A new paradigm of AI hardware to power human progress.

AI co-designed silicon built on new device physics. Partnering with the world's most advanced institutions.

ABOUT
DEPLOYMENT
In Production
with the world’s largest semiconductor companies
IMPACT
2x Faster
verification on the most complex IP and SoCs
HARDWARE
Reference Systems
available , prioritizing long-context memory inside AI inference
Backing
$85M+
funding raised. Strategic round led by Samsung Catalyst
EFFICIENCY
Intelligence / $ / W
inference per second, per watt and per mm² of silicon
Normal EDA · In production
AI-accelerated co-design for silicon engineering teams.

Normal EDA proves what works and explores what's possible in silicon. It learns continually, post-training our AI models on your design data and your team's implicit knowledge, deployed on-premises.

State of the art on industry benchmarks for RTL design and verification
READ THE RESEARCH
A NEW EDA STACK
We build fast, AI-native simulation and synthesis engines.  Agents learn from every run, the loop closes from architecture to signoff.
Structured representations of the design
The Ontology grounds every downstream artifact, from test plans to stimulus to RTL, in your team’s single source of truth for the design.
From spec to plan to simulation
Extract design requirements from specs, generate test plans, then run simulations. Leverage the simulation output as feedback to improve the specs and optimize generated collateral.
Debug + fix — in your terminal
The agent runs your regression, root-causes the failure against the spec, and proposes a fix for review.
The semiconductor industry’s leading companies use Normal EDA to scale new hardware and accelerate time-to-market.
GET in Touch
NORMAL ASICs
Normal ASICs compute the way physical systems do: noise as a resource, compute with memory, asynchronous. The result is orders-of-magnitude more efficient compute.

Normal ASICs build what Normal EDA makes possible: custom silicon, with novel device physics, built toward a 10-100× gain in AI inference per dollar, per watt. Designed to break the long-context wall, and scale to future multi-modal and world modeling workloads.

Reference system available
TALK TO OUR ENGINEERS
Pilot deployments open. We prioritize by application and impact, including long-context agentic reasoning.
01
WORkload
Thermo processing-with-memory runs the model’s heaviest operations inside the memory itself, starting with long-context decoding.
03
CAPACITY
Transformer attention computed in memory for up to ~500B parameter models at rack scale.
02
Physics
Stochastic analog computation in memory: thermal noise is a computational resource.
04
SYSTEM
PCIe accelerator cards in standard air-cooled servers, scaling from card to server to rack. Normal forward-deployed engineers install in your environment.
THE CO-DESIGN ERA
Frontier AI models are converging to physics. We close the gap between software and hardware, fundamentally changing how hardware is built
Hardware
GENERATIVE AI
TEAM
An applied AI company solving the hardest problems in AI and silicon.

Normal is built by engineers at the frontier of physical and mathematical intelligence: co-creators of core TensorFlow frameworks, co-founders of Meta's Probability team, architects of Google's first production-scale AI deployments, pioneers of NISQ at Los Alamos, and co-designers of Haskell and C#.

Our silicon team built the chips that power modern computing: CPU and GPU IP at NVIDIA and Apple, AI accelerators at Graphcore, and tape-outs across analog and mixed-signal designs. Alongside these builders are operators who took frontier work to market from zero to one at Palantir, Google X, and high-growth deep tech companies.

PLay Video
Writing
Insights & breakthroughs, shaping the future of computing.
All posts
4.23.2026
From Specifications to Formal Models: Autoformalizing Memory Chips with DRAMBench
Read more
Research
All
3.25.2026
Fortune Exclusive: Normal Computing raises $50M from Samsung Catalyst to tackle soaring AI chip costs and power demands
Read more
Press
All
3.25.2026
Normal Computing Raises $50M Led By Samsung Catalyst to Accelerate Silicon Design and Solve AI Hardware Energy Crisis
Read more
Announcements
All
3.2.2026
Building an Open-Source Verilog Simulator with AI: 580K Lines in 43 Days
Read more
Engineering
All
1.12.2026
A Complete Decomposition of Stochastic Differential Equations
Read more
Research
All