Kinetica GPU-Accelerated Real-Time Database

Kinetica DB, Inc., formerly known as GPUdb, lists its office at Leesburg Pike in Vienna, Virginia. Its co-founders first built a real-time tracking system for US Army intelligence and the NSA, and the company now sells a distributed database that uses GPUs to analyze very large, fast-changing datasets without ETL or pre-aggregation.

One engine handles SQL, vector search, graph, time-series and geospatial workloads, with tiered storage across GPU memory, RAM, disk and cold storage. Users query in SQL-92 or natural language converted to SQL, through JDBC, ODBC, REST and SDKs for Python, Java and JavaScript, and AI agents can combine vector, graph and spatial retrieval in a single query plan. It is offered as Kinetica Cloud, a free Developer Edition and an Enterprise Edition. Logos on the site include NVIDIA, Citi, Verizon, USPS and the FAA, and Lippo Group Digital's OVO in Indonesia is a published customer.

• GPU acceleration: Massively parallel query execution on NVIDIA GPUs with tiered storage.
• Geospatial analytics: Native spatial functions for tracking, routing and location joins.
• Vector and graph: HNSW vector search plus PGQ and Cypher-compatible GQL graph queries.
• Time-series and streaming: Continuous ingest from Kafka with sub-50ms queries on billions of records.
• Natural language to SQL: Prompts are converted into single SQL statements for analysts and agents.
• Enterprise operations: Hot backups, column compression, Kafka-based replication and OIDC single sign-on.

A Vietnamese telecom operator can analyze network events and subscriber locations in real time to detect congestion and plan capacity. A logistics or ride-hailing company can track vehicles and optimize routes using GPU-accelerated spatial queries. A retail conglomerate can unify loyalty, payment and store data into one customer view for real-time offers, as OVO did in Indonesia.

Organizations can join streaming, location and transactional data in sub-second queries on billions of rows, the vendor citing ingest of 800,000 records per second on an 8-node cluster while serving more than 40 dashboards. Consolidating vector, graph, time-series and spatial workloads in one engine reduces the number of databases to run. Vendor benchmarks claim spatial queries up to 240x faster than PostGIS and graph queries 6x faster than Neo4j. A free Developer Edition runs on Docker without a credit card.

Telecom networks, fleets, sensors and digital payments generate location-tagged data at a scale that CPU databases struggle to analyze in real time. GPU infrastructure is spreading rapidly for AI, and organizations want to reuse it for analytics and agent retrieval. In Southeast Asia, conglomerates, telecom operators and smart city programs increasingly combine customer, transaction and location data for real-time decisions.

Kinetica DB, Inc., formerly GPUdb, lists its office at 8219 Leesburg Pike, Vienna, Virginia. It carries an Accelerated by NVIDIA partner badge, released a Confluent-certified Kafka connector and was awarded a US Air Force Digital Directorate contract with a 100 million USD ceiling.