Enfabrica Corporation
Enfabrica connects GPUs and memory at multi-terabit speeds for AI hyperscalers.
Enfabrica develops specialized networking chips and systems that solve critical data movement bottlenecks for AI and machine learning workloads at hyperscale. Their Accelerated Compute Fabric (ACF) SuperNIC chip delivers multi-terabit Ethernet connectivity and memory disaggregation, enabling GPUs, CPUs, and memory to communicate 4x faster than competing solutions. Built by former Broadcom and Google infrastructure leaders, the company bridges PCIe/CXL memory semantics with RDMA networking to reduce AI processing costs and improve efficiency.
Problem solved
AI infrastructure teams face severe data movement bottlenecks when scaling GPU clusters, limiting throughput and increasing costs per compute unit.
Target customer
Hyperscale cloud providers, AI infrastructure companies, and data center operators building large-scale GPU clusters for machine learning and generative AI applications.
Founders
R
Rochan Sankar
Founder, President & CEO
26-year semiconductor veteran who led Broadcom's Data Center Ethernet switching business and brought four generations of Tomahawk/Trident chips to market; holds B.A.Sc. from University of Toronto and MBA from Wharton.
S
Shrijeet Mukherjee
Co-Founder, Chief Development Officer
Previously headed networking platforms and architecture at Google, recognized the need for infrastructure optimized for parallel, accelerated, and heterogeneous compute paradigms.
Funding history
Series A
$50M
2022
Led by Unknown
· Unknown
Series B
$125M
June 2023
Led by Atreides Management
· Nvidia, Sutter Hill Ventures, IAG Capital Partners, Liberty Global Ventures, Valor Equity Partners, Infinitum Partners, Alumni Ventures
Series C
$115M
November 2024
Led by Spark Capital
· Arm, Cisco Investments, Maverick Silicon, Samsung Catalyst Fund, VentureTech Alliance, Atreides Management, Alumni Ventures, IAG Capital, Liberty Global Ventures, Sutter Hill Ventures, Valor Equity Partners
Acquisition by Nvidia
$900M+
September 2025
Led by Nvidia
· N/A
Total raised:
$290M
Pricing
Not publicly available; enterprise hardware/system pricing model.
Notable customers
Not publicly disclosed; target includes major hyperscalers and AI infrastructure providers.
Integrations
Accton (SuperNIC system partnership), Ultra Ethernet Consortium (Technical Advisory Committee), Ultra Accelerator Link (UALink) Consortium, Fujitsu
Tech stack
GSAP (JavaScript frameworks)
React (JavaScript frameworks)
web-vitals (JavaScript libraries)
Microsoft ASP.NET (Web frameworks)
Webpack
Open Graph
Module Federation
Sitecore (CMS)
Magento (Ecommerce)
Google Analytics (Analytics)
Sentry (Issue trackers)
Google Font API (Font scripts)
PHP (Programming languages)
Google Workspace (Email)
MySQL (Databases)
Google Cloud (IaaS)
Google Cloud Trace (Performance)
Website
Competitors
Broadcom Tomahawk/Trident
General-purpose data center switching; Enfabrica specializes in GPU-optimized interconnect with 4x bandwidth advantage for AI workloads.
Nvidia BlueField
GPU-attached DPU for networking; Enfabrica's ACF SuperNIC combines PCIe/CXL bridging with RDMA for superior memory disaggregation capabilities.
Marvell Prestera
Cloud networking switch; lacks Enfabrica's specialized GPU cluster interconnect and memory offload functionality.
Why this matters: Enfabrica is notable because it identified and solved a fundamental infrastructure gap as AI workloads scaled beyond single-node limits—data movement became the bottleneck. The September 2025 Nvidia acquisition at $900M+ validates the strategic importance of specialized GPU interconnect silicon, and signals how critical distributed memory and networking architectures are to the next generation of AI infrastructure.
Best for: Hyperscalers and AI infrastructure teams building next-generation GPU clusters requiring multi-terabit bandwidth, memory disaggregation, and resilient interconnect.
Use cases
Large-scale GPU cluster interconnect
Training facilities with hundreds of GPUs need ultra-low-latency, high-bandwidth connections between compute nodes. Enfabrica's ACF SuperNIC delivers 3.2 Tbps with multipath resiliency, eliminating network as a bottleneck during distributed training jobs.
Remote memory pooling for AI
The EMFASYS system allows GPU servers to access abundant remote DRAM over high-speed networks, enabling disaggregated inference and fine-tuning workloads without over-provisioning local GPU memory, reducing per-unit infrastructure costs.
Heterogeneous compute architecture
Modern AI workloads mix GPUs, CPUs, and accelerators requiring efficient data movement between different compute types. Enfabrica bridges PCIe/CXL memory semantics with RDMA networking, allowing seamless integration of diverse processors in a single fabric.
Alternatives
Nvidia NVLink
Point-to-point GPU interconnect; limited to Nvidia ecosystems and lower port counts than Enfabrica's multi-vendor, multi-port approach.
Mellanox InfiniBand
Established HPC interconnect with lower bandwidth (400G vs. 3.2T) and not optimized for GPU-centric memory disaggregation workloads.
Standard Ethernet switches
Generic data center fabric; lacks GPU-aware optimizations, multipath resiliency, and memory semantics bridging critical for AI efficiency.
FAQ
What does Enfabrica do? +
Enfabrica designs specialized networking chips and systems that connect GPUs, CPUs, and memory at multi-terabit speeds. Their ACF SuperNIC chip delivers 4x the bandwidth and resiliency of competing GPU-attached NICs, while their EMFASYS system enables memory disaggregation for AI workloads. The technology bridges traditional PCIe/CXL memory protocols with RDMA networking to solve critical bottlenecks in hyperscale AI infrastructure.
How much does Enfabrica cost? +
Pricing is not publicly disclosed. Enfabrica sells enterprise-grade hardware and systems to hyperscalers and infrastructure providers; contact sales for custom quotes.
What are alternatives to Enfabrica? +
Nvidia NVLink (GPU-to-GPU interconnect with lower bandwidth), Mellanox InfiniBand (established HPC fabric with lower performance), and standard Ethernet switches (generic but lacking GPU optimization and memory semantics).
Who uses Enfabrica? +
Target customers are hyperscale cloud providers and AI infrastructure operators building large-scale GPU clusters. Specific customer names are not publicly disclosed, though partnerships with Accton and Fujitsu indicate OEM adoption.
Why was Enfabrica acquired by Nvidia? +
Nvidia's September 2025 acquisition of Enfabrica for $900M+ reflects the critical importance of GPU interconnect and memory fabric for AI scaling. The acquisition consolidates Enfabrica's ACF SuperNIC technology with Nvidia's ecosystem, enabling tighter integration of networking, compute, and memory for next-generation AI infrastructure.
Tags
GPU interconnect
AI infrastructure
data center networking
memory disaggregation
semiconductor
hyperscaler
distributed computing