We use third-party cookies in order to personalize your site experience. See our Privacy Policy.

Technology thesis · Semiconductors & Chips

low conviction concept

Neuromorphic computing

Brain-inspired chips slash power for always-on edge sensing, but Intel Loihi and IBM NorthPole stay research-only; the real test is whether BrainChip and Innatera win high-volume edge designs.

Position maintained continuously · last reviewed Jun 24, 2026

The thesis

State of the art (2026)

Neuromorphic remains a research field with a thin commercial edge. The flagship system, Intel's Hala Point (1.15 billion neurons across 1,152 Loihi 2 chips), runs at Sandia as a research platform, not a product, and Intel has shipped no official Loihi 3 silicon. IBM's NorthPole is likewise a research inference architecture. The real traction is at the edge: BrainChip, the only listed pure-play, raised $25m in December 2025 to scale Akida 2; Innatera began shipping its Pulsar SNN microcontroller into consumer wearables and smart-home devices in early 2026; SpiNNcloud is commercialising SpiNNaker 2. Patent filings surged in 2025, but spiking-network software tooling and a killer high-volume application are still missing.

Core thesis

Neuromorphic chips process information with spiking neural networks, trading dense clocked matrix maths for sparse, event-driven activity that can cut inference power by one to two orders of magnitude on the right workloads. The hardware is real: Intel's Hala Point packs 1.15 billion neurons across 1,152 Loihi 2 chips at Sandia, and IBM's NorthPole shows roughly 25x the energy efficiency of a comparable 12nm GPU on ResNet-50. But general-purpose neuromorphic remains a long-horizon bet. Spiking-network software tooling lags PyTorch by years, mainstream models do not map cleanly onto spikes, and the advantage is confined to narrow problem classes. The live commercial slice is the edge: always-on sensing silicon (BrainChip Akida, SynSense, Innatera Pulsar) and event-based vision (Prophesee/Sony, iniVation), where sub-milliwatt, microsecond-latency sensing is a genuine physics win rather than a research demo.

The rest of the file

Everything below is live inside CanaryIQ

The full analysis behind the verdict — the structure is real; the content unlocks when you log in.

Signal stack

Evidence stacked leading → lagging

6 signals
talent
research
patent
expert
market

Technology-native KPIs

Metrics that predict trajectory, tracked over time

5 tracked
Largest single neuromorphic system, neuron count
Energy efficiency vs GPU (NorthPole, ResNet-50)
Edge SNN MCU energy/latency advantage (Innatera Pulsar)
Event-based vision sensor market size
Public pure-play traction (BrainChip funding)

Landscape map

Who builds what — and who depends on whom

17 players · 4 layers

Catalyst calendar

Dated events that will move the position

6 ahead

Technology roadmap

Milestones on the path to maturity

9 milestones

Watchlists

Companies, people and papers — each with a remove-by condition

19 · 20
Companies · 19
People · 20

Decision frameworks

The same call, framed for your desk

Locked
Public Equity
PE / VC
Corporate Leader

Thesis changelog

When our view changed, and why

5 updates

Change our mind

2 disconfirming conditions

The rest is inside

You've read the verdict. The file is much deeper.

The full signal stack, technology-native KPIs tracked over time, the landscape of who depends on whom, the dated catalyst calendar, decision frameworks for every desk, live watchlists and the changelog of every time our call on Neuromorphic computing has changed — all live inside CanaryIQ.