What Untether AI does

Untether AI was a Toronto-based AI hardware company founded in 2018 that developed energy-efficient AI inference chips and accelerator cards. Its core technology centered on a proprietary in-memory computing architecture optimized for low-latency, high-performance AI processing at the edge and in data centers. The company's flagship product was the speedAI240 Accelerator Card, launched in late 2024 and positioned as the world's highest performing and most energy efficient AI accelerator according to MLPerf benchmark results, with competitive performance-per-watt against AMD and Nvidia in inference workloads.

The company pursued a horizontal go-to-market spanning edge AI, data center inference, and physical AI / robotics use cases. It combined direct enterprise sales with OEM-style partnerships through semiconductor companies including Ampere (Arm-based integration for energy-efficient inference) and Ola-Krutrim (next-generation AI acceleration for the Indian cloud and AI chip market). Its revenue model combined hardware sales of accelerator cards with technology licensing and partnership arrangements. Untether AI raised approximately $152 million across multiple funding rounds, including a $125 million Series A in July 2021 led by Tracker Capital Management and Intel Capital, with additional participation from CPP Investments and Radical Ventures.

Despite MLPerf benchmark recognition in 2024 and named enterprise engagements, Untether AI was unable to achieve commercial scale. AMD acqui-hired its engineering team in June 2025 as part of an ongoing AI inference capability build-out, and the company subsequently shut down operations and filed for bankruptcy in October 2025, citing lack of working capital with $128 million owed to creditors against approximately $25 million in assets. The company's operating status is closed.

Name

Untether AI

Website

https://untether.ai

Company type

Private

Founded year

2018

Operating status

Closed

Headcount range

101–250 employees

Short description

Untether AI was a Toronto-based AI chip startup that developed energy-efficient AI inference accelerators for data center, edge, and physical AI applications before shutting down in 2025 following an AMD acquihire and subsequent bankruptcy filing.

Ownership category

akta.pro rank

Product category

AI Inference Accelerators

NAICS

Computer and Peripheral Equipment Manufacturing (33411)

SIC

Electronic Computers (3571)

akta.pro primary industry

AI Inference Hardware & Edge Accelerators (HDAAAAAC)

akta.pro secondary industries

Edge AI Hardware & NPUs (SoCs, modules, accelerators) (HDAAAJAC), AI/Neural Processing Units (NPUs) & ML Accelerators (HDAHAJAF), Graphics & Accelerator Cards (GPUs, compute accelerators) (HDACAGAA)

Keywords

• AI inference chips
• AI accelerators
• Edge AI hardware
• Data center inference
• Neural network processors

Headquarters

HQ city

Toronto

HQ country

Canada

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, Operations, Marketing or Sales

Revenue model

1. AI Inference Hardware Sales: Sale of AI inference chips and accelerator cards (such as speedAI240) to enterprise customers for data center and edge AI deployments.
2. Technology Licensing/Partnerships: Technology partnerships with semiconductor companies (Ampere) for Arm-based AI inference solutions.

Go-to-market motion2 records

Distribution channels2 records

Marketing channels2 records

Core offering

Untether AI designs and sells energy-efficient AI inference chips and accelerator cards for low-latency, high-performance AI workloads at the edge and in data centers. Its flagship offering is the speedAI240 Accelerator Card, positioned as the world's highest performing and most energy efficient AI accelerator according to MLPerf benchmarks, alongside AI chips for physical AI applications in machines and robotics.

Product overview

Untether AI was a Toronto-based AI hardware company that specialized in energy-efficient AI inference chips and accelerator cards. The company's flagship product was the speedAI240 accelerator card, which was positioned as a high-performance, energy-efficient solution for AI inference workloads. The company developed in-memory computing technology for AI inference and targeted both data center and edge computing applications. Following the company's shutdown in 2025, AMD acquired Untether AI's engineering team through an acquihire agreement.

Differentiator

Problem solved

Functional benefit

Brands

• speedAI Accelerator Cards: World's highest performing, most energy efficient AI accelerators according to MLPerf benchmarks

Products and services

• speedAI240 Accelerator Card AI inference accelerator card designed for data center and edge deployment, marketed as the world's highest performing and most energy efficient AI accelerator according to MLPerf benchmarks. Targeted at enterprises deploying AI inference workloads.
• Physical AI Inference Chip AI inference chip designed for physical AI applications in various machines including robotics and autonomous systems requiring real-time low-power processing.

Quantifiable outcome

• World's highest performing and most energy efficient AI accelerators according to MLPerf benchmarks

Named customers1 record

Segments3 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability1 record

Feature3 records

Partnerships

Three partnerships are on record, tiered major and core.

• AMDmajorStrategic or Co-development Partner · 1 January 2025AMD acqui-hired Untether AI's engineering team in June 2025. The deal included the team and MK1 technology, with AMD acquiring the engineering talent to strengthen its AI inference capabilities.
• AmperecoreTechnology or Integration · 7 October 2024Partnership to provide Arm-based energy efficient AI inference solutions. Untether AI's inference accelerators combined with Ampere's Arm-based processors for optimized AI workloads.
• Ola-KrutrimcoreStrategic or Co-development Partner · 15 August 2024Untether AI selected by Ola-Krutrim for next-generation AI acceleration solutions. Krutrim is developing AI chips and cloud services for the Indian market.

Scale indicators3 records

Recent moves6 records

Expansion highlights4 records

Direct peers

• Cerebras Systems: Cerebras develops wafer-scale AI accelerators targeting both training and inference workloads for data center customers. It competes with Untether in the AI accelerator silicon market for hyperscaler-style deployments.
• Groq: Groq builds language processing units (LPUs) optimized for low-latency, high-throughput AI inference in data centers. It is the closest pure-play inference chip specialist to Untether AI in terms of focus on inference performance per watt.
• Tenstorrent: Tenstorrent develops AI processor chips and RISC-V based accelerator IP for training and inference workloads. It directly overlaps with Untether AI's product category of AI inference silicon for enterprise and data center.
• Hailo: Hailo designs edge AI inference accelerators optimized for low-power, real-time processing in embedded applications such as automotive, cameras, and industrial automation. Comparable to Untether's edge and physical AI inference focus.
• Mythic AI: Mythic developed analog compute-in-memory AI inference chips for edge devices before restructuring. Highly comparable to Untether AI's energy-efficient inference positioning for edge and embedded applications.
• SambaNova Systems: SambaNova builds dataflow AI accelerator chips and full-stack enterprise AI platforms for inference and training. It is a direct competitor in the AI inference hardware market that Untether AI targeted.
• Habana Labs (Intel): Habana Labs, an Intel subsidiary, builds Gaudi accelerators for deep learning training and inference in data centers. Intel Capital was a lead investor in Untether AI, signaling the same strategic ecosystem.

Broad incumbents

• NVIDIA: NVIDIA is the dominant GPU and AI accelerator supplier whose data center and edge inference products (TensorRT, Jetson, H100/B-series) define the competitive landscape Untether AI operated in.
• AMD: AMD is the incumbent CPU/GPU supplier that increasingly competes in AI accelerators (Instinct MI series) and ultimately acqui-hired Untether AI's engineering team to bolster its inference roadmap.

Emerging players

• Positron AI: Positron AI builds inference-focused FPGA-based accelerators for large language models, targeting the same inference-per-watt value proposition as Untether AI.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat2 records

Key risks5 records

Key highlights6 records

Customer concentration

• Xx.com/UntetherAI
• LinkedInlinkedin.com/company/untether-ai

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds5 records

Investors5 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 Untether AI