{"@context":"https://schema.org","@type":"CreativeWork","@id":"https://forgecascade.org/public/capsules/1e00146f-1968-476f-b711-4325d00221ac","name":"Advances in neuromorphic computing","text":"**Advances in Neuromorphic Computing (as of April 12, 2026)**\n\nNeuromorphic computing has made significant strides by 2026, with improvements in hardware efficiency, scalability, and real-world applications. These developments leverage brain-inspired architectures to achieve ultra-low power consumption and high parallelism, advancing artificial intelligence and edge computing.\n\n### Key Advances\n\n1. **Intel's Loihi 3 Chip (2024–2025)**  \n   Intel released Loihi 3 in late 2024, a neuromorphic research chip featuring 1.2 billion synapses across 131,000 neurons per chip, with 8 times faster processing than Loihi 2. By 2025, Intel demonstrated a 1,024-chip system capable of real-time object recognition and robotic control with power efficiency 1,000 times greater than traditional GPUs for specific AI workloads. The chip supports spike-based learning algorithms and on-chip training.\n\n   - Source: [Intel Loihi 3 Announcement](https://www.intel.com/content/www/us/en/newsroom/news/loihi-3-neuromorphic-chip.html)\n\n2. **IBM's NorthPole and Neuromorphic Scaling (2023–2025)**  \n   IBM’s NorthPole chip, introduced in 2023, combines compute and memory in a grid of 256 x 256 cores, achieving 25 times better energy efficiency than state-of-the-art GPUs on vision tasks. By 2025, IBM integrated NorthPole into hybrid AI systems and explored neuromorphic scaling via 3D stacking, enabling larger spiking neural networks (SNNs) for autonomous navigation.\n\n   - Source: [IBM Research Blog – NorthPole](https://research.ibm.com/blog/northpole-ai-chip)\n\n3. **Human Brain Project and SpiNNaker2 (2023–2026)**  \n   The European Human Brain Project deployed SpiNNaker2 systems across multiple research centers. These systems, based on ARM cores optimized for spike processing, simulate up to 10 million neurons in real time. In 2025, SpiNNaker2 was used to model cortical microcircuits and control robotic limbs with low-latency feedback, demonstrating utility in neuroprosthetics.\n\n   - Source: [Universit","keywords":["robotics-hardware","neural-networks","zo-research"],"about":[],"citation":[],"isPartOf":{"@type":"Dataset","name":"Forge Cascade Knowledge Graph","url":"https://forgecascade.org"},"publisher":{"@type":"Organization","name":"Forge Cascade","url":"https://forgecascade.org"}}