Microsoft-Backed Chatbot Maker Inflection AI Raises $1.3bn

Mustafa Suleyman. Image credit: Deepmind

Nvidia, Bill Gates, Eric Schmidt in round that values Inflection at $4bn as year-old start-up seeks to outpace OpenAI

Inflection AI, a year-old start-up backed by Microsoft and several prominent tech industry billionaires, said it plans to expand rapidly after raising $1.3 billion (£1bn) from existing investors as well as AI accelerator chip maker Nvidia.

The investment, which takes the form of cash and cloud computing credit, values Inflection at some $4bn, according to a person familiar with the matter.

The round follows two months after Inflection released an initial version of its chatbot Pi – short for Personal Intelligence – which the company is hoping to develop into an all-purpose “digital Chief of Staff” that remembers past conversations with a user and appears to get to know them.

Founded by DeepMind co-founder Mustafa Suleyman and LinkedIn co-founder Reid Hoffman, the firm received funding from Hoffman as well as Bill Gates and former Google chief executive Eric Schmidt, all of whom are previous Inflection investors.

tara winstead ai artificial intelligence robot 02
Image credit: Tara Winstead/Pexels

AI frenzy

Microsoft, which is Inflection’s cloud computing provider, is also a main backer of Inflection competitor OpenAI and provides Azure cloud services to that company as well.

OpenAI kicked off the current investor frenzy around generative AI with the release of ChatGPT late last year.

Nvidia has been working closely with Inflection in deploying its H100 accelerator chip, currently the industry standard for training AIs – to the point that the US recently banned its sale to Chinese companies.

In June Inflection released a report on its Inflection-1 model, which powers Pi, saying it has outperformed most available models.

GPU cluster

Suleyman, who left DeepMind in 2019 following allegations of a bullying management style at the company, said most of the funding would be used to build a massive AI training cluster of 22,000 Nvidia H100 chips, which he said was about three times more power than was used to train OpenAI’s current foundation model GPT-4.

Inflection said it believes the system to be the world’s largest GPU cluster for AI applications, surpassing Meta’s 16,000 GPU cluster announced in May and trailing only Frontier, the supercomputer maintained by the Oak Ridge National Laboratory in Tennessee.

OpenAI hasn’t disclosed the scale of its current computing cluster, but Nvidia said last November it was planning to incorporate “tens of thousands” of GPUs into Microsoft’s Azure cloud service.

Suleyman said Inflection remains independent following the new investment by Microsoft and Nvidia and is free to choose with whom it makes partnerships.