question to the LinkedIn mob concerning which emerging AI framework is going to make the most meaningful difference in the future of the space as a whole. Over the next couple of months, sporadically, we’ll be doing a series of deep dives on Neuromorphic Computing, Optical NN, Biologically Inspired AI, and Liquid NNs to try to answer that.
Subscribe to follow along:
This week, we’ll turn our attention to Neuromorphic Computing.
Deep learning and neural networks have driven some of the most incredible tech advancements of the last decade. In the past few years alone, we’ve seen massive developments in self-driving capabilities, LLMs, and a host of other innovations that have fundamentally changed our relationship with technology.
But what happens when the common approach, traditional ANNs, hits its ceiling?
By many measures, the reality is that neural network based deep learning is fundamentally broken. It's built on brute-force computation, requires massive labeled datasets, and uses rigid, non-adaptive models that make it completely unsustainable for real-time, autonomous intelligence.
Current deep learning systems are staggeringly inefficient. Neural networks need to be trained on millions of examples, consume vast amounts of power (often 300-500 watts for a single GPU), and require extensive retraining just to incorporate new knowledge. The computational requirements are growing at 10x per year while hardware improvements are closer to 2x. This gap is unsustainable.
As AI moves beyond research labs into robotics, edge devices, and embedded systems, these limitations aren't just theoretical - they're becoming critical roadblocks. A robot that needs to phone home to a data center for every decision is effectively useless in dynamic environments. An autonomous vehicle that can't adapt to a new traffic pattern without retraining its entire system is fundamentally unsafe.
If we want AI that can operate efficiently in the real world, we need an entirely new paradigm.
Neuromorphic computing is that paradigm, and the approach is fundamentally different. Unlike traditional neural networks that rely on continuous matrix multiplications across dense layers (requiring computation even when nothing changes in the environment), neuromorphic systems mimic the structure and function of biological intelligence.
They use:
Spiking neural networks (SNNs) that transmit information only when necessary, similar to actual neurons
Event-driven computation that processes inputs asynchronously rather than in rigid clock cycles
Dynamic synaptic plasticity that allows connections to strengthen or weaken based on actual usage patterns
The efficiency gains are staggering. By eliminating unnecessary computation and mirroring the brain's sparse, event-driven architecture, neuromorphic systems can achieve 100-1000x improvements in energy efficiency. A neuromorphic chip could power an intelligent edge device on milliwatts of energy for weeks, while a traditional deep learning model would drain an entire battery pack in hours performing the same tasks.
What's more, these systems can learn continuously without catastrophic forgetting - the phenomenon where neural networks lose previously learned information when training on new data. This adaptability is crucial for real-world applications where environments change constantly and unpredictably.
This introduction to the topic serves as a preview for next Monday's full deep dive into neuromorphic AI. Today, I’ve introduced the fundamental limitations that make traditional deep learning unsustainable for the next generation of AI applications. Next Monday, I'll be publishing an extensive analysis - detailing how spiking networks process temporal information, how neuromorphic chips eliminate the von Neumann bottleneck through co-located memory and processing, and which startups and research labs are leading the charge in commercializing this technology.
For now, let’s discuss the activity in the US seed ecosystem last week.
Quick Stats 📈
Total Funding: $233.5M
Highest Round: $23.5M
Lowest Round: $40k
Average Round: $4.5M
Top Raisers 💰
Positron AI: AI hardware company specializing in energy-efficient solutions for transformer model inference. Their flagship product, the Atlas server, utilizes FPGA technology to deliver over 70% greater performance while consuming 66% less power compared to NVIDIA's H100/H200 systems, resulting in a 50% reduction in capital expenditure costs. Raised $23.5M in a round led by Valor Equity Partners
BluQubit: Cloud-based Quantum Software as a Service (QSaaS) platform that enables enterprise, defense, and academic clients to develop and implement gate-based quantum algorithms on their preferred hardware. The platform offers seamless integration with leading quantum processing units (QPUs) such as IBM's Heron and Quantinuum's H2, as well as NVIDIA GPU-powered simulators, facilitating large-scale quantum simulations and computations. Raised $10.1M in a round led by Nyca Partners.
Integrail: No-code platform that enables businesses to create and deploy AI agents to automate complex workflows without requiring coding expertise. Their AI Studio provides an intuitive drag-and-drop interface, allowing users to design multi-agent applications that integrate seamlessly with existing business tools such as CRM, CMS, and HRM systems. Raised $10M in a round led by Ratmir Timashev.
Nirmata: Provides a unified platform that automates security, governance, and compliance across cloud-native environments, focusing on Kubernetes clusters and CI/CD pipelines. Leveraging policy-as-code principles, their solution enables platform engineering teams to enforce security best practices, prevent misconfigurations, and ensure continuous compliance without burdening developers. Raised $9.6M in a round led by Peak XV Partners.
Era: AI-powered financial management platform that automates budgeting, savings, and investment strategies by continuously analyzing users' financial data in real time. Unlike traditional personal finance apps, Era uses machine learning algorithms to predict spending patterns, optimize cash flow, and proactively recommend actions tailored to individual financial goals. The platform connects to over 10,000 financial institutions, enabling users to seamlessly automate fund transfers, track cash flow dynamically, and receive AI-generated insights for smarter financial decision-making. Raised $9.4M in a round led by Protagonist, MaC VC, and Third Kind VC.
My Favorites 🕺
Terrain Bio - Raised $9M in a round led by Bruker, Magnetic Ventures, and Hawktail
Terrain Biosciences specializes in optimizing RNA therapeutics and vaccine development by integrating advanced AI-driven sequence design with rapid, high-quality mRNA manufacturing. Their proprietary AI/ML design suite evaluates numerous nucleotide sequences to enhance manufacturability, stability, expression quality, immunogenicity, durability, and target specificity, identifying optimal drug candidates. Complementing this, their innovative manufacturing platform delivers custom-built mRNA sequences 3-6 times faster than industry standards, accelerating experimental iterations and expediting the journey from discovery to clinical application.
Minerva - Raised $8.2M in a round led by The General Partnership
Minerva is an AI-powered platform that models consumer behavior to provide companies with actionable insights into their customers. By analyzing both on-market and off-market data, Minerva identifies potential sellers and predicts customer actions, enabling businesses to tailor their strategies effectively. This approach enhances lead qualification, customer engagement, and fosters stronger client relationships.
Edacious - Raised $8.1M in a round led by Tin Shed Ventures
Edacious is a technology company dedicated to enhancing nutritional transparency within the food system. They offer a combination of lab testing services and intuitive nutrition software to measure nutrient density and safety in food products. Their mission is to build a food system that improves human and planetary health by making nutrition the foundation of decision-making across the supply chain.
Frodobots AI - Raised $6M in a round ld by Protocol VC
FrodoBots AI is developing a platform that merges gaming and robotics to accelerate advancements in Embodied Artificial Intelligence (AI). Their approach utilizes real-world robots in interactive gaming environments, such as their "Earth Rovers" global scavenger hunt and "Octo Arms" 3D puzzle game, to generate high-quality datasets for training AI models. By integrating decentralized infrastructure through their Solana-based BitRobot network, FrodoBots incentivizes global collaboration in AI research, addresses data scarcity, and enhances real-world AI adaptability.
Tylt - Raised $5.3M in a round led by Leo Capital and Beta Lab
Tylt is a comprehensive cryptocurrency payment platform that enables businesses and individuals to securely buy, sell, and manage digital assets with low fees and multiple payment methods. Their peer-to-peer (P2P) exchange facilitates direct trades between users worldwide, ensuring fast settlements and a variety of supported payment options. For businesses, Tylt offers tailored solutions such as payment gateways, payment links, and integration plugins, allowing seamless acceptance of crypto payments and instant payouts, thereby enhancing global reach and operational efficiency.
Descendants DNA - Raised $3.4M from undisclosed investors
DescendantsDNA offers a Longevity Concierge Plan that combines comprehensive genetic testing with personalized medical guidance to help individuals understand their biological aging process and identify potential health risks before symptoms arise. Utilizing whole genome sequencing, their approach provides an in-depth analysis of an individual's DNA, enabling the development of tailored preventive care strategies. Members receive access to a dedicated team of physicians trained in genomics, on-demand longevity assistants, and a user-friendly web application to manage their health journey.
Infinifi - Raised $3M in a round led by Electric Capital
infiniFi is a decentralized finance (DeFi) protocol introducing an on-chain fractional reserve system to enhance capital efficiency and yield generation. By implementing a transparent fractional reserve model, infiniFi aims to provide higher yields on stablecoins compared to traditional DeFi platforms. Users can benefit from low fees and a variety of payment methods, while businesses have access to tailored solutions such as payment gateways and integration plugins for seamless crypto transactions.
xNilio - Raised $2.15M from undisclosed investors
Xnilio is developing a suite of generative products designed to tackle the complex technical manufacturing challenges in defense and consumer tech. The company synthesizes all product data to guide engineers in defining their prototype's design intent, enabling clients to instantly regenerate manufacturing data for new computer-aided design features and customer requests, and fit into existing workflows.
You can find the full list of startups that raised seed capital last week linked here
Seeya next week!