Content
summary Summary

British AI chipmaker Graphcore is considering a sale to foreign owners.

Ad

According to The Telegraph, Graphcore is in talks with major technology companies about a possible deal to raise fresh capital to cover heavy losses. A sale of the company is also being considered.

British chipmaker Arm, Japanese technology conglomerate SoftBank and AI company OpenAI have all been mentioned as potential buyers. It is unclear how far the sale talks have progressed. Talks about further venture capital may be running parallel to the sale talks, according to the Telegraph.

London-based investment fund Chrysalis, which invested in Graphcore, announced in December that one of its portfolio companies was in sale negotiations. Chrysalis and another investor, Baillie Gifford, are said to have raised the valuation of their Graphcore stake after writing it down significantly. Investor Sequoia reportedly wrote down the value of its stake in the start-up to zero.

Ad
Ad

A sale to foreign bidders would likely be scrutinized by national security agencies, as AI technology is considered strategically important, according to the Telegraph.

Graphcore has raised more than $700 million in the past from investors including Microsoft and Silicon Valley venture capital giant Sequoia, and was valued at $2.8 billion at the end of 2020.

But the company has struggled to sell its intelligence processing units, which compete with GPUs sold by Nvidia. Graphcore says it's "doing something different" than Nvidia because it builds a graph processor, not a GPU, which has a different set of strengths and weaknesses. Ultimately, though, they are competing for the same budgets.

In 2022, Graphcore's loss increased 11 percent to $204.6 million, while revenue fell from $5 million to $2.7 million. The company said it would have $157 million in cash by the end of the year and would need to raise more funds by May, the Telegraph reports.

Recently, Graphcore reportedly pulled out of China due to US export regulations.

Recommendation

AI chips are rare, but Graphcore can't compete with Nvidia yet

Graphcore is a British manufacturer of AI hardware, founded in 2016. The company makes chips specifically for artificial intelligence development, and in the past has marketed them as a competitor to Nvidia's DGX systems.

One of Graphcore's main products is the Intelligence Processing Unit (IPU) technology, which Graphcore aims to establish as the global standard for AI computing.

Its latest innovation is the Bow IPU, the world's first wafer-on-wafer processor using TSMC's new 3D wafer-on-wafer technology. This processor promises up to 40 percent higher performance with 16 percent better power efficiency.

According to Graphcore, the Bow-Pod16 is clearly ahead of Nvidia's DGX-A100 server in the ResNet-50 benchmark and offers competitive prices.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.

However, Graphcore's chips have yet to gain a foothold in the market. One reason for this may be that, in addition to the hardware itself, the software that controls the hardware plays an important role. Nvidia executive Manuvir Das described his company as 80 percent software, optimizing its hardware for AI training and inference.

Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • British AI chipmaker Graphcore is considering a sale to foreign owners, possibly including Arm, SoftBank or OpenAI.
  • Graphcore has raised more than $700 million from investors in the past, and was valued at $2.8 billion at the end of 2020. However, the company has struggled to sell its intelligence processing units, which compete with Nvidia's GPUs.
  • A sale to foreign bidders would likely be scrutinized by national security agencies, as AI technology is considered strategically important.
Sources
Online journalist Matthias is the co-founder and publisher of THE DECODER. He believes that artificial intelligence will fundamentally change the relationship between humans and computers.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.