GRAPHCORE

AI Focus Areas

• Specialized AI accelerators (IPUs)
• Graph‑based compute for ML
• AI cloud and bare‑metal services
• High‑performance training and inference
• AI system software (Poplar SDK)

Key Products

• Graphcore IPU hardware
• IPU‑Pod systems
• Poplar software stack
• Cloud IPU services via partners

Market Position

Graphcore was one of Europe’s most prominent AI hardware startups, raising around $700M and reaching valuations up to ~$2.8B before being acquired by SoftBank in 2024. Its IPU architecture was designed from the ground up for AI workloads, emphasizing massive parallelism and fine‑grained compute over large parameter graphs. While Graphcore faced intense competition from Nvidia and TPU‑based systems, it secured strategic partnerships with Dell, Microsoft Azure and others. As part of SoftBank, its technology may now be integrated into a broader AI hardware strategy, leveraging SoftBank’s capital and ecosystem relationships to find new niches where IPUs outperform generic GPUs.

AGI Relevance

Graphcore’s work explored alternative hardware architectures for large‑scale AI, including memory, interconnect and programming model choices optimized for neural networks rather than legacy workloads. Such exploration is crucial for identifying compute substrates that could support AGI‑scale models efficiently and safely. Insights from Graphcore’s successes and struggles—around software tooling, ecosystem building and customer education—highlight how difficult it is to shift the industry off incumbent GPUs even when technical advantages exist, informing how future AGI‑oriented hardware platforms might need to be designed and commercialized.

Investment Highlights

Graphcore raised multiple rounds including a $222M Series E at a $2.77B valuation in 2020, backed by Ontario Teachers’, Sequoia and others. Crunchbase records show a subsequent private equity round in 2021 and an acquisition by SoftBank announced in July 2024.

Tags

• IPU
• AI Chips
• UK Tech