Speedata - The Analytics Processing Unit (APU)

What Speedata - The Analytics Processing Unit (APU) does

Speedata is an Israeli semiconductor company that has built the Analytics Processing Unit (APU), the first processor purpose-built for Apache Spark SQL, batch ETL, and AI data preparation workloads. Its flagship C200 APU pairs a custom CGRA-based Callisto ASIC on a PCIe Gen 5 accelerator card (16-lane, 32GB onboard RAM) with firmware running on a bare-metal RISC-V VeeR EH1 core. The architecture is designed to handle the branch-divergent, control-flow-heavy patterns of SQL analytics that GPUs cannot efficiently execute, and it drops into existing Spark environments with zero code changes on Kubernetes, YARN, or standalone Spark deployments.

The company's commercial model rests on three deployment paths: direct sales of a pre-configured 2U server with dual C200 cards, OEM integrations through Dell and HPE server configurations, and cloud delivery via Speedata Acceleration for AWS and the European sovereign cloud operated by Nebul. The free Workload Analyzer functions as a product-led growth engine, letting prospects upload Spark event logs to predict APU acceleration and TCO gains before purchase. Customer evidence is early but high-quality: an unnamed pharmaceutical company cut a multi-petabyte batch job from 90 to 8 hours by replacing 100 CPU servers with a single 4-APU node, and an unnamed global adtech firm reduced a 1,200-node analytics workload from 8 minutes 44 seconds to 48 seconds. Microsoft integrated the C200 APU into Microsoft Fabric, and the resulting CoddSpeed benchmark paper won the SIGMOD 2026 Best Industry Paper Award, documenting up to 259x speedups on TPC-H queries.

Speedata has raised $114M across a 2020 seed, a $55M 2021 Series A, and a $44M Series B in June 2025, and is led by a team with prior semiconductor exits including Provigent (acquired by Broadcom for $300M+) and Centipede. The company employs 51–100 people, headquartered in Haifa, Israel, and operates across Israel, the US, France, Germany, the Netherlands, Austria, and the UK. Revenue is not publicly disclosed.

Name

Speedata - The Analytics Processing Unit (APU)

Legal name

Speedata Ltd.

Website

https://speedata.io

Company type

Private

Founded year

2019

Operating status

Operating

Headcount range

51–100 employees

Short description

Speedata is an Israeli semiconductor company that builds the Analytics Processing Unit (APU), a purpose-built PCIe accelerator for Apache Spark SQL, batch ETL, and AI data preparation, selling to enterprise data teams via direct sales, Dell and HPE OEM channels, and cloud deployments.

Ownership category

akta.pro rank

Product category

Data Analytics Acceleration Hardware

NAICS

Electronic Computer Manufacturing (334111), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)

SIC

Electronic Computers (3571)

akta.pro primary industry

AI/Neural Processing Units (NPUs) & ML Accelerators (HDAHAJAF)

akta.pro secondary industries

AI Server Systems & HGX/Accelerator Platforms (HDAAAAAB), Graphics Processing Units (GPUs) & Parallel Accelerators (HDAHAJAE), GPU/Accelerator Compute Servers (HDABADAC)

Keywords

• Analytics processing unit
• Apache Spark acceleration
• AI data preparation
• Semiconductor ASIC accelerator
• Batch ETL hardware

Headquarters

HQ city

Netanya

HQ country

Israel

HQ region

Middle East

Offices1 record

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Technology or R&D, Personnel, Supply Chain, Marketing or Sales, Operations

Revenue model

1. Hardware Sales - C200 APU Cards: Sale of C200 APU accelerator cards, either as pre-configured 2U servers from Speedata or through OEM partners Dell and HPE server configurations. The hardware delivers up to 100x faster Apache Spark performance and 90% lower TCO.
2. Speedata Acceleration for AWS: Cloud-based acceleration service where customers keep data in S3, sync deltas to Speedata-hosted environment, accelerate ETL and AI data prep on the APU, and push processed data back to AWS.

Pricing tiers

Go-to-market motion4 records

Distribution channels8 records

Marketing channels7 records

Core offering

Speedata designs, manufactures, and sells the C200 Analytics Processing Unit (APU), a purpose-built ASIC accelerator card (Callisto chip on PCIe Gen 5) that accelerates Apache Spark SQL, batch ETL, and AI data preparation workloads by up to 100x compared with CPUs and GPUs. The product is delivered via pre-configured 2U servers, OEM configurations with Dell and HPE, and cloud-based acceleration through AWS and Nebul sovereign cloud, with the Workload Analyzer provided as a free planning tool.

Product overview

Speedata offers a purpose-built semiconductor product portfolio centered on the C200 Analytics Processing Unit (APU) — the first processor designed specifically for Apache Spark SQL, batch ETL, and AI data preparation workloads. The APU hardware is the Callisto ASIC on a PCIe Gen 5 accelerator card. Customers access the APU through three deployment paths: (1) a pre-configured 2U server with dual C200 accelerator cards sold directly by Speedata, (2) OEM server configurations through Dell and HPE, and (3) cloud options including Speedata Acceleration for AWS and Nebul's European sovereign cloud. Supporting tools include the Workload Analyzer (a free performance simulation CLI/browser tool for predicting APU gains on customer Spark workloads) and a live APU demo running real Apache Spark on silicon. The APU integrates directly into existing Apache Spark environments with zero code changes required.

Differentiator

Problem solved

Functional benefit

Products and services

• C200 Analytics Processing Unit (APU) The C200 is the first processor architected specifically for the data layer — Apache Spark, batch ETL, and AI data preparation pipelines. It uses a CGRA-based dataflow architecture with row processing mapped directly into hardware, achieving over one billion rows per second throughput. Packaged as a PCIe Gen 5 accelerator card (Callisto ASIC, 16-lane, 32GB onboard RAM) that drops into existing Apache Spark environments with zero code changes.
• Workload Analyzer A free tool that predicts performance and TCO gains when running Apache Spark workloads on the C200 APU. Operates in three modes: browser-based Spark event log upload, local CLI execution in the customer's environment, or TPC-DS benchmark comparison. Produces per-query acceleration reports, workload-level performance summaries, and breakdowns by compute, I/O, and overhead.
• Speedata Acceleration for AWS A cloud-based acceleration service where customers keep data in Amazon S3, sync deltas to a Speedata-hosted environment, accelerate ETL and AI data preparation on the APU, and push processed data back to AWS. Built for AWS-dependent customers seeking APU acceleration without on-premises infrastructure investment.
• Pre-Configured APU Server (2U, Dual C200 Accelerator Cards) A fully integrated 2U server bundled with two C200 APU accelerator cards, ready to run Apache Spark workloads out of the box. Sold directly by Speedata to enterprises as the fastest on-premises path to APU acceleration.
• Speedata APU via Dell Server Configurations Dell-branded server configurations embedding Speedata C200 APU accelerator cards, enabling enterprises to procure purpose-built analytics acceleration through their existing Dell data center supply chain.
• Speedata APU via HPE Server Configurations HPE-branded server configurations integrating Speedata C200 APU accelerator cards, bringing purpose-built analytics acceleration to enterprises through HPE's global infrastructure and support ecosystem.
• Speedata APU via Nebul European Sovereign Cloud Speedata APU deployed within Nebul's European sovereign cloud infrastructure, enabling European enterprises to run advanced analytics and AI data preparation workloads while maintaining full jurisdictional control over data under European legal authority.

Quantifiable outcome

• Up to 100x faster Apache Spark performance vs CPUs/GPUs
• +5 more outcomes

Named customers2 records

Segments1 record

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration1 record

AI capability3 records

Feature6 records

Partnerships

Seven partnerships are on record, tiered core and secondary.

• NebulcoreTechnology or Integration · 19 February 2026Strategic partnership to deploy Speedata's APU technology within Nebul's sovereign cloud infrastructure. Nebul is a European Private AI Neocloud provider, and will be the first to offer the APU in Europe, enabling enterprises to run advanced analytics and AI workloads while maintaining full jurisdictional control over their data under European legal authority. The partnership addresses surging European demand for AI data center capacity.
• DellcoreOEM/ Whitelabel/ Licensing PartnerSpeedata APUs are available through Dell server configurations, enabling enterprises to deploy purpose-built analytics acceleration within their existing data center infrastructure.
• HPE (Hewlett Packard Enterprise)coreOEM/ Whitelabel/ Licensing PartnerSpeedata APUs are available through HPE server configurations, bringing purpose-built analytics acceleration to enterprises through HPE's trusted infrastructure ecosystem.
• CarahsoftcoreChannel Partner/ Reseller/ DistributorUS public-sector distributor delivering Speedata's analytics acceleration hardware to government customers through its reseller network, contracts, and procurement expertise.
• Systemic.frsecondaryChannel Partner/ Reseller/ DistributorFrench value-added reseller providing consulting, integration, and support services for data infrastructure, helping enterprises deploy Speedata within their environments.
• KodatasecondaryChannel Partner/ Reseller/ DistributorEuropean distributor bringing Speedata's analytics acceleration hardware to resellers and system integrators across the region.
• MicrosoftcoreTechnology or IntegrationMicrosoft integrated Speedata C-200 APU into its Fabric architecture, validating the APU technology through their CoddSpeed benchmark paper which won SIGMOD 2026 Best Industry Paper. Microsoft tested the APU against TPC-H queries and documented up to 259x speedups. 16 Speedata engineers contributed to the project.

Scale indicators7 records

Recent moves6 records

Expansion highlights6 records

Emerging players

• Cerebras Systems: Cerebras builds wafer-scale AI accelerators (CS-3) for large model training and inference. Like Speedata, it is a venture-backed emerging custom-silicon AI infrastructure company pursuing a new processor category, but at much larger scale and focused on training rather than analytics.
• Groq: Groq develops LPU (Language Processing Unit) inference accelerators and pursues a similar custom-silicon, single-workload-optimization strategy. Both are venture-backed emerging players creating new silicon categories, though Groq targets inference while Speedata targets the upstream data preparation layer.
• SambaNova Systems: SambaNova builds dataflow architecture AI accelerators (RDU) using reconfigurable computing concepts similar to Speedata's CGRA approach. Both target data center AI workloads with custom silicon, though SambaNova focuses on training/inference while Speedata targets analytics — making them adjacent emerging competitors in the reconfigurable AI accelerator category.
• Graphcore: Graphcore's Intelligence Processing Unit (IPU) is another venture-funded emerging custom-silicon AI accelerator targeting data center workloads. Both companies aim to displace GPUs in specific workload categories through purpose-built architectures with different design philosophies.
• Tenstorrent: Tenstorrent develops RISC-V based AI accelerators and pursues a similar custom-silicon strategy targeting data center AI workloads. Both are emerging category creators, though Tenstorrent focuses on training/inference workloads and has pursued a different go-to-market (open hardware IP licensing).
• Hailo: Hailo builds AI processors initially for edge inference and expanding into data center. As an Israel-based emerging custom-silicon AI accelerator peer, Hailo shares Speedata's regional talent ecosystem and startup-stage venture backing profile.

Broad incumbents

• NVIDIA: NVIDIA's GPUs (H100, H200, B200) dominate AI infrastructure spend and are the primary alternative to Speedata for analytics acceleration. While not purpose-built for SQL, NVIDIA's CUDA ecosystem and aggressive roadmap make it the closest broad incumbent Speedata competes with on data center acceleration workloads.
• Intel: Intel competes through Xeon CPUs (the baseline Speedata displaces) and Habana Gaudi accelerators. As the incumbent CPU supplier and Speedata's VP R&D Dani Voitsechov's former employer, Intel represents both the legacy baseline and a potential competing accelerator pathway for analytics workloads.
• AMD: AMD's Instinct MI300/MI325 accelerators compete for the same data center AI and analytics budgets. AMD is an established incumbent with broader portfolio reach but, like NVIDIA, lacks a purpose-built analytics workload match against Speedata's CGRA architecture.

Others

• Databricks: Databricks is the dominant Apache Spark platform vendor and Speedata's primary ecosystem partner. While not a direct accelerator competitor, Databricks controls the workload routing decisions that determine whether customers adopt APU acceleration, making it a critical ecosystem enabler (and potential competitive threat if Databricks pursues its own silicon path).

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/Speedata1
• GitHubgithub.com/Speedata-io
• YouTubeyoutube.com/@SpeedataAPU
• Facebookfacebook.com/speedata.io
• LinkedInlinkedin.com/company/speedataio

Revenue estimate

Valuation estimate

Number of profiles

Profiles9 records

Funding overview

Funding rounds3 records

Investors9 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Speedata - The Analytics Processing Unit (APU)