Femtosense
femtoAI (formerly Femtosense) builds a full-stack edge AI inference platform centered on its proprietary Sparse Processing Unit (SPU-001) chip, delivering sub-mW AI for hearing aids, hearables, smart home appliances, security cameras, and robotics. Founded 2018 out of Stanford's Brains in Silicon Lab.
- Company typePrivate
- Founded2018
- HeadquartersPalo Alto, United States
- Headcount11–50
- GTM typeB2B
- OfferingHardware or Manufacturing
What Femtosense does
femtoAI (formerly Femtosense) is a Silicon Valley-based fabless semiconductor company founded in 2018 by researchers from Stanford University's Brains in Silicon Lab. The company designs a full-stack edge AI inference platform anchored on its proprietary Sparse Processing Unit (SPU-001), a 22nm AI co-processor (1.6mm x 2.2mm, sub-mW power, 1MB on-chip SRAM with 10MB effective via sparsity) that applies neuromorphic-inspired sparsity to deliver claimed 100x lower power consumption, 10x smaller memory footprint, and up to 10x larger model capacity versus standard edge AI solutions. The product stack spans silicon (SPU-001), the AI-ADAM-100 system-in-package (a joint SiP with ABOV Semiconductor), the Femtocrux compiler and Femtodriver simulator, a Python SDK, a developer portal, and a library of pre-trained models including Clara AI speech enhancement, voice interface, sound event detection, and a hearing aid suite.
The company targets ultra-low-power always-on AI use cases where traditional NPUs or cloud inference are impractical — hearing aids (notably Xiamen NewSound's AI-powered OTC hearing aids on Amazon), TWS earbuds and smart glasses, voice-controlled home appliances, security cameras, and battery-constrained robotics. Revenue is generated through hardware sales of SPU chips and evaluation kits, SDK and pre-trained model licensing, and embedded OEM partnerships, distributed via direct enterprise sales, the five-year strategic partnership with ABOV Semiconductor that co-packages SPU-001 with ABOV MCUs for global consumer-electronics OEMs, and a self-service developer portal.
femtoAI has raised approximately $12M-$16M+ across a 2018 seed, a 2022 $8M Series A led by Fine Structure Ventures (with Kleiner Perkins, Draper Associates, J2 Ventures, Quest Venture Partners, VanEdge Capital, Gaingels, Amino Capital participating), a 2023 strategic investment from ABOV Semiconductor, a $4.26M round in Nov 2024, and an undisclosed round closed in Dec 2025. The company rebranded from Femtosense to femtoAI in June 2025, crossed 100,000 cumulative SPU units shipped by end of 2025, and won CES 2026 Best-in-Show recognition for the SPU-001. The firm is founder-controlled, headquartered in San Bruno, California, and operates with 11-50 employees.
Femtosense firmographics
Firmographics- Name
- Femtosense
- Legal name
- Femtosense, Inc.
- Website
- https://femto.ai
- Company type
- Private
- Founded year
- 2018
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- femtoAI (formerly Femtosense) builds a full-stack edge AI inference platform centered on its proprietary Sparse Processing Unit (SPU-001) chip, delivering sub-mW AI for hearing aids, hearables, smart home appliances, security cameras, and robotics. Founded 2018 out of Stanford's Brains in Silicon Lab.
- Ownership category
- akta.pro rank
Femtosense industry classification
Industry- Product category
- Edge AI Inference Silicon
- NAICS
- Electromedical and Electrotherapeutic Apparatus Manufacturing (334510), Printed Circuit Assembly (Electronic Assembly) Manufacturing (334418)
- SIC
- Electromedical & Electrotherapeutic Apparatus (3845), Measuring & Controlling Devices, Nec (3829)
- akta.pro primary industry
- Edge AI Hardware & NPUs (SoCs, modules, accelerators) (HDAAAJAC)
- akta.pro secondary industries
- AI Inference Hardware & Edge Accelerators (HDAAAAAC), AI/Neural Processing Units (NPUs) & ML Accelerators (HDAHAJAF), On-Device Inference Runtimes & SDKs (mobile/embedded) (HDAAAJAB), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)
Keywords
Where Femtosense is headquartered
LocationHeadquarters
- HQ city
- Palo Alto
- HQ country
- United States
- HQ region
- North America
Offices2 records
Markets served
Femtosense business model
Business model- GTM type
- B2B
- Offering type
- Hardware or Manufacturing
- Cost components
- Technology or R&D, Personnel, Supply Chain, Marketing or Sales, Operations
Revenue model
- Hardware Sales (SPU Chips): Sale of Sparse Processing Unit (SPU-001) AI inference accelerator chips to consumer electronics manufacturers. Chips are sold directly to OEMs and through semiconductor partners like ABOV Semiconductor.
- Software Development Kit (SDK) Licensing: Licensing of SDK containing model optimization tools, compiler, and performance simulator. Developers pay for access to tools for pruning, quantizing, and deploying sparse AI models.
- Pre-trained AI Models: Sale of pre-trained, pre-optimized, and pre-compiled AI models for voice interfaces, speech enhancement, sound event detection, and hearing aid applications. Ready-to-deploy models for specific use cases.
- Evaluation Kits and Boards: Sale of evaluation boards (EVB) and evaluation kits (EVK2) for system integration testing. PCB with SPU-001 WLCSP for integrating with external host systems.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Multi-year contract | Custom enterprise pricing based on volume and integration requirements |
Go-to-market motion4 records
Distribution channels4 records
Marketing channels6 records
Femtosense product offering
Product offeringCore offering
femtoAI builds and sells a full-stack AI inference platform for extreme-edge devices, centered on the Sparse Processing Unit (SPU-001) AI accelerator chip. The platform includes pre-trained production-ready AI models, a sparse model optimization SDK with compiler and simulator, and reference evaluation hardware, and is packaged with ABOV Semiconductor MCUs as the AI-ADAM-100 system-in-package for voice-based consumer electronics and home appliances.
Product overview
femtoAI (formerly Femtosense) offers a full-stack edge AI inference platform combining proprietary sparsity-exploiting AI accelerator hardware with optimized models and development tools. The core SPU-001 chip delivers up to 100x lower power consumption through neuromorphic-inspired sparsity acceleration, while the AI-ADAM-100 (with ABOV Semiconductor) provides a system-in-package solution for voice-based consumer applications. The embedded platform includes ready-to-deploy AI models (Voice Interface, Speech Enhancement, Sound Event Detection, Hearing Aid Suite), the SDK with Femtocrux compiler and Femtodriver simulator, and evaluation hardware (EVB, EVK2). Clara AI Speech Enhancement technology provides deep neural network-based noise reduction for hearing aids and earbuds.
Differentiator
Problem solved
Functional benefit
Products and services
- SPU-001 (Sparse Processing Unit) AI Accelerator First-generation AI co-processor chip (WLCSP package) with 1MB on-chip SRAM (10MB effective with sparsity), 22nm process, sub-mW power consumption for speech/audio workloads, SPI interface, and a 1.6mm x 2.2mm footprint. Achieves up to 100x lower power consumption and 10x larger model capacity through dual-sparsity (weight and activation) acceleration, and is sold to OEMs integrating on-device AI in hearing aids, earbuds, wearables, home appliances, and IoT devices.
- AI-ADAM-100 System-in-Package AI MCU system-in-package jointly developed with ABOV Semiconductor, integrating the SPU-001 with an ABOV MCU. Enables on-device voice-based control, 'say what you mean' natural language processing, and voice-cleanup capabilities for home appliances and consumer electronics, delivering up to 100x improvement in AI model complexity without increasing power or cost.
- Software Development Kit (SDK) Full software development platform containing sparse model optimization tools (prune, quantize, sparsify), the Femtocrux custom compiler, the Femtodriver fast performance simulator, and a Python API for power/latency/footprint estimation. Supports PyTorch models through a conversion, simulation, and deployment workflow.
- AI Model Collection Pre-trained, pre-optimized, and pre-compiled production-ready AI models for deployment on the SPU. Categories include Voice Interface (wake word detection, natural language understanding), Speech Enhancement (Clara AI), Sound Event Detection (baby cries, glass breaking, alarms), and the Hearing Aid Suite (wide dynamic range compression, adaptive feedback cancellation, beamforming, dynamic EQ, sound scene classification).
- Clara AI Speech Enhancement Deep neural network-based speech enhancement technology that separates speech from background noise, delivering improved speech quality and comprehension for hearing aids, earbuds, and smart glasses. Runs on SPU-001 with sub-mW power consumption via 27x quantization and pruning compression, achieving PESQ 1.57, STOI 0.799, and SI-SDR 7.74, with 7-13 dB background noise suppression while preserving speech from any direction.
- Evaluation Board (EVB) Printed circuit board with SPU-001 in WLCSP form factor for integrating with external host systems. Provides full pin access via PMOD header, 16 Mb onboard SPI Flash, 1.8-3.3V operation, and headers for power measurement, sold to product developers for system integration testing.
- Evaluation Kit 2 (EVK2) Evaluation board bundled with the Tympan open-source hearing aid platform and Tympan connector board. Includes a Cortex-M7 host processor with FPU, 2x onboard MEMS microphones, 3.5mm audio jacks, and a micro SD card reader for development and testing of hearing aid and audio AI applications.
Quantifiable outcome
- 100x lower power consumption compared to standard AI solutions
- +6 more outcomes
Companies that use Femtosense
Customer profileNamed customers4 records
Segments5 records
Ideal customer profiles4 records
Femtosense technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
Integration1 record
AI capability7 records
Feature5 records
Femtosense partnerships and signals
Strategic signalPartnerships
Three partnerships are on record, tiered core and minor.
- ABOV SemiconductorcoreFive-year strategic partnership to deliver ultra-efficient edge AI solutions for consumer electronics. femtoAI's SPU-001 is packaged alongside ABOV's MCU series, enabling brands to add AI applications such as human language interfaces and predictive diagnostics to devices at low cost and power efficiency. Follows initial collaboration in 2024.
- Edge AI FoundationminorPartnership with the Edge AI Foundation allows femtoAI to expand its global thought leadership in edge AI computing.
- ABOV SemiconductorcoreInitial collaboration to offer energy- and cost-saving AI for home appliances. Femtosense's SPU-001 integrated with ABOV MCUs enables ultra-low-power AI features ready for deployment in hearables, IoT products, and home appliances.
Scale indicators7 records
Recent moves6 records
Expansion highlights6 records
Femtosense competitors and assessment
Company assessmentMarket position
Competitive moat5 records
Key highlights6 records
Customer concentration
Femtosense social profiles
Digital presenceFemtosense financial estimates
Financial estimateRevenue estimate
Valuation estimate
Femtosense leadership team
Management profileNumber of profiles
Profiles11 records
Femtosense funding detail
Funding detailFunding overview
Funding rounds9 records
Investors9 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Femtosense M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Femtosense
What does Femtosense do?
femtoAI builds and sells a full-stack AI inference platform for extreme-edge devices, centered on the Sparse Processing Unit (SPU-001) AI accelerator chip. The platform includes pre-trained production-ready AI models, a sparse model optimization SDK with compiler and simulator, and reference evaluation hardware, and is packaged with ABOV Semiconductor MCUs as the AI-ADAM-100 system-in-package for voice-based consumer electronics and home appliances.
Is Femtosense a public or private company?
Femtosense is a private company. It is classified as venture growth investor backed and is currently operating.
When was Femtosense founded?
Femtosense was founded in 2018. It employs 11 to 50 people.
Where is Femtosense based?
Femtosense is headquartered in Palo Alto, United States, in the North America region.
How does Femtosense make money?
Four revenue lines are on record. Hardware Sales (SPU Chips) is the primary driver. The others are software Development Kit (SDK) Licensing, pre-trained AI Models and evaluation Kits and Boards.
Does Femtosense have an API?
No public API is recorded for Femtosense.
What industry is Femtosense in?
Femtosense's product category is Edge AI Inference Silicon. Its primary akta.pro industry code is HDAAAJAC, Edge AI Hardware & NPUs (SoCs, modules, accelerators), with a secondary code of HDAAAAAC, AI Inference Hardware & Edge Accelerators. Its NAICS code is 334510 and its SIC code is 3845.