generated: '2026-07-19'
method: searched
source: https://docs.nvidia.com/sdgm/quotas-and-limits
docs: https://docs.nvidia.com/sdgm/quotas-and-limits
style: quotas-and-concurrency
note: Kumo publishes data-scale quotas and job concurrency limits rather than HTTP
  request-rate limits. Limits vary by deployment mode and a given account may be configured
  lower; increases are handled through the customer's Kumo contact.
rate_limits:
- name: Queued jobs
  limit_count: 15
  unit: jobs
  scope: training or batch prediction jobs queued at once
- name: Parallel job execution (SaaS)
  limit_count: 10
  unit: jobs
  scope: SaaS deployment
- name: Parallel job execution (Databricks Native)
  limit_count: 10
  unit: jobs
  scope: Databricks Native deployment
- name: Parallel job execution (Snowflake Native)
  limit_count: 1
  unit: jobs
  scope: Snowflake Native deployment; parallelism achieved by launching multiple Kumo
    applications
quotas:
- name: Rows in a table
  saas: 15000000000
  snowflake_native: 1000000000
  databricks_native: 10000000000
- name: Columns in a table
  saas: 2000
  snowflake_native: 100
  databricks_native: 2000
- name: Table size
  saas: 2 TB
  snowflake_native: 100 GB
  databricks_native: 2 TB
- name: Row size
  saas: 10 KB
  snowflake_native: 10 KB
  databricks_native: 10 KB
- name: Tables in a graph
  saas: 12
  snowflake_native: 12
  databricks_native: 12
- name: Rows in a graph
  saas: 30000000000
  snowflake_native: 1000000000
  databricks_native: 30000000000
- name: Graph size
  saas: 2 TB
  snowflake_native: 200 GB
  databricks_native: 2 TB
- name: Edges in a graph
  saas: 30000000000
  snowflake_native: 4000000000
  databricks_native: 30000000000