Groq: Improving AI and Machine Learning
Providing unique and deterministic performance for compute-intensive workloads
Companies are increasingly turning to compute-intensive applications such as AI and machine learning to improve customer experience, competitive advantage, security, and more. However, due to the growing intricacy of hardware processor models, achieving the high-performance these workloads require is difficult.
Alumni Ventures’ portfolio company Groq invented the Tensor Streaming Processor (TSP) architecture-based chip to address this pain point. This simplified processing architecture provides a unique and deterministic (state or condition that always produces the same results) performance for compute-intensive workloads to help facilitate AI and deep learning technology.
Outperforming Hardware
Groq’s TSP chip surpasses industry standards by increasing computing response speed and performance, touting 2x to 18x faster and 3x to 20x performance superiority. Moreover, Groq’s chip design reduces the complexity of the traditional hardware-focused development, so developers can concentrate on algorithms instead of adapting their solutions to the hardware.
Some of the benefits of Groq’s architecture include:
- Saving developer resources by eliminating the need for profiling
- Better deploying of AI solutions
- Higher responsiveness with lower TCO
- More efficiently processing small batch workloads
You can read more about Groq’s technology in this white paper.
What We Liked About the Deal
Significant Market Potential: The global AI chip market size is projected to reach $91+ billion by 2025. With more companies adopting AI and machine learning practices, the need for Groq’s technical problem-solving TSP chip is increasing.
Competitive Moat and Customer Traction: Groq already has partnerships with autonomous vehicle companies, financial services organizations, and national laboratories. The company currently performs better than Habana, Nvidia, Graphcore, and Cabricon chips, with 16x better latency and IPS (instruction per second), 5x better IPS/Watt (lower power consumption), and 10x better TCO.
Experienced Team: Groq was formed in 2017 by a strong team of ex-Google engineers with experience in technical and corporate development, as well as a successful track record in software development, chip architecture, and scaling chips at the commercial level. Additionally, company board members include prior Chief Supply Chain Officer for HP Stuart Pann and acclaimed Venture Capitalist Andy Rappaport (who came out of retirement to join the company).
How We Are Involved
Alumni Ventures’ Castor Ventures (for MIT alumni and friends of the community) and sibling fund Ring Ventures (for Texas A&M former students and friends of the community) deployed capital in Groq’s $300 million funding round led by Tiger Global Management and billionaire investor Dan Sundheim’s D1 Capital.