CDN — content delivery
Low-latency global content delivery on a Tier-1 backbone. The fastest win for any relocating product with users worldwide.
Core offer
Full-stack infrastructure
One contract for your whole edge-to-AI stack.
Built on Tier-1 global backbone capacity, billed through a single B2B invoice. Lead with content delivery, servers and security; add GPU compute when you need it.
Cloud & bare-metal servers
Virtual and dedicated servers on demand, scalable per project. Provisioned fast, billed monthly, no card on file.
Core offer
WAF & DDoS protection
Server-level Web Application Firewall and always-on DDoS mitigation. Keeps client workloads online under attack.
Core offer
Managed DNS
Anycast DNS with global resolution and failover. Ships alongside CDN so domains and delivery move together.
Core offer
Cyber security
Edge protection, bot mitigation and traffic filtering across the whole perimeter. Compliance-friendly for regulated clients.
Core offer
GPU compute
H200 / H100 / L40S / A100 / RTX nodes for fine-tuning and inference. Single nodes to small clusters — our premium add-on.
★ Premium add-on
Why we exist
Frontier models on Asian silicon, with a contract that clears globally with flexible B2B terms.
We are not a reseller landing on top of someone else's API. We hold capacity contracts with two hyperscalers in Greater China, rack the boxes ourselves, and bill through one of our contracting entities — a Hong Kong or Singaporean company with flexible cross-border banking. The exact entity is picked at signing to fit your accounting; the list is sent on request. No card declines, no "we'll get back to you" sales loop.
- Fixed monthly rate locked in your cart — no surprise overages.
- Wire transfer in USD, CNY or USDT, closing docs the next business day.
- SSH and Jupyter access in under an hour after payment lands.
- Direct line to a sysadmin on Telegram, not a ticket queue.
B2B cabinet · why register
Your fleet, your invoices, your SLA — in one window.
After registration we open a personal B2B cabinet. Live GPU-hour load, every running and provisioning instance with its SSH endpoint, every invoice with its status, SLA credits — no chasing accountants over email, no Excel spreadsheets.
WEST-EAST CLOUD · B2B CABINET
Welcome, Global AI Solutions Ltd.
alex@globalai.io · contract #WEC-2026-0187
SZ · Futian
GPUs in use
8
+2 / wk
Spend MTD
$14,100
78% of budget
Active instances
2
0 provisioning
SLA · 30d
99.98%
0 incidents
GPU-hours · 30 days
1,344 h
Aggregated across all running instances
-
$
Invoice INV-WEC-20260522-1794-3
NVIDIA RTX A6000 24GB · 4× · HK · 6d -
$
Invoice INV-WEC-20260514-3594-2
NVIDIA L40S 48GB · 4× · SZ · 2w -
▶
Instance provisioned · crimson-comet-02
L40S 48GB · 4× · SZ · 2w -
$
Invoice INV-WEC-20260502-1794-1
NVIDIA A100 80GB · 8× · SZ · 3w
| Name | GPU | Region | Uptime | Utilisation | SSH | Rate | Status |
|---|---|---|---|---|---|---|---|
| tundra-beacon-01 | A100 80GB · 8× | SZ | 23d 11h | 92% | 10.109.133.228 | $3,200/mo | running |
| crimson-comet-02 | L40S 48GB · 4× | SZ | 14d 02h | 78% | 10.132.51.117 | $1,580/mo | running |
| orion-edge-04 | H100 SXM · 8× | HK | 06d 18h | 65% | 10.211.44.18 | $9,200/mo | running |
| silica-aurora-07 | H200 SXM · 8× | HK | 00h 14m | 12% | 10.211.44.41 | $12,800/mo | provisioning |
| iris-nebula-12 | RTX 4090 · 2× | SZ | — | offline | — | $420/mo | stopped |
5 of 8 GPUs visible in this view · sorted by uptime · a row click would open the console and metrics
Paid YTD
$48,200
8 invoices
In flight
$14,100
cleared in last 30d
Due
$0
all current
| # | SKU / GPU | Region | Period | Amount | Status | |
|---|---|---|---|---|---|---|
| INV-WEC-20260522-1794-3 | RTX A6000 24GB · 4× | HK | May 2026 | $1,856 | paid | |
| INV-WEC-20260514-3594-2 | L40S 48GB · 4× | SZ | May 2026 | $1,582 | paid | |
| INV-WEC-20260502-1794-1 | A100 80GB · 8× | SZ | May 2026 | $10,662 | paid | |
| INV-WEC-20260601-0014-D | H200 SXM · 8× · 1 mo | HK | June 2026 | $12,800 | draft |
Account details
- Legal name
- Global AI Solutions Ltd.
- Registration Number
- 2024-HK-471829
- Contract
- #WEC-2026-0187
- Account manager
- Marina K. · @operator_X2026
SLA credits
$0
No incidents in the last 90 days. Credits accrue automatically whenever uptime drops below 99.9%.
-
GPU-hour usage
Live chart of your monthly load across every instance. Spot spikes before the controller does.
-
Running instances
Every node, its SSH endpoint, uptime, region and rate. Stop a box or pull logs with one click.
-
Invoice history
Every invoice issued under your contract: paid, pending or draft. Copy the wire reference straight into your bank app.
-
SLA credits & account
Uptime-credit accrual, contract reference, account ID, contact channel. Everything an accountant or auditor would ask for.
Access is granted within one business day after we verify the legal entity. No card on file, no auto-renew traps — the cabinet only shows what you actually rent.
Catalog
Pick a SKU. The rate is what you'll pay.
Sixteen production SKUs across two regions. Filter, add to cart, and the invoice generates against the rate visible right now — not a quote we'll send "shortly".
- 1Filter the catalog
- 2Add SKU to cart
- 3Sign in as a legal entity
- 4Generate & pay invoice
17 SKUs shown
| Configuration | Region | vCPU | RAM | GPU mem | Net | Price / month | Status |
|---|
Hourly rate = monthly ÷ 730. The rate in your cart is the rate on the invoice. Provisioning starts when payment clears — typically within one business day for SWIFT, minutes for USDT (TRC-20).
Footprint
Two regions, one fabric.
Both sites carry the same SKU families, so you can grow or migrate workloads without re-architecting. RTT to major global hubs is measured monthly from a probe in each facility.
HK
Hong Kong
- Facility
- Tier-3 · Tseung Kwan O
- Power
- 2N · 16 MW
- Fabric
- InfiniBand NDR 400 Gb/s
- Peering
- HKIX · 4× 100 GbE upstream
- RTT to Frankfurt
- 186 ms (median)
SZ · Futian
22.54°N · 114.06°E
HK · TKO
22.32°N · 114.27°E
< 2 ms
© WEST-EAST CLOUD · Greater Bay Area
SZ
Shenzhen
- Facility
- Tier-3 · Futian
- Power
- 2N · 22 MW
- Fabric
- RoCE v2 · 200 Gb/s
- Peering
- CN backbone · 200 Gb/s
- RTT to Frankfurt
- 194 ms (median)
Benchmarks
Throughput on Llama-3 70B fine-tune.
Tokens per second on a single node, batch 16, sequence 4096, FP8 where supported. Numbers are from our own runs against a reference dataset, not vendor slides.
H200 SXM · 8×
14 820 tok/s
H100 SXM · 8×
11 540 tok/s
L40S · 8×
6 210 tok/s
A100 80GB · 8×
5 080 tok/s
RTX 4090 · 8×
2 810 tok/s
V100 32GB · 8×
1 740 tok/s
Reference: torch 2.3 · transformers 4.41 · FA-2 · BF16/FP8. Raw logs available on request.
What teams actually run on us
Six workloads, one billing entity.
LLM fine-tuning
QLoRA on 70B-class models, full FT on 7–13B. NVLink within node, InfiniBand between nodes.
H100 8× · H200 8×
Inference at scale
vLLM / TGI behind your own gateway. L40S is the price/perf pick for 7–13B serving.
L40S 4–8× · A100 80GB
Computer vision
Detection, tracking, video pipelines. CUDA / TensorRT image pre-built on request.
A100 80GB · RTX 6000 Ada
Generative media
SDXL / Flux / video-diffusion render farms. 4090 wins on cost per frame.
RTX 4090 4–8× · L40S
Scientific compute
Molecular dynamics, CFD, genomics. FP64-heavy jobs land on A100 / H100 PCIe nodes.
A100 · H100 PCIe
Render & VFX
Blender, Houdini, Unreal frame servers. A6000 for VRAM-hungry scenes.
RTX 4090 · RTX A6000
Trust & SLA
The boring details that matter at month-end.
99.95%
Uptime SLA
Power, network and hypervisor availability, measured per calendar month. Credits apply automatically against the next invoice.
< 1 bd
Closing docs
Акт + счёт-фактура / invoice + statement sent the next business day after payment lands.
ISO 27001
Compliance
Upstream facilities certified ISO/IEC 27001 and 27017. We sign DPA on request before provisioning.
HK · SG
Contracting entity
A Hong Kong or Singaporean company with flexible cross-border banking. We pick the entity that fits your accounting at signing.
Full panel and reqs on request: wet4kg@gmail.com
Straight answers
FAQ
How do you handle cross-border payments for emerging markets?
You wire USD, CNY or USDT to one of our contracting entities — a Hong Kong or Singaporean company with flexible cross-border banking. The exact entity is picked at signing per your accounting; we send the panel and reqs on request. We hold capacity contracts directly with Alibaba Cloud / Tencent Cloud and settle in CNY on our side. Closing docs come in your currency. No card processors, no offshore wrappers.
How long from invoice paid to SSH?
USDT (TRC-20) — minutes. SWIFT in USD / CNY — same business day if before 14:00 HKT, otherwise next morning. We notify you on Telegram the moment funds land, then push credentials within 30–60 minutes.
Can I bring my own container / OS image?
Yes. We ship Ubuntu 22.04 with CUDA 12.4, cuDNN, NCCL and the usual ML stack pre-baked, but you can upload a custom image (Docker, qcow2, ISO) and we'll boot it on bare metal.
What if I need to scale mid-month?
Add SKUs to the cart, generate a top-up invoice — we pro-rate to the existing billing cycle so everything aligns. Downgrades take effect from the next cycle.
Do you do reserved capacity or 12-month contracts?
Yes, on H200 and H100 8× we discount roughly 10% for a 12-month reservation and 18% for 24 months. Email wet4kg@gmail.com with your spec.
Trial or pilot?
After KYC we can give a verified legal entity a 24-hour pilot on a 4090 or A100 node. We don't do free credits for individuals — sorry.