Keep EDA Licenses Working. Turn Fabrication Data Into Yield.

WEKA® NeuralMesh™ runs EDA, lithography, and inspection AI from one platform, on premises or in any major cloud.

USE CASES

One Platform for All Semiconductor Fab and Design Use Cases

CPU design farms and GPU AI clusters break storage the same way: millions of concurrent file lookups. NeuralMesh spreads data and metadata across every server, so both read one dataset at full speed, even mid-rebuild.

A colorful, close-up view of an intricate computer chip's circuitry.

EDA Verification, Signoff, and Tape-Out

Nightly regressions create, open, and delete millions of small files. NeuralMesh spreads lookups across every server, so runs finish overnight and tape-out dates hold.

A silicon wafer being processed by technicians in a semiconductor cleanroom under amber light.

Computational Lithography

Lithography is moving to GPUs that need data far faster. NeuralMesh keeps those jobs fed and carries multi-day runs through failures, so you keep full model detail.

A person in a cleanroom suit uses a green laser to work on a wafer, next to a robotic arm.

Process and Equipment Simulation

Simulations of process steps and fab equipment run for days. NeuralMesh rebuilds across every node while jobs keep running, so a drive failure doesn't force a restart.

A semiconductor wafer being scanned by a green laser on a testing machine.

Wafer Inspection and Defect Detection

Inspection tools capture terabytes of wafer images a day. NeuralMesh feeds those images to GPU defect models for training and live checks, without staging copies.

Digital illustration of data streaming from server racks and fiber optic cables to a glowing cloud.

Hybrid Cloud Chip Design

NeuralMesh copies the file index first, so design data shows up in the cloud almost at once. Burst jobs start before a full transfer finishes, in AWS, Azure, Google Cloud, or OCI.

NeuralMesh Results for Chip Design and Fabrication Teams

Examine the results from semiconductor and electronics customers.

0→90%
Production yield at an electronics maker
A leading electronics manufacturer raised production yield from 50% to 90% by feeding inspection data to AI on NeuralMesh.
0TB
Same-day sensor data analysis
Analysis of sensor data now finishes the same day, down from more than two days.
0x
Faster rebuilds as clusters double
NeuralMesh rebuilds across every node, so recovery gets 2x to 10x faster each time the cluster doubles in size.

WHY WEKA

Faster Tape-Outs, Higher Yield, One Platform

A NAS per project and new storage for every GPU cluster means duplicate data and slowdowns at tape-out peak. A namespace on top hides the silos, not their limits.

Abstract glowing blue and green circuit board.

Keep EDA Licenses Working at Tape-Out
NeuralMesh spreads data and file lookups across every server, so jobs don’t queue behind one controller. Run EDA up to 10x faster than NAS, free licenses sooner, and hold tape-out dates.

1 / 9

Keep EDA Licenses Working at Tape-Out
NeuralMesh spreads data and file lookups across every server, so jobs don’t queue behind one controller. Run EDA up to 10x faster than NAS, free licenses sooner, and hold tape-out dates.
Turn Inspection Images Into Yield Answers
Feed GPU defect models at full speed and keep years of inspection images in reach, so you train on all your fab history and find root causes faster.
Unify Design and Fab Data on One Platform
File and object access work on the same data, so CPU design farms and GPU AI clusters read from one system. You cut the second storage stack and the copies that come with it.

Handle Millions of Small Design Files at Once
NeuralMesh splits every file into the same small pieces and spreads them evenly. Small files read up to 77x faster than NFS, and lookups complete in microseconds.
Keep Full Speed Through Drive and Server Failures
NeuralMesh repairs itself across every node while jobs keep running. Rebuilds get 2x to 10x faster as the cluster doubles, so multi-day simulations finish instead of restarting.
Scale to 14 Exabytes Without Hand-Tuning
One system scales to trillions of files and adapts to each job on the fly. NeuralMesh Observe shows the health of every cluster from one dashboard.

Burst to Any Major Cloud Without Copying
The same software runs on premises and in AWS, Azure, Google Cloud, and OCI. Design data appears in the cloud almost at once, with the same performance and no hardware lock-in.
Isolate Each Project on Shared Hardware
Dedicate resources to your most sensitive programs or foundry customers while everything else shares one pool. Per-tenant encryption helps protect each customer’s IP.
Fit More Fab History in Less Space
Always-on data reduction stores up to 6x more AI data with under 5% write impact, guaranteed in contract. WEKApod® 3 fits over 1.1 exabytes in a single rack.

1 / 3

Jiyeon JungSr. Engineer, AI Infra • LG AI Research

“Choosing WEKA in a public cloud environment was the optimal decision for maximizing performance, and it really made a difference in accelerating our multi-node AI training.”

Read the Full Story

“Choosing WEKA in a public cloud environment was the optimal decision for maximizing performance, and it really made a difference in accelerating our multi-node AI training.”

Jiyeon JungSr. Engineer, AI Infra • LG AI Research

Read the Full Story