Bare metal
Single-tenant rack. AMD Genoa + Intel Sapphire Rapids — no hypervisor tax.
0% virt overheadGBCORP.CLOUD · IN PRODUCTION ACROSS 17 REGIONS
Compute, storage, network, and edge across GBCORP's seventeen-country office grid. One API. One bill. One SLA. Data stays in the jurisdiction you choose — no fine print.
FOOTPRINT
Same engineering team, same SLA, same console — in every region. Pick where your data lives. Deploy in any subset of seventeen.
THE PLATFORM
Pick a category to explore. Every service runs on the same private backbone, ships in the same SDK, and shares one console.
Run any workload on any region. Live-migrate between metal and VMs without redeploying the application.
Single-tenant rack. AMD Genoa + Intel Sapphire Rapids — no hypervisor tax.
0% virt overheadFifteen instance families. Live-migrate across regions with zero downtime.
15 familiesManaged Kubernetes with anti-entropy replication across three availability zones.
99.99% control planeNVIDIA H100 / H200 reserved by the hour. No spot games, no preemption.
A100 · H100 · H200Ampere Altra at roughly 40% lower cost-per-core versus equivalent x86.
128 vCPU / instanceServerless V8 isolates. Same TypeScript on the edge and at the origin.
<10ms cold startObject, block, file, and managed databases — replicated across four regions by default; deeper on request.
S3-compatible API. Async multi-region replication with strong read-after-write.
11 9s durabilityNVMe-backed SSD with snapshots, live resize, and encryption-at-rest.
<0.4ms p95 latencyNFSv4 + SMBv3 shares with auto-scale and full POSIX semantics.
Up to 10 PiB / sharePoint-in-time recovery, read replicas, automated failover, logical replication.
PG 14 · 15 · 16Document DB and in-memory cache, oplog-based replication, TLS by default.
100 shards · 2 TB cacheSeven-year-plus retention with WORM lock, audit trails, and regulatory templates.
$0.004 / GB / monthA sovereign private backbone with anycast routing and DDoS scrubbing — all included, no add-on SKUs.
Inter-region private network. Encrypted overlay; zero internet hops.
99.99% backbone SLOA single IP routed to the nearest healthy region. Multi-region failover.
342 PoPsL3 / L4 / L7 mitigation absorbed at the edge before reaching origin.
∞ Tbps absorbedApplication, network, and global load balancing under one console.
L4 + L7DNSSEC-signed authoritative DNS served from the private backbone.
<10ms global resolveRate limiting, JWT validation, request transformation, OpenAPI introspection.
OpenAPI 3.1Run code, transforms, and cache logic at 342 points of presence — same deployment model as the rest of the stack.
V8 isolates running TypeScript. Same artifact at edge and origin.
<10ms cold startStale-while-revalidate, granular cache keys, real-time invalidation.
95% origin offloadResize, format-convert, watermark on demand at the closest PoP.
WebP · AVIF · JPEG-XLHLS / DASH packaging for live and on-demand at the edge.
H.264 · H.265 · AV1Per-request country, region, and ASN policies.
GeoIP at the edgePer-key, per-IP, per-region — enforced at the edge before origin.
<1ms decisionSovereign by default. Per-region data residency, customer-managed keys, audited at every layer.
OIDC, SAML, and SCIM. Role-based access; every API call audit logged.
SCIM 2.0HashiCorp Vault-compatible API. HSM-backed roots, automatic rotation.
FIPS 140-3 HSMsBYOK and HYOK. Your keys live in your HSMs, never ours.
BYOK + HYOKL7 attack rules + your own signatures. OWASP Top 10 covered.
OWASP Top 10Every API call captured. Immutable, exportable, streamable to your warehouse.
7-year retentionSOC 2, ISO 27001, HIPAA, PCI DSS — reports available on demand.
8 frameworksHosted inference, vector search, and stream processing on the same control plane as the rest of the platform.
Host Llama, Mistral, Qwen, or your own fine-tune behind an OpenAI-compatible API.
OpenAI-compatible APIpgvector-compatible vector search; per-region embeddings; hybrid retrieval.
1B+ vectors / indexColumnar OLAP on an S3-compatible data lake. ANSI SQL :: 2023.
SQL :: 2023Kafka-compatible streams and Flink-style stateful processing.
Kafka API + Flink SQLReserved GPU clusters for fine-tuning. No queue, no pre-emption.
H100 / H200 clustersManaged JupyterHub with per-user GPU quotas for data teams.
Per-user GPU poolARCHITECTURE
The whole stack — from the physical rack in your jurisdiction up to the control-plane API — is owned and operated by one engineering team. No third-party seams, no escalation handoffs, no two-vendor billing.
CONTROL PLANE
L06API · console · IAM · billing · ops automation
OBSERVABILITY
L05Logs · metrics · traces · SLO budgets · alerting
COMPUTE & EDGE
L04Metal · VMs · containers · K8s · V8 isolates at 342 PoPs
STORAGE & DATA
L03Object · block · file · DBaaS · 4-region replication
NETWORK
L02Private backbone · anycast · DDoS · IPv6 · sovereign routing
PHYSICAL FOOTPRINT
L0117 regions · 342 edge PoPs · 4 continents · GBCORP office grid
DEPLOY
The same CLI runs against every region on the network. Pick a region. Pick a service. Ship. The control plane handles replication, networking, and observability automatically.
$ npm i -g @gbcorp/cloud-cli
# Authenticate once.
$ gbcorp auth login
# Pick a region and deploy a container.
$ gbcorp deploy my-api \
--region eu-vilnius \
--image ghcr.io/me/api:v0.4.2 \
--scale auto \
--sovereign
→ provisioning ✓ region selected · eu-vilnius (sovereign-eu)
→ building ✓ container · 412 MB · sha256:a3e29c…
→ rolling out ✓ 3/3 healthy · canary · main · prev
→ ready ✓ https://my-api.eu-vilnius.gbcorp.cloud
p95: 38ms · cost: $0.0042 / request INTEGRATIONS
COMPLIANCE
THE PROMISE
Pick the country. We bring the cloud.
START
Most pilots take a single discovery call to scope. A solutions engineer joins by day three. Production-ready inside two weeks.