What Sagence AI does

Sagence AI is a Santa Clara, California-based semiconductor company founded in 2018 (originally as Analog Inference) that develops analog in-memory computing (AIMC) accelerators for AI inference. The company's core architecture executes AI operations directly within multi-level non-volatile flash memory cells operating at deep subthreshold voltage levels, eliminating data movement between memory and compute units that constrains conventional digital GPU and CPU architectures. The flagship product is a modular chiplet-based accelerator that scales from edge deployments to multi-rack data-center configurations, paired with a software stack that imports pre-trained models from PyTorch, ONNX, and TensorFlow and pre-allocates neural network resources at compile time for deterministic, low-latency inference. Sagence AI reports performance claims of 100X lower MAC power, 20X lower price, and 10X smaller rack space versus leading GPU processing when normalized on Llama 3.1 70B inference at 666K tokens/sec.

The company's go-to-market is direct enterprise field sales targeting hyperscale data-center operators and enterprise data centers facing energy, cooling, performance, and capital-cost constraints on AI workloads, with secondary positioning in edge AI use cases including recommendation systems, defect detection, and video security. The business model is hardware sales of AI inference accelerators, with pricing quote-based and negotiated per deployment (no public pricing disclosed). Sagence AI sells through direct engagement and proof-of-concept demonstrations rather than channel or e-commerce distribution.

Sagence AI has raised US$58M in total funding from seed through Series B, including marquee Silicon Valley investors (Vinod Khosla via Khosla Ventures, Andy Bechtolsheim, Atiq Raza) and strategic deep-tech investors (TDK Ventures, Prosperity7/Aramco Ventures, Cambium Capital, New Science Ventures, Alumni Ventures). The company operates with 51–100 employees, was named to the EE Times 2025 Silicon 100 Top 25, and is led by founder and CEO Vishal Sarin, who holds 100+ patents in semiconductor technology, supported by a VP bench with experience at Intel, AMD, Hynix, GlobalFoundries, UMC, SanDisk, Cadence, Siemens EDA, and Altera/Xilinx. The company emerged from stealth in November 2024 and has not yet disclosed commercial customer wins.

Name

Sagence AI

Legal name

Sagence AI

Website

https://www.sagence-ai.com/

Company type

Private

Founded year

2018

Operating status

Operating

Headcount range

51–100 employees

Short description

Sagence AI is a Santa Clara-based semiconductor company building analog in-memory computing accelerators that execute AI inference inside multi-level non-volatile memory, targeting hyperscale and enterprise data centers with claimed 100X lower power and 20X lower cost versus GPUs.

Ownership category

akta.pro rank

Product category

AI Inference Accelerators

NAICS

Computer Storage Device Manufacturing (334112), Computer Systems Design and Related Services (54151)

SIC

Computer Storage Devices (3572)

akta.pro primary industry

AI Inference Hardware & Edge Accelerators (HDAAAAAC)

akta.pro secondary industries

AI Accelerators (GPUs/TPUs/NPUs/ASICs) (HDAAAAAA), AI/Neural Processing Units (NPUs) & ML Accelerators (HDAHAJAF), Edge AI Hardware & NPUs (SoCs, modules, accelerators) (HDAAAJAC), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)

Keywords

• Analog in-memory computing
• AI inference chips
• AI hardware accelerators
• Neural network accelerators
• Chiplet AI processors

Headquarters

HQ city

Santa Clara

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, Personnel, Supply Chain, Marketing or Sales, Infrastructure

Revenue model

1. AI Inference Hardware Sales: Sagence AI generates revenue through the sale of analog in-memory compute acceleration hardware. The company sells AI inference accelerators designed for data center and edge applications, offering hardware solutions that deliver high performance at significantly lower power and cost than GPU-based alternatives.

Go-to-market motion1 record

Distribution channels1 record

Marketing channels6 records

Core offering

Sagence AI designs and sells analog in-memory computing (AIMC) AI inference accelerator chips built on a modular chiplet architecture. The hardware executes neural network operations directly inside multi-level non-volatile memory cells, eliminating data movement between memory and compute, and is paired with a software stack that imports models from PyTorch, ONNX, and TensorFlow. The accelerators target generative AI inference at hyperscale and enterprise data centers, delivering 100X lower MAC power, 20X lower price, and 10X smaller rack space than GPU-based solutions.

Product overview

Sagence AI offers a unified hardware and software platform based on analog in-memory computing technology. The core product is the Sagence AI Analog In-Memory Compute Architecture, which uses chiplet-based accelerators that execute AI operations directly within multi-level non-volatile memory cells. The Sagence AI Software Stack supports model import from PyTorch, ONNX, and TensorFlow frameworks. The platform powers four vertical-specific solutions: Content Generation, Recommendation Systems, Defect Detection, and Video Security. Two deployment solutions address market segments: Hyper-scale Data Centers and Enterprise Data Centers. The architecture achieves 100X lower MAC power, 10X smaller rack space, and 5X lower price compared to GPU-based solutions, scaling from data center to edge applications.

Differentiator

Problem solved

Functional benefit

Products and services

• Sagence AI Analog In-Memory Compute Architecture Analog in-memory computing platform that executes AI operations directly within multi-level non-volatile memory cells, eliminating data movement between memory and compute. Delivers 100X lower MAC power and 20X lower cost versus GPU-based AI inference processing.
• Sagence AI Chiplet Accelerators Modular chiplet-based AI accelerator hardware that scales from single-chip edge deployments to multi-rack data center installations for generative AI inference workloads.
• Sagence AI Software Stack Software platform that imports pre-trained neural networks using standards-based interfaces (PyTorch, ONNX, TensorFlow), converts them into Sagence format, and pre-allocates resources at compile time for low, consistent latency without runtime scheduling delays.

Quantifiable outcome

• 100X lower MAC power consumption compared to traditional digital GPU/CPU architectures
• +4 more outcomes

Segments3 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability8 records

Feature5 records

Scale indicators4 records

Recent moves7 records

Expansion highlights6 records

Direct peers

• Cerebras Systems: Cerebras builds wafer-scale AI compute systems for training and inference, competing for the same hyperscale and enterprise AI compute budgets as Sagence AI with a different (digital wafer-scale) architecture.
• SambaNova Systems: SambaNova offers AI accelerator systems (RDU) for enterprise and hyperscale generative AI workloads, positioning itself on inference economics and ease of deployment in a manner directly comparable to Sagence AI's enterprise data center pitch.
• Groq: Groq develops LPU-based AI inference accelerators targeting low-latency, cost-efficient inference at scale - directly comparable to Sagence AI's value proposition of economic at-scale AI inference for hyperscalers and enterprises.
• Mythic AI: Mythic AI builds analog compute-in-memory AI inference chips for edge devices, making it the closest direct analog-AIMC peer to Sagence AI. Both pursue deep-subthreshold analog matrix multiplication to deliver low-power inference, though Mythic focuses primarily on edge while Sagence targets data center scale.
• Blaize: Blaize develops programmable AI inference accelerators for edge and data center applications. Sagence AI's VP of Strategy, Richard Terrill, has prior experience at Blaize, illustrating the close overlap in programmable inference silicon.
• Tenstorrent: Tenstorrent designs RISC-V-based AI accelerators for training and inference, targeting data center and edge customers with a comparable enterprise-AI infrastructure value proposition.

Emerging players

• Syntiant: Syntiant produces ultra-low-power neural decision processors for edge AI/audio/vision applications. While much smaller in scope than Sagence AI, it shares the design philosophy of always-on, low-power inference at the edge.

Broad incumbents

• Intel (Habana Labs): Intel's Habana Labs Gaudi accelerators are positioned for cost-efficient AI training and inference, representing an incumbent alternative for enterprises evaluating non-NVIDIA AI silicon at data center scale.
• AMD: AMD's MI300/MI325 series AI accelerators compete for the same hyperscale and enterprise inference workloads that Sagence AI targets, with a broader CPU/GPU portfolio that includes data center, embedded, and adaptive computing lines.
• NVIDIA: NVIDIA dominates the AI inference and training market with H100/B200 GPUs and the CUDA software stack. Sagence AI explicitly benchmarks against the NVIDIA B200 and positions its analog architecture as a 100X power / 20X cost alternative for LLM inference.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks6 records

Key highlights7 records

Customer concentration

• YouTubeyoutube.com/@sagenceai
• LinkedInlinkedin.com/company/sagence-ai

Revenue estimate

Valuation estimate

Number of profiles

Profiles6 records

Funding overview

Funding rounds4 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 Sagence AI