Enfabrica

What Enfabrica does

Enfabrica is a networking silicon and software company founded in 2019 and headquartered in Mountain View, California, specializing in infrastructure components for AI data centers. The company's core technology is the patented Accelerated Compute Fabric (ACF) architecture, which delivers high-throughput, fault-tolerant, multi-path networking designed specifically for AI workloads. Its flagship product is the ACF-S SuperNIC, which provides 3.2 Tbps throughput with multi-port 800GbE, PCIe Gen5, and CXL 2.0+ interfaces, engineered to connect 100,000+ GPUs as a unified supercomputer and support up to 524,288 accelerators on a 2-layer switched network. In July 2025, Enfabrica introduced EMFASYS (Elastic Memory Fabric System), the first commercially available Ethernet-based memory fabric for AI inference, which uses the same ACF-S silicon to pool up to 18 TB of DDR5 DRAM via CXL, enabling dynamic offloading from expensive GPU HBM with microsecond-level latency.\n\nEnfabrica's revenue model centers on hardware sales of its ACF-S SuperNIC chips and EMFASYS systems to enterprise customers, with an enterprise field sales motion targeting major AI cloud providers and hyperscalers. Pricing is custom B2B and not publicly disclosed. At the time of acquisition, products were in pilot deployment with major AI cloud providers, with broader commercial availability of ACF-S expected in Q1 2025. The company raised approximately $240 million across two rounds — a $125M Series B in September 2023 and a $115M Series C in November 2024 led by Spark Capital — with strategic backing from Arm Holdings, Cisco Investments, Samsung Catalyst Fund, and Nvidia (prior investor).\n\nIn September 2025, Nvidia acquired Enfabrica in a $900M+ acquihire deal, hiring CEO Rochan Sankar and licensing the company's AI networking technology. The acquisition represented one of Nvidia's largest acquihire transactions and was positioned to extend Nvidia's strategy from selling processors into offering complete AI infrastructure solutions.

Name

Enfabrica

Legal name

Enfabrica Corporation

Website

https://enfabrica.net

Company type

Private

Founded year

2019

Operating status

Acquired

Headcount range

101–250 employees

Short description

Enfabrica develops high-throughput AI networking silicon (ACF-S SuperNIC, 3.2 Tbps) and the first Ethernet-based memory fabric (EMFASYS) for hyperscaler AI data centers, targeting major AI cloud providers. Acquired by Nvidia in September 2025 for $900M+.

Ownership category

akta.pro rank

Product category

AI Networking Silicon

NAICS

Semiconductor and Related Device Manufacturing (334413)

SIC

Computer Communications Equipment (3576)

akta.pro primary industry

AI Networking for Clusters (InfiniBand/Ethernet/RDMA) (HDAAAAAE)

Keywords

• AI networking silicon
• Data center networking
• GPU interconnect chips
• SuperNIC hardware
• Memory fabric systems

Headquarters

HQ city

Mountain View

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

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

Revenue model

1. AI Networking Silicon Hardware: Enfabrica generates revenue through the sale of its ACF-S SuperNIC chips and related networking hardware to AI data center operators. The company was ramping up production of ACF SuperNIC chips with initial availability expected in Q1 2025.

Go-to-market motion1 record

Distribution channels1 record

Marketing channels5 records

Core offering

Enfabrica designs and sells AI networking silicon hardware built on its patented Accelerated Compute Fabric (ACF) architecture. Its flagship ACF-S SuperNIC delivers 3.2 Tbps throughput with multi-port 800GbE, PCIe Gen5 and CXL 2.0+ interfaces to connect 100,000+ GPUs as a unified supercomputer, and its EMFASYS (Elastic Memory Fabric System) uses the same silicon to pool up to 18 TB of DDR5 DRAM for AI inference memory offload. Products are sold to enterprise AI cloud providers and hyperscalers.

Product overview

Enfabrica offered a unified portfolio of AI infrastructure solutions centered on two core products: the ACF-S SuperNIC for high-performance AI networking and the EMFASYS for elastic AI memory fabric. The ACF-S provides 3.2 Tbps throughput for GPU connectivity and rack-scale networking, while EMFASYS leverages the same silicon to pool up to 18 TB of DDR5 DRAM via CXL and Ethernet for cost-effective AI inference memory offloading. Both products use the proprietary Accelerated Compute Fabric (ACF) architecture to enable fault-tolerant, congestion-free data movement across large GPU clusters. The company was acquired by Nvidia in September 2025 for over $900 million.

Differentiator

Problem solved

Functional benefit

Brands

• ACF-S: World's highest throughput AI SuperNIC with 3.2 Tbps throughput, multi-port 800GbE, PCIe Gen5 and CXL 2.0+ interfaces, based on Enfabrica's patented Accelerated Compute Fabric architecture.
• EMFASYS

Products and services

• ACF-S SuperNIC (Accelerated Compute Fabric SuperNIC) The world's highest throughput AI SuperNIC achieving 3.2 Tbps throughput with 8x elastic bandwidth for GPUs. Features multi-port 800GbE, PCIe Gen5 and CXL 2.0+ interfaces based on Enfabrica's patented Accelerated Compute Fabric (ACF) architecture, designed to connect 100,000+ GPUs as a unified supercomputer for AI data centers and hyperscale AI training clusters.
• EMFASYS (Elastic Memory Fabric System) The industry's first RDMA Ethernet-based AI Memory Fabric architecture that offloads GPU, HBM, and local head node DDR resource consumption. Based on the 3.2 Tbps ACF SuperNIC silicon, it elastically connects up to 144 CXL memory lanes per system to pooled memory targets up to 18 TB CXL DDR5 DRAM, enabling dynamic offloading from expensive GPU HBM to more affordable DRAM with microsecond-level access latency for AI inference workloads.

Quantifiable outcome

• CapEx reduction of up to 29%, OpEx reduction of up to 55%
• +2 more outcomes

Named customers1 record

Segments1 record

Ideal customer profiles1 record

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Feature3 records

Scale indicators5 records

Recent moves5 records

Expansion highlights5 records

Direct peers

• Broadcom: Broadcom is a leading supplier of networking silicon — including Ethernet NICs, switches, and custom AI ASICs — for hyperscale data centers and AI clusters. It competes head-to-head with Enfabrica's ACF-S SuperNIC for high-throughput AI cluster connectivity and serves the same hyperscaler buyer set.
• Marvell Technology: Marvell provides networking connectivity silicon — including Inphi-based DSPs, NVMe-oF controllers, and high-speed Ethernet silicon — that is deeply embedded in AI and cloud infrastructure. Its portfolio overlaps directly with Enfabrica's AI networking and CXL-oriented roadmap and addresses the same hyperscaler customer base.
• Nvidia Networking (Mellanox): Nvidia's Mellanox-derived ConnectX and Spectrum-X platforms are the dominant InfiniBand and Ethernet networking solutions for AI clusters. Enfabrica's ACF-S SuperNIC directly competes with ConnectX-7/8 and Spectrum-X for GPU cluster connectivity and inference networking.
• Astera Labs: Astera Labs designs connectivity silicon for AI and cloud infrastructure — including PCIe retimers, CXL controllers, and Ethernet connectivity ICs — overlapping directly with Enfabrica's CXL and high-speed Ethernet portfolio. Both companies target hyperscaler AI infrastructure customers with similar connectivity silicon.
• Cornelis Networks: Cornelis Networks (formerly Intel's Omni-Path architecture) provides high-performance networking for HPC and AI workloads, including ultra-scale fabric solutions. It overlaps with Enfabrica's positioning around high-throughput, low-latency, multi-path networking for very large GPU/HPC clusters.

Broad incumbents

• Cisco (Silicon One): Cisco's Silicon One networking ASIC portfolio targets hyperscale data center and AI routing/switching applications, including AI back-end networks. While part of Cisco's broader networking business, Silicon One competes with Enfabrica's ACF architecture in high-radix, high-throughput switching silicon for AI clusters. Cisco Investments was also a backer of Enfabrica.
• AMD (Pensando): AMD's Pensando DPU/data processing unit portfolio provides networking, storage, and security offload for cloud and AI infrastructure. While broader in scope, Pensando competes with Enfabrica's networking silicon for hyperscaler AI server deployments.
• Microchip Technology: Microchip supplies a portfolio of networking ICs — including Ethernet PHYs, switches, and timing products — used broadly in data center and AI infrastructure. While less directly competitive on AI SuperNICs, Microchip overlaps with Enfabrica's high-speed interconnect portfolio for cloud and AI server platforms.

Emerging players

• XConn Technologies: XConn Technologies is an emerging startup building CXL switch and fabric silicon for memory pooling and disaggregation in AI and cloud servers. It overlaps with Enfabrica's EMFASYS memory fabric thesis and shares the same CXL-based vision of decoupling memory from compute.
• MemVerge: MemVerge builds memory pooling and tiering software/systems for AI and big-data workloads, targeting similar memory-fabric and DRAM-tiering use cases that EMFASYS addresses from the silicon layer. Adjacent ecosystem participant in the AI memory fabric category.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat3 records

Key risks6 records

Key highlights6 records

Customer concentration

• Xx.com/enfabrica_
• LinkedInlinkedin.com/company/enfabricacorporation

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds2 records

Investors15 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 Enfabrica