Key Takeaways
- • Sequoia Capital backed Graphcore, a Bristol-based AI chip startup, with a $50 million Series C investment in November 2017, positioning the company to develop specialized Intelligence Processing Units (IPUs) designed specifically for machine learning workloads rather than general-purpose computing.
- • Graphcore's IPU technology claimed to deliver between 10x to 100x acceleration compared to existing hardware at the time, leveraging a graph architecture with thousands of cores and memory distributed across the chip to address the computational demands of AI applications.
- • Sequoia partner Matt Miller stated the firm believed Graphcore had "the potential to be the most transformative chip company we have partnered with since we invested in NVIDIA over 20 years ago," signaling the venture investor's conviction in the competitive opportunity within specialized AI chip design.
- • Graphcore was subsequently acquired by SoftBank in 2024, though Sequoia later wrote off its stake in the company, reflecting challenges in the competitive AI chip market despite the initial strategic thesis.
- • The investment thesis centered on the observation that GPUs, while superior to CPUs for machine learning, were not purpose-built for AI workloads—creating an opening for specialized chip architectures to capture meaningful performance gains and market share.
Summary
Sequoia Capital invested $50 million in Graphcore during the company's Series C funding round in November 2017, backing the Bristol-based startup's mission to develop specialized microprocessors for artificial intelligence applications. The investment reflected Sequoia's thesis that the rapidly expanding machine intelligence market would support dedicated chip architectures designed specifically for AI workloads, rather than relying on general-purpose processors or graphics processing units adapted for machine learning.
Graphcore's core technology was the Intelligence Processing Unit (IPU), a microprocessor utilizing a graph architecture with thousands of cores and memory distributed across the chip. According to the company's claims at the time, the IPU technology would deliver between 10x to 100x acceleration compared to then-existing hardware solutions. The company projected shipping its first IPU products to early access customers at the start of 2018, with production scaling and developer community building as immediate priorities following the capital infusion.
Sequoia's conviction in Graphcore's potential was explicit and expansive. Matt Miller, a Sequoia partner who joined Graphcore's board, stated that the firm believed Graphcore represented "the most transformative chip company we have partnered with since we invested in NVIDIA over 20 years ago." This positioning reflected the venture capital firm's assessment that machine intelligence would drive explosive growth in new applications and services across multiple industries, from autonomous vehicles to medical imaging and air traffic control systems. The firm's due diligence involved meetings with teams across the United States, Europe, Israel, and China, as well as extensive consultation with the machine learning community, enterprise technology landscape, and major cloud providers.
However, the long-term outcome diverged significantly from Sequoia's initial thesis. Graphcore was acquired by SoftBank in 2024, and Sequoia subsequently wrote off its stake in the company. This outcome underscores the challenges facing specialized AI chip startups in competing against both established semiconductor manufacturers and well-capitalized new entrants in the rapidly evolving artificial intelligence hardware market.
Graphcore's core technology was the Intelligence Processing Unit (IPU), a microprocessor utilizing a graph architecture with thousands of cores and memory distributed across the chip. According to the company's claims at the time, the IPU technology would deliver between 10x to 100x acceleration compared to then-existing hardware solutions. The company projected shipping its first IPU products to early access customers at the start of 2018, with production scaling and developer community building as immediate priorities following the capital infusion.
Sequoia's conviction in Graphcore's potential was explicit and expansive. Matt Miller, a Sequoia partner who joined Graphcore's board, stated that the firm believed Graphcore represented "the most transformative chip company we have partnered with since we invested in NVIDIA over 20 years ago." This positioning reflected the venture capital firm's assessment that machine intelligence would drive explosive growth in new applications and services across multiple industries, from autonomous vehicles to medical imaging and air traffic control systems. The firm's due diligence involved meetings with teams across the United States, Europe, Israel, and China, as well as extensive consultation with the machine learning community, enterprise technology landscape, and major cloud providers.
However, the long-term outcome diverged significantly from Sequoia's initial thesis. Graphcore was acquired by SoftBank in 2024, and Sequoia subsequently wrote off its stake in the company. This outcome underscores the challenges facing specialized AI chip startups in competing against both established semiconductor manufacturers and well-capitalized new entrants in the rapidly evolving artificial intelligence hardware market.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.