BaseTensor markBASETENSORRequest capacity
COMPUTE

Dedicated clusters, operated for you.

BaseTensor deploys and operates high-density GPU clusters and delivers them as dedicated capacity. Choose the level of operation you want.

01 · RESERVEDNOT YET IN SERVICE

Reserved dedicated clusters

A dedicated cluster reserved for your organisation under a multi-year agreement. You get exclusive use of the whole cluster — compute, fabric and storage — with an SLA and ramp schedule agreed at reservation. Hardware remains owned and operated by BaseTensor at BaseTensor facilities.

  • +Dedicated, physically contiguous cluster
  • +Multi-year reservation
  • +SLA and ramp agreed jointly
  • +Expansion priority on later phases

FORModel companies and enterprises that need guaranteed capacity for training and serving at scale.

02 · MANAGEDNOT YET IN SERVICE

Managed capacity

Dedicated capacity delivered as a managed service: BaseTensor owns, deploys and operates the hardware throughout the term; you bring the workload. No hardware procurement, ownership or operations on your side.

  • +Dedicated high-density capacity
  • +Operated and maintained by BaseTensor
  • +Flexible ramp within the reservation term
  • +Single point of accountability

FORTeams that want to focus on models and product.

What is included.

The same cluster under either offering. What changes is how much of the operation you run.

CUSTODY

Hardware stays with BaseTensor

Owned and operated by BaseTensor at BaseTensor facilities for the full term. Nothing is shipped, relocated or transferred.

ACCESS

Bare metal or managed orchestration

Chosen at reservation. Bare metal gives your team root on the nodes and your own orchestration; managed orchestration is operated by BaseTensor.

Option ABare metal, own orchestration
Option BManaged orchestration
TELEMETRY

A monitoring feed to your team

Node, fabric and facility telemetry is exposed to the customer for the term, so you see what we see.

SLA

Availability and replacement time

Two measures, agreed at reservation: cluster availability and time to replace a failed node. Figures are shared under NDA.

Availabilityunder NDA
Node replacementunder NDA
RAMP

Capacity delivered on a schedule

The ramp is fixed in the reservation agreement, phase by phase, and reported against.

EXPANSION

Priority on later phases

Reserved customers hold first call on capacity in subsequent phases, in the order of reservation.

RESERVATION SUMMARY

What a reservation fixes.

Every field below is set jointly and written into the reservation agreement. None of them is set by a price list.

Request capacity

Reservation summary

ILLUSTRATIVE
Offering
Reserved dedicated cluster / Managed capacity
Capacity
Agreed at reservation
Access
Bare metal / Managed orchestration
Term
Multi-year, agreed at reservation
Ramp
Phased, dated in the agreement
SLA
Availability · node replacement time
Telemetry
Customer feed, full term
Expansion
Priority on later phases
Region
Southeast Asia

Illustrative structure only. No field carries a committed value until a reservation agreement is executed.

Designed for full-rack training topologies.

PLANNED
COMPUTE RACKS

High-density, liquid-cooled, sized for full-rack training topologies.

FABRIC

High-bandwidth, low-latency interconnect for distributed training and hyperscale inference.

STORAGE

Local high-speed storage for weights and datasets.

Detailed configurations, delivery schedules and pricing are shared under NDA.
Capacity is provided to screened customers for their own workloads; hardware is not resold, sublet or transferred.

Request capacity