# NeuralMesh for Semiconductor Design and Fabrication

WEKA’s NeuralMesh™ gives semiconductor design and fabrication teams one namespace with file and object access to the same data to speed EDA verification, feed inspection AI that lifts yield, and burst to any major cloud without copying files first.

**Type:** Solution Brief

## Benefits

- Keep EDA Licenses Working at Tape-Out Peak
- Turn Inspection Images into Yield Answers
- Unify Design and Fab Data on One Platform
- Handle Millions of Small Design Files Concurrently
- Maintain Full Speed Through Drive and Server Failures
- Store and Tune Massive Datasets Easily
- Run On-Premises and Burst to the Cloud Seamlessly
- Isolate Each Project on Shared Hardware

## NeuralMesh™ Speeds Chip Verification, Raises Yield for Semiconductor Fabrication

WEKA® NeuralMesh™ serves Electronic Design Automation (EDA), lithography, and inspection AI from a single namespace with file and object access to the same data, all on-site or in any major cloud. Design and fabrication teams stop copying files between systems, cloud jobs start before a full transfer finishes, and years of inspection images and equipment records remain accessible for yield analysis.

## Key Benefits of NeuralMesh in Semiconductor Fab

**Keep EDA Licenses Working at Tape-Out Peak**

- Run EDA 10x faster than NAS
- Free-up licenses sooner
- Protect tape-out dates

**Turn Inspection Images into Yield Answers**

- Keep years of images in reach
- Feed GPUs for defect checks
- Train on all your fab history

**Unify Design and Fab Data on One Platform**

- Share file and object access
- Serve CPU and GPU jobs alike
- Eliminate duplicate storage

**Handle Millions of Small Design Files Concurrently**

- Split every file evenly
- Read files 77x faster
- Run lookups in microseconds

**Maintain Full Speed Through Drive and Server Failures**

- Rebuild across every node
- Finish multi-day simulations
- Recover faster as you grow

**Store and Tune Massive Datasets Easily**

- Scale up to 14 exabytes
- Adapt to each job on the fly
- Track the cluster in Observe

**Run On-Premises and Burst to the Cloud Seamlessly**

- Access data where it lives
- Use AWS, Azure, GCP, or OCI
- Avoid hardware lock-in

**Isolate Each Project on Shared Hardware**

- Dedicate nodes to key work
- Encrypt each tenant's data
- Securely share infrastructure

```json
{
  "_key": "rp00031",
  "_type": "statsBlock",
  "markDefs": null,
  "stats": [
    {
      "_key": "rp00028",
      "_type": "statItem",
      "label": "EDA job sets on SPECstorage 2020",
      "value": "17,000"
    },
    {
      "_key": "rp00029",
      "_type": "statItem",
      "label": "Production yield at a leading electronics maker",
      "value": "50%→90%"
    },
    {
      "_key": "rp00030",
      "_type": "statItem",
      "label": "Reduction in overall data-operations costs",
      "value": "90%"
    }
  ]
}
```

```json
{
  "_key": "rp00034",
  "_type": "statsBlock",
  "markDefs": null,
  "stats": [
    {
      "_key": "rp00032",
      "_type": "statItem",
      "label": "Sensor data analyzed same day vs. 2+ days",
      "value": "100TB"
    },
    {
      "_key": "rp00033",
      "_type": "statItem",
      "label": "Faster rebuilds as clusters double in size",
      "value": "10x"
    }
  ]
}
```

## Semiconductor Design and Fabrication Use Cases

**Accelerated EDA Verification, Signoff and Tape-Out**

NeuralMesh spreads file lookups across every server and node, so nightly regressions finish overnight and tape-out dates hold.

**Computational Lithography and Process Simulation**

NeuralMesh keeps lithography jobs fed and carries multi-day simulations through failures, so teams keep full model detail.

**Wafer Inspection, Defect Detection and Yield Analysis**

NeuralMesh feeds live floor data and fab history to inspection AI that flags deviations in real time and speeds root-cause work.

**Hybrid Cloud Design and AI-assisted Engineering**

NeuralMesh presents design data in the cloud almost at once, so burst jobs start early and archives answer as fast as live work.

```json
{
  "_key": "a2f904fae4b5",
  "_type": "customerQuoteBlock",
  "markDefs": null,
  "quote": {
    "anonymizedAttribution": "Fortune Global 500 Semiconductor Company",
    "anonymous": true,
    "authorName": "Patrick Bangert",
    "companyLogo": {
      "_type": "image",
      "asset": {
        "_id": "image-c7baa141e12ef5217829ccdf93eab592030d4497-300x46-svg",
        "metadata": {
          "dimensions": {
            "_type": "sanity.imageDimensions",
            "aspectRatio": 6.521739130434782,
            "height": 46,
            "width": 300
          },
          "lqip": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAADCAYAAACTWi8uAAAACXBIWXMAAAsTAAALEwEAmpwYAAAAs0lEQVR4nD2OO24CQRBE9whUzxW2e/ZaSMYyAbIjJ05YC1iQzCKt/JGl6RofFfUEDip7VfU6aJ2TcpTe16J+hnJG5ijGX1F+i3JK5icoR2jdiPKC3peV+hvMa+MHfxQtU/BdlKF8h/EhmR9F/UtymSTzM4L/Md9Do+jXYGD+KsafdpzLczBi/OjS8Lcg8yZad+21GdY9jIdV5iXswipZOSTzbbAwX9CXpxBpo72/QMuIzPkOQM5Q8RRpKeQAAAAASUVORK5CYII="
        },
        "url": "https://cdn.sanity.io/images/ult5g8gw/production/c7baa141e12ef5217829ccdf93eab592030d4497-300x46.svg"
      },
      "crop": null,
      "hotspot": null
    },
    "companyLogoMono": {
      "_type": "image",
      "asset": {
        "_id": "image-7df38191f92e36212a5951f5903b9e38c93bcc25-300x46-svg",
        "metadata": {
          "dimensions": {
            "_type": "sanity.imageDimensions",
            "aspectRatio": 6.521739130434782,
            "height": 46,
            "width": 300
          },
          "lqip": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAADCAYAAACTWi8uAAAACXBIWXMAAAsTAAALEwEAmpwYAAAAc0lEQVR4nE3OuwoCQRBE0fOrgooGopGJiSu7PkBXEF0x8lelpQSDYeiZW7caejQY4fg3vzDggH3eJjjhhg3e4afhiv+CO4zR4ZHPe85Ptk3wEmaNZ4qXYc4lrLYrFmmtxgq32aaEddc8D1uZWRYp6SrC/gOcrR8S1zUcDgAAAABJRU5ErkJggg=="
        },
        "url": "https://cdn.sanity.io/images/ult5g8gw/production/7df38191f92e36212a5951f5903b9e38c93bcc25-300x46.svg"
      },
      "crop": null,
      "hotspot": null
    },
    "companyName": "Samsung SDS",
    "jobTitle": "Chief of AI, Data Science",
    "text": "For the same hardware, [WEKA] can deliver data faster, thus allowing you to feed more of these GPUs and fill their memory just in time for you to do the processing."
  }
}
```

[Download PDF](/api/resource-pdf?slug=neuralmesh-for-semiconductor-design-and-fabrication)