Revolutionary AI Infrastructure
Infercom is powered by SambaNova's dataflow architecture — purpose-built for AI inference, delivering unprecedented performance and efficiency.
Up to 10x
Faster Inference
Up to 5x
Energy Efficient
Up to 25TB
Memory per Rack
Dataflow vs. GPU Architecture
Why purpose-built dataflow beats general-purpose GPUs for AI inference
Dataflow Architecture
Purpose-built for AI
Purpose-built for AI workloads, creating custom processing pipelines for entire computation graphs while minimizing data movement.
- ✓ Entire model resident in memory
- ✓ Data flows through operations without intermediate writes
- ✓ Operator fusion: hundreds of operations in single kernel
- ✓ Software-defined hardware optimizes for each workload
Traditional GPU
General-purpose design
General-purpose design requiring kernel-by-kernel execution creates bottlenecks for AI inference workloads.
- ✗ Kernel-by-kernel execution creates overhead
- ✗ Excessive data movement between processor and memory
- ✗ Memory bandwidth bottleneck limits performance
- ✗ Underutilization of compute resources
SN40L Reconfigurable Dataflow Unit
Built on TSMC's 5nm process with 1,040 compute cores per chip, delivering 638 BF16 TFLOPS per chip — 10.2 PetaFLOPS per rack.
SambaNova SN40L RDU — dual-die CoWoS package on TSMC 5nm
638
TFLOPS/Chip (BF16)
1,040
PCUs/Chip
10.2
PFLOPS/Rack
TSMC 5nm
Process
Three-Tier Memory per Chip
520MB
SRAM
Ultra-fast on-chip cache
64GB
HBM
High-bandwidth co-packaged memory
Up to 1.5TB
DDR
Off-package DIMM storage
Rack Configuration (SN40L-16)
16
RDU Chips
Up to 25TB
Total Memory
~10kW
Typical Power
Air
Cooled
World-Record Performance
Performance measured in output tokens per second. MiniMax and DeepSeek-V3.1 benchmarks from Infercom EU infrastructure. DeepSeek-R1 and gpt-oss from Artificial Analysis (SambaNova Cloud).
MiniMax M2.5
NEW — EU HostedHigh-performance multimodal model now hosted on Infercom's EU infrastructure. Independently measured at 400+ tokens/sec by Artificial Analysis.
EU Sovereign
DeepSeek-R1 671B
10x vs GPUThe world's largest reasoning model at unprecedented speed. Up to 10x faster than GPU-based providers.
671B params
DeepSeek-V3.1
EU Hosted128K context, full function calling, and JSON mode — one of Europe's most capable sovereign LLMs.
EU Sovereign
gpt-oss-120b
Fastest EU ModelOpenAI's open-source 120B parameter model. Exceptional throughput for high-volume sovereign workloads.
EU Sovereign
Sustainable AI Infrastructure
Up to 5x better energy efficiency than GPU-based inference
Lower Power Consumption
Typical 10kW per rack versus GPU racks consuming 40–50kW+ for equivalent workloads. Dataflow architecture requires fewer chips, translating to dramatic power savings.
Smaller Footprint
Dramatically reduced physical space, simplified cooling, and lower total infrastructure costs.
Air-Cooled Design
No liquid cooling infrastructure required. Standard air cooling simplifies deployment, reduces maintenance complexity, and lowers operational overhead.
"Not all tokens are created equal. The real value lies not in measuring tokens generated, but in the quality of intelligence delivered per unit of energy consumed."
SambaNova — "Intelligence per Joule"Advanced Model Capabilities
Massive Model Support
Run large models with hundreds of billions of parameters on a single rack. Support for Composition of Experts (CoE) with 100+ expert models hosted simultaneously.
Long Context Windows
Handle up to 164K token context windows (MiniMax M2.5) and 128K on most other models. Massive memory capacity enables document analysis, code generation, and reasoning tasks.
Millisecond Model Switching
Multiple models resident in memory simultaneously with millisecond switching latency — orders of magnitude faster than GPU systems. Perfect for agentic AI and multi-model workflows.
European Infrastructure
Hosted in Equinix Munich 4 — Tier III+ certified, carrier-neutral datacenter
SambaNova rack infrastructure — air-cooled, 10kW per rack
Munich-Based Hosting
All data and processing remains within German borders under EU jurisdiction.
No US Jurisdiction
Protection from CLOUD Act, PATRIOT Act, and foreign intelligence access.
AI Act Ready
Infrastructure prepared for EU AI Act requirements and compliance.
Tier III+ Certified
99.982% uptime guarantee with redundant power and cooling systems.
Learn More About the Technology
Explore in-depth resources from SambaNova and independent sources
SambaNova Blog
Intelligence per Joule
Why energy efficiency is the new benchmark for AI infrastructure. Includes independent performance data.
Read moreSambaNova Blog
Dataflow Architecture for Inference
How dataflow eliminates GPU bottlenecks and enables efficient multi-model hosting at scale.
Read moreSambaNova Blog
Why SN40L Is Best for Inference
Full chip specifications, benchmark data, and architectural deep-dive into the SN40L RDU.
Read moreSambaNova Blog
SambaNova & Infercom Partnership
Building Germany's premier sovereign inference cloud — our partnership story in SambaNova's own words.
Read morearXiv
SN40L Architecture Paper
Peer-reviewed technical paper with full chip specifications, CoE benchmarks, and memory hierarchy details.
Read moreArtificial Analysis
Independent Performance Benchmarks
Third-party verified inference speed, throughput, and quality metrics across all SambaNova models.
Read more