Why Banyan AI Fund I Invested in SpiNNcloud

The Shift: From Dense to Sparse Computing

The next generation of GenAI models isn’t dense—it’s sparse. Instead of activating every layer or weight, sparse models compute only what’s needed, dynamically skipping irrelevant parts. This shift enables massive energy savings and faster inference—but only if the hardware is designed to support it.

GPUs were built for dense workloads. They waste energy and compute power on zeros. Dynamic sparsity requires a new kind of silicon.

SpiNNcloud’s Edge: Purpose-Built for Sparse AI

SpiNNcloud builds brain-inspired neuromorphic hardware from the ground up to run sparse GenAI workloads with unprecedented energy efficiency.

• 78x more efficient than NVIDIA H100
• 7x more efficient than Groq’s LPU (344.1 vs. 44.6 Token/s/W)
• Event-based processing that mirrors how the brain activates only what’s relevant
• Dedicated sparse compute architecture—not repurposed GPU logic

Not Just Theory — Already Deployed

“We have reached the commercialization of neuromorphic supercomputers ahead of other companies.”
— Hector Gonzalez, Co-CEO of SpiNNcloud BBC

‍

• SpiNNcloud has built the world’s largest brain-inspired supercomputer
• 5 million cores, 34,000 chips, 720 boards, 8 racks
• Used by top institutions across Europe and the U.S., including the U.S. Department of Energy
  
  ‍

Strategic Recognition

BBC Technology recently profiled SpiNNcloud as a potential “NVIDIA killer,” highlighting its commercial lead in neuromorphic computing:

“In May, SpiNNcloud Systems announced it will begin selling neuromorphic supercomputers…
Companies like Intel and IBM are still in R&D—SpiNNcloud is shipping hardware.”
— Zoe Corbyn, BBC Technology Reporter

The Market Opportunity

The AI inference market is projected to exceed $72.6B by 2030. SpiNNcloud is one of the few companies not just chasing that market—But re-architecting how it will be won with a diverse customer base and a trajectory to reach nine figures of revenue in the coming years.

Our Investment

Banyan AI Fund I is investing in the seed round, ahead of a Series A.

We believe SpiNNcloud is a generational company positioned to reshape inference infrastructure. It aligns with three megatrends driving our thesis:

1. The move from dense to sparse computation
2. The rise of energy-proportional, AI-native hardware
3. The commercialization of neuromorphic architecture for real-world use cases

A World-Class Team

SpiNNcloud is led by Dr. Hector Gonzalez, a deep technologist with a decade of experience in neuromorphic engineering, energy-efficient computing, and hybrid AI. A graduate of TU Dresden, Hector previously co-led Europe’s Human Brain Project and helped drive the transition of SpiNNaker from academic prototype to commercial platform.

He’s supported by an exceptional team of scientists, engineers, and operators including co-founders from Amazon, GlobalFoundries, RacyICs, and TU Munich, and advisors like Steve Furber (inventor of the ARM CPU) and experts from DeepMind, IQM, and ETH Zurich. With over 50 employees and 30% holding PhDs, SpiNNcloud has the technical depth and execution capacity to deliver.

Why It Fits Banyan AI Fund I Thesis

We invest in AI-native infrastructure that redefines what’s possible—not incremental GPU improvements. SpiNNcloud represents a full-stack rethink of inference compute, aligned with the most fundamental trend in AI today: energy-efficient, sparse computation at scale.

This is what infrastructure for the Third Wave of AI looks like.

‍

Sam Awrabi

Founder and MD, Banyan