FemtoAI Reports 5X Growth in H1 2026 as SPU AI Inference Platform Expands Across Samsung, Marshall and Edge AI Markets

FemtoAI, a leading AI inference platform and creator of the Sparse Processing Unit (SPU), announced 5X growth in 1H 2026, driven by engagements with Samsung and Marshall, new customer use cases and increased developer adoption.

The demand for AI inference globally continues to outpace supply, and femtoAI is filling a gap. With 200K+ chips already deployed, and customers across data centre, enterprise and consumer electronics, femtoAI delivers AI inference solutions that consume 100X less power & energy with 10X less memory to make AI economically viable on every device.

"We are proud that we are winning and scaling across several markets because of our unique acceleration of sparsity in the AI stack to reduce memory and power without compromising performance," said Sam Fok, co-founder and CEO at femtoAI. "From our initial wins with our first-generation SPU with companies such as NewSound and now Marshall, to new use cases such as fault monitoring for AI factories, to smart glass and robotics, femtoAI is solidifying its place as one of the leading solutions for on-device, low-power, low-memory AI on every device. Our next-generation SPU will be even more groundbreaking."

Highlights for femtoAI in 2026 include:

Customer Growth

• Over 200K+ SPU-001 chips currently in market, underscoring the growing market demand for a differentiated AI inference solution
• Six tier-1 next-generation smart glass vendors, including Orka, have selected femtoAI as their on-device AI solution
• AI-enhanced audio will reach more consumers as femtoAI and Marshall partner across premium headphone and earbud families
• 5-year partnership with ABOV Semiconductor and Samsung continues to evolve across a variety of home appliances

Developer Demand & Engagement at developer.femto.ai:

• Increased developer engagement with nearly half of all customers building their own algorithms and AI models in the femtoAI developer portal
• femtoAI partners directly with developers to uncover new inference solutions, both locally and in the cloud, such as a new data centre energy management system, robotics and vision denoising, among others

Product Innovation:

• Launch of femtoAI ClaraCall 3.0 and ClaraCall 3.0-S models for next-generation AI call quality enhancement, available on the femtoAI developer platform
• The next-generation SPU is currently underway to meet increasing customer demand for low-energy, low-power, and low-memory AI inference
• Decrease memory usage via model compression, resulting in 10X less memory required for open-source models such as OpenAI Whisper and custom solutions like local voice commands

Recognition & Awards:

• femtoAI was awarded the CES 2026 Best-in-Show award for SPU-001
• Named a cutting-edge AI inference solution by Bessemer Venture Partners
• Honoured as EDGE AI Foundation "Rookie of the Year" 2026

Dual Sparsity is the Key Ingredient
femtoAI is able to reduce memory by up to 10X and increase energy efficiency by up to 100X with its SPU Platform by applying a unique concept: dual sparsity. Sparsity is a known process of reducing unnecessary compute and storage, enabling the fastest and most efficient path to task completion. femtoAI's dual sparsity approach delivers these gains by leveraging this concept simultaneously at both hardware (chip) and software (tooling, applications) levels to ensure minimal unnecessary processing occurs when completing a task.

Swetha Srinivasan, Author of The Thesis, a technology and semiconductor newsletter, recently stated, "Getting to the next level means looking past the standard playbook of just scaling compute. What's exciting about femtoAI is the brain-inspired bet on sparsity, and the fact that they're solving it at every layer of the stack: sparsifying the model itself and designing silicon specifically for sparse compute. That full-stack alignment is a holistic approach that AI actually needs."