Right-sized for your workload
Four clear hardware tiers
Your capabilities select the smallest tier that fits. You can move up for headroom, switch hardware vendors, or reserve a Blackwell configuration without changing the workload plan.
Choose by sizing trigger
Ada configurations are available now and shown as the starting price.
Google Distributed Cloud connected ready
- 1Per node
Value
Integration and data-only capabilities. Single node, or 3-node HA for small-format retail.
- Licensed capacity
- 8–16 vCPU
- Video
- Up to 4 streams
- GPU
- 1
- Storage
- 1 TB
Ada baseline from
$1,600 or $95/mo
Dell Pro Micro QCM1250 (Core Ultra 5 235T, 16–32 GB DDR5, 512 GB NVMe)
- 2Per node
Performance
Any computer-vision capability. One GPU-capable sled per site.
- Licensed capacity
- 16–48 vCPU
- Video
- Up to 16 streams
- GPU
- 1
- Storage
- 4 TB
Ada baseline from
$7,000 or $260/mo
Dell PowerEdge XR8000r + 1× XR8620t sled (5th Gen Xeon, 128–256 GB) + 1× NVIDIA L4
- 3Per site
Platinum
HA required, more than 16 streams, manufacturing and campus.
- Licensed capacity
- 64–128 vCPU
- Video
- Up to 48 streams
- GPU
- 3
- Storage
- 12 TB
Ada baseline from
$24,000 or $820/mo
XR8000r + 3× XR8620t sleds + 3× NVIDIA L4 (GDC connected servers G2 three-machine default)
- 4Per site
Titanium
Regional hub, distribution center, plant with multiple lines, multi-model inference, sovereign AI.
- Licensed capacity
- 128–256 vCPU
- Video
- Up to 100 streams
- GPU
- 8
- Storage
- 24 TB
Ada baseline from
$85,000 or $2,900/mo
PowerEdge R760xa with 4× NVIDIA L40S, or 2× XR8000r fully populated (6× XR8620t, 6× L4)
Compare every Ada, Blackwell and Blackwell Ultra configuration
All reference SKUs and display prices come from the catalog. Blackwell configurations can be reserved now for the first validated wave.
| Tier and generation | Reference configuration | Capacity | Sizing trigger | One-time | As a service |
|---|---|---|---|---|---|
| Value | Dell Pro Micro QCM1250 (Core Ultra 5 235T, 16–32 GB DDR5, 512 GB NVMe) | 8 → 16 vCPU · 1 GPU · up to 4 streams · 1 TB · 1 or 3 nodes | Integration and data-only capabilities. Single node, or 3-node HA for small-format retail. | $1,600per node | $95/mo60 mo, 5-yr maintenance |
| BlackwellComing soon, reserve now | Dell Pro Max Micro with NVIDIA RTX PRO Blackwell (light inference) | Same tier capacity | Integration and data-only capabilities. Single node, or 3-node HA for small-format retail. | $2,400per node | $110/mo60 mo, 5-yr maintenance |
| BlackwellComing soon, reserve now | Dell Pro Max with GB10 (Grace Blackwell, 128 GB unified, ~1 PFLOP FP4) | Same tier capacity | Integration and data-only capabilities. Single node, or 3-node HA for small-format retail. | $4,000per node | $150/mo60 mo, 5-yr maintenance |
| Performance | Dell PowerEdge XR8000r + 1× XR8620t sled (5th Gen Xeon, 128–256 GB) + 1× NVIDIA L4 | 16 → 48 vCPU · 1 GPU · up to 16 streams · 4 TB · 1 or 3 nodes | Any computer-vision capability. One GPU-capable sled per site. | $7,000per node | $260/mo60 mo, 5-yr maintenance |
| BlackwellComing soon, reserve now | XR8000r + 1× XR8720t sled (Xeon 6 SoC, up to 72 cores) + 1× RTX PRO 6000 Blackwell SE class | Same tier capacity | Any computer-vision capability. One GPU-capable sled per site. | $9,500per node | $350/mo60 mo, 5-yr maintenance |
| Platinum | XR8000r + 3× XR8620t sleds + 3× NVIDIA L4 (GDC connected servers G2 three-machine default) | 64 → 128 vCPU · 3 GPU · up to 48 streams · 12 TB · 3 nodes | HA required, more than 16 streams, manufacturing and campus. | $24,000per site | $820/mo60 mo, 5-yr maintenance |
| BlackwellComing soon, reserve now | XR8000r + 3× XR8720t + 3× Blackwell GPUs | Same tier capacity | HA required, more than 16 streams, manufacturing and campus. | $32,000per site | $1,090/mo60 mo, 5-yr maintenance |
| Titanium | PowerEdge R760xa with 4× NVIDIA L40S, or 2× XR8000r fully populated (6× XR8620t, 6× L4) | 128 → 256 vCPU · 8 GPU · up to 100 streams · 24 TB · 6 nodes | Regional hub, distribution center, plant with multiple lines, multi-model inference, sovereign AI. | $85,000per site | $2,900/mo60 mo, 5-yr maintenance |
| BlackwellComing soon, reserve now | PowerEdge XE7745 with up to 8× RTX PRO 6000 Blackwell Server Edition | Same tier capacity | Regional hub, distribution center, plant with multiple lines, multi-model inference, sovereign AI. | $185,000per site | $6,300/mo60 mo, 5-yr maintenance |
| Blackwell UltraComing soon, reserve now | NVIDIA HGX B300 on PowerEdge XE9785 (named for GDC at Next ’26) | Same tier capacity | Regional hub, distribution center, plant with multiple lines, multi-model inference, sovereign AI. | $450,000per site | $15,000/mo60 mo, 5-yr maintenance |
Vendor equivalents
Same capacity, no lock-in. Dell matches Zones’ pipeline; the others change the name and a small hardware delta.
- Dell baseline
- Intel-based −3%
- HPE +5%
- Lenovo +3%
- MSI −5%
Intel exited the NUC line in 2023, so custom small-form-factor designs are labeled Intel-based.
Blackwell for Google Distributed Cloud
Blackwell for Google Distributed Cloud was announced at Next ’26. Reserve it now and your order slots into the first validated wave.
Titanium adds a Blackwell Ultra configuration (NVIDIA HGX B300) recommended for on-prem Gemini at hub scale. The Value tier also lists the Dell Pro Max with GB10, the single-site AI in a box, quoted separately in this release.
Default Zones services by tier
- Valuesimple staging · simple deployment · simple field
- Performancestandard staging · standard deployment · standard field
- Platinumcomplex staging · complex deployment · gold field
- Titaniumcomplex staging · complex deployment · gold field · 4-hour
Intermittent connectivity, no problem
Intermittent connectivity, no problem. GDC connected keeps your workloads running for up to 7 days if the site loses its link to Google Cloud.
You never have to pick a server. Pick capabilities and the tier follows.