Hazelcast

Hazelcast is a distributed in-memory data grid and streaming platform founded in 2008 by Talip Ozturk, Fuad Malikov, and Enes Akar.

Updated 6 Jul 2026 Reviewed by 7wData

Visit Website

Profile

Unified in-memory platform combining distributed caching, stream processing, and real-time computation in Java for enterprises needing sub-millisecond data access.

Hazelcast is a distributed in-memory data grid and streaming platform founded in 2008 by Talip Ozturk, Fuad Malikov, and Enes Akar. The company released its first open-source implementation in early 2009 under an Apache 2.0 license. The platform provides sub-millisecond data access, stream processing, and distributed computing—combining in-memory storage with real-time computation in a single Java-based runtime. Hazelcast enables enterprises to process data in motion before it reaches a database, critical for low-latency applications in finance, e-commerce, and supply chains.

The company, headquartered in Palo Alto with an R&D center in Istanbul, raised $88.4M across eight funding rounds, with the most recent being a $50M Series D expansion in February 2020 from investors including Bain Capital Ventures, C5 Capital, Earlybird Venture Capital, and EBRD. Kelly Herrell served as CEO from August 2018 until 2024; Greg Luck, who joined as CTO in 2014 and became CEO in June 2014, later returned to the role. The company operates with approximately 156 employees as of late 2025 and generated $9.2M in revenue in 2024, growing from $7.3M in 2023.

Hazelcast competes with Redis and Apache Kafka by offering a unified platform that handles both caching and stream processing—capabilities Redis lacks—and distributed architecture that scales beyond single-node deployments. In May 2025, Hazelcast partnered with IBM to bring real-time data processing to IBM LinuxONE and Linux on Z mainframes, targeting use cases like instant payments and fraud detection. In December 2025, Forrester Research named Hazelcast a Strong Performer in streaming data platforms, recognizing its extensibility and fault tolerance for low-latency workloads. The company is now focused on real-time machine learning inference, positioning itself as a feature serving layer where predictions execute on live data rather than batch processing.

Who buys this

Global financial services (banks, trading firms, payment processors) requiring instant caching and fraud detection
Retailers and e-commerce platforms processing high-velocity transaction and inventory data
Supply chain and logistics companies with real-time visibility requirements
Enterprises modernizing mainframe workloads with real-time capabilities

Publicly disclosed clients

JPMorgan Chase
Deutsche Bank
Credit Suisse
HSBC
BNP Paribas
Swedbank
Amazon

Strengths and what to watch

Strengths

Unified architecture eliminating separate caching and streaming layers, reducing customer infrastructure footprint by ~50%
Multi-threaded Java implementation with distributed fault tolerance and auto-discovery, scaling beyond single-node Redis competitors
Recognized by Forrester Q4 2025 for extensibility and fault tolerance in low-latency streaming; endorsed by IBM for mainframe real-time processing

Watch for

Revenue growth at $9.2M (2024) remains modest relative to $88.4M raised; no disclosed IPO or exit timeline despite 16-year history
Narrow customer concentration risk: 45% of 1,597 users are US-based; exact percentage from Fortune 500 vs. mid-market not disclosed
Shifting leadership: Kelly Herrell CEO 2018–2024; Greg Luck returned to CEO role; unclear product strategy continuity amid real-time ML pivot

Recent moves

Key Information

Industry: Streaming/In-Memory
Founded: 2008
Headquarters: Palo Alto

Frequently Asked Questions

What is Hazelcast?

Hazelcast is a distributed in-memory data platform founded in 2008 that combines caching, stream processing, and real-time computation. It enables enterprises to access data in sub-millisecond speeds, processing information in motion before it reaches databases. Built on Java, it serves finance, e-commerce, and supply chain sectors requiring ultra-low latency.

How does Hazelcast compare to Redis?

Hazelcast offers unified caching and stream processing in one platform, while Redis focuses on caching alone. Unlike Redis's single-node architecture, Hazelcast scales with distributed, multi-threaded Java implementation providing fault tolerance. Hazelcast reduces infrastructure footprint by approximately 50% by eliminating separate streaming layers.

What industries use Hazelcast?

Hazelcast powers instant fraud detection in banking, real-time inventory management in e-commerce, and visibility across supply chains. Financial institutions like JPMorgan Chase and Deutsche Bank use it for sub-millisecond caching and payment processing. It also enables modernizing mainframe workloads with real-time capabilities.

Does Hazelcast run on IBM mainframes?

Yes. In May 2025, Hazelcast partnered with IBM to bring real-time data processing to IBM LinuxONE and Linux on Z mainframes. The partnership targets use cases including instant payments and fraud detection, modernizing legacy mainframe systems with real-time computing capabilities while preserving existing infrastructure investments.

Was Hazelcast recognized by Forrester?

Yes. In December 2025, Forrester Research named Hazelcast a Strong Performer in its Wave for streaming data platforms. The recognition highlighted its extensibility and fault tolerance for low-latency workloads. This validates Hazelcast's industry leadership in real-time data processing for enterprises requiring computational reliability.

How does Hazelcast support machine learning?

Hazelcast is positioning itself as a feature serving layer for real-time machine learning inference. Instead of batch processing, predictions execute on live data, enabling models to make decisions instantly. This supports fraud detection, personalized recommendations, and other latency-sensitive AI applications across finance and e-commerce.

How Hazelcast compares

Direct head-to-head against 3 competitors. Picked by 7wData.

Positioning: Unified in-memory platform combining distributed caching, stream processing, and real-time computation in Java for enterprises needing sub-millisecond data access.
Customer segments: Global financial services (banks, trading firms, payment processors) requiring instant caching and fraud detection
Strengths: Unified architecture eliminating separate caching and streaming layers, reducing customer infrastructure footprint by ~50%
Watch for: Revenue growth at $9.2M (2024) remains modest relative to $88.4M raised; no disclosed IPO or exit timeline despite 16-year history
Recent moves: Hazelcast Named a Strong Performer in Forrester Wave: Streaming Data Platforms, Q4 2025

Positioning: In-memory data store repositioned as real-time AI agent memory layer; surpassed $300M ARR in 2025.
Customer segments: Enterprise developers, financial services, and e-commerce teams needing sub-millisecond cache and AI context retrieval.
Strengths: Present in 43% of enterprise AI agent stacks; unified cache, vector, and feature-store operations in a single runtime.
Watch for: March 2024 SSPL license change drove 70% of users to evaluate alternatives; Valkey fork now production-ready under Linux Foundation, backed by AWS and Google.
Recent moves: Acquired Featureform, October 9, 2025, adding ML feature store for real-time structured data in AI production workflows.

Positioning: Distributed in-memory computing for HTAP workloads; absorbed into MariaDB following March 2026 acquisition.
Customer segments: Enterprises requiring co-located SQL analytics on live operational data, particularly in financial services and retail.
Strengths: Native ANSI SQL over distributed in-memory data, enabling transactional and analytical queries on the same dataset without extraction.
Watch for: MariaDB acquisition completed March 2026; standalone GridGain roadmap continuity and enterprise pricing post-acquisition remain unconfirmed.
Recent moves: Acquired by MariaDB plc, completed March 24, 2026, targeting an AI-ready unified transactional and analytical platform.

Positioning: Commercial event streaming platform built on Apache Kafka; wholly owned by IBM as of March 2026.
Customer segments: Enterprises streaming operational data for real-time analytics, financial services, and supply chains; now bundled into IBM enterprise accounts.
Strengths: Managed Flink stream processing with Flink ARR growing approximately 3x in two quarters of 2025; native stream governance tooling.
Watch for: IBM acquisition completed March 17, 2026; Confluent delisted from Nasdaq, with product priorities now subject to IBM roadmap control.
Recent moves: IBM completed $11B acquisition March 17, 2026; Confluent became an IBM wholly owned subsidiary and delisted from Nasdaq.

Sources

hazelcast.com — Company overview, products, customers, cloud deployments
getlatka.com — 2024 revenue ($9.2M), employee headcount (189), 2023 revenue ($7.3M)
www.prnewswire.com — Series D funding details, $50M raised February 2020, investors (EBRD, DEG, C5 Capital, Bain Capital Ventures, Earlybird Venture Capital, Capital One Growth Ventures)
www.businesswire.com — Kelly Herrell appointed CEO August 2018
www.prnewswire.com — Greg Luck appointed CEO June 2014, Talip Ozturk became CTO and board member
venturebeat.com — Hazelcast founded 2008 by Turkish engineers, open-source IMDG for Java
www.prnewswire.com — Forrester Wave Q4 2025 recognition, CEO Eric Bochner quote, real-time AI/ML inference strategy
www.dbta.com — IBM LinuxONE partnership May 2025, mainframe real-time data processing, use cases (instant payments, fraud detection)