Skip to main content

MK1

Category: AI Infrastructure

MK1 develops high-speed AI inference and reasoning technologies optimized for AMD Instinct GPUs, enhancing performance and efficiency for large-scale AI deployments. MK1 was founded in 2023. The company is led by Paul Merolla. Based in Mountain View, United States. Team size: N/A. Total funding raised: Undisclosed. Latest round: Undisclosed. Key investors include Ascent Equity Partners, Interface Capital, Rsquared, Susa Ventures.

Founded
2023
Headquarters
Mountain View, United States
Team size
N/A
Total funding
Undisclosed

Value proposition

Optimizes AI inference workloads on AMD hardware, enabling trillion-token-scale processing with low latency and high efficiency

Products and solutions

Flywheel Inference Engine, Comprehension Reasoning Engine, Software stack for heterogeneous compute environments

Unique value

Founded by Neuralink veterans with expertise in neuromorphic engineering and low-power neural hardware design; proprietary optimization for AMD GPU architecture

Target customer

Enterprises leveraging AI for cloud computing, automation, and intelligent analytics

Industries served

Enterprise AI, Cloud Computing, Large-scale AI Reasoning Systems, Automotive (via AMD partnerships)

Technology advantage

Patent-optimized software stack for AMD Instinct GPUs, enabling 1+ trillion tokens/day processing; hybrid inference-reasoning architecture

How they differentiate

MK1 differentiates through neuromorphic engineering and low-power neural hardware design, leveraging expertise from Neuralink co-founder Paul Merolla. Focus on enterprise AI inference optimization for AMD's cloud infrastructure.

Main competitors

ESSENTIAL, Knostic, Unbound

Key partnerships

AMD (2025 acquisition), TensorWave (2024 cloud inference partnership), OpenAI (via AMD infrastructure deals)

Notable customers

TensorWave partnership, AMD enterprise clients

Major milestones

2023 Founding by Neuralink alumni, 2024 Partnership with TensorWave for AI inference solutions, 2025 Acquisition by AMD for enterprise AI inference capabilities

Growth metrics

Acquired by AMD in 2025; processes over 1 trillion tokens daily

Market positioning

Positioned as a specialized AI inference solutions provider for enterprise markets, targeting performance and efficiency improvements for cloud AI workloads

Geographic focus

Primary focus on North American markets, with potential expansion into global enterprise clients via AMD's distribution channels

Patents and IP

No public patents mentioned in available sources (as of Nov 2025)

About Paul Merolla

Co-founder of Neuralink (2016-2022), where he co-led chip design and developed algorithms for decoding brain activity. Former lead designer on over 10 neuromorphic chips with expertise in low-power neural hardware design.

Official website: