Why Intel, IBM And MythWorx Are Betting On 20 Watt AI
getty
The neuromorphic computing market is experiencing significant growth, with revenues expected to increase from $5.1 billion in 2023 to $29.2 billion by 2032, per SNS Insider. Neuromorphic computing is the practice of designing chips and AI systems that mimic the structure of the human brain, using networks of artificial neurons that process information in parallel and consume a fraction of the power of traditional architectures.
The human brain runs on roughly 20 watts, about the same as a dim light bulb, According to IFLScience. It rewires its own pathways as it learns, with no pretraining phase and no massive energy draw. Electricity consumption from AI optimized servers is projected to rise nearly fivefold by 2030, according to Gartner.
The gap between those two realities is where a growing class of companies sees the biggest opportunity in AI.
Neuromorphic computing is moving from research labs into real world deployment. The companies building in this space are making a compelling case that the smartest AI investment may also be the leanest.
MythWorx has built a neuromorphic and generative AI platform that delivers real reasoning at a fraction of the compute, power and cost of significantly larger systems by modeling its architecture after the brain itself.
"We looked at how biological systems actually process information and asked why AI couldn’t work the same way," Jason Williamson, CEO of MythWorx, told me last month at SXSW. "The result is a platform that reasons through problems efficiently, the way the brain does, without brute-forcing its way through trillions of parameters."
Jason Williamson, CEO of MythWorx
Alyssa Brown
The efficiency gains open doors to use cases where cloud-dependent AI simply cannot go. At SXSW, Williamson participated in the Combating Trafficking Accelerator, where he discussed MythWorx's work with the Tim Tebow Foundation to deploy AI in the global fight against human trafficking.
"When you reduce the compute barrier, you put powerful AI in the hands of organizations doing life-saving work," Williamson said. "That is the unlock nobody in the industry is talking about."
MythWorx is far from alone. The neuromorphic computing market is accelerating, with investment, production hardware and enterprise-grade platforms all emerging in parallel.
BrainChip's Akida processor is already in mass production, making it one of the first neuromorphic chips shipping at commercial scale. In 2025, the company launched Akida Cloud, giving developers instant access to neuromorphic computing without specialized hardware. Its IP has been licensed for space-grade processors, taking brain-inspired AI beyond Earth's atmosphere.
Unconventional AI raised $475 million in seed funding this year to build brain-inspired analog computing systems. Founder Balaraman Rao told Bloomberg that the energy problem is the fundamental scaling bottleneck for AI, and that neuromorphic architecture is the way through it.
Even the largest chipmakers are placing bets. Intel’s Hala Point neuromorphic research system, launched in 2024, simulates 1.15 billion neurons and is being tested across robotics, healthcare and IoT applications.
Innovative Laboratory Research Utilizes Advanced Eeg Headset for Cognitive Assessment. Futuristic Lab With Advanced Brain-computer Interface and Neural Signal Analysis in Medicine
getty
IBM's NorthPole chip has demonstrated dramatic efficiency gains in inference workloads by eliminating the traditional separation between memory and processing.
The pattern across these companies points to a question every business leader should be asking before the next AI investment. Does this problem actually require the scale I am paying for?
Neuromorphic systems will coexist with large language models and cloud-based AI, each suited to different workloads. A growing number of high-impact applications, from anti-trafficking operations to autonomous edge devices to space-grade processors, are better served by intelligence that is smaller, more efficient and closer to the problem.
For enterprises watching compute costs climb alongside their carbon footprint, neuromorphic computing may turn out to be the best blueprint the AI industry ever ignored.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。