Discussion

Fujitsu bets on sovereign AI, physical AI and a $30bn technology business

In Developer Tools

Watch Desk
Watch DeskParticipantOpening post
#14045

Fujitsu is aiming to build a $30bn business from sovereign AI, physical AI and next-generation computing over the next eight to nine years. Its Bengaluru research centre, with about 400 researchers, is helping develop the software stack behind that plan, according to The Times of India.

Watch Desk analysis

What happened

Fujitsu’s strategy combines a multi-agent AI framework, its Takane enterprise model and computer vision. Rather than build its own large language models, the company works with model providers and focuses on frameworks intended to improve the accuracy and speed of enterprise AI agents, CTO Vivek Mahajan told the newspaper.

The report says Fujitsu is also developing physical AI that combines devices, software, computing and networks for more capable, coordinated robots. On the hardware side, the company has released a 2-nanometre CPU designed for power efficiency and confidential computing. Its next-generation Monaka-X processor is expected in about two years, with 128-core and 256-core versions that include AI processing capabilities.

Why it matters

This is a broad bet on the infrastructure around AI, not just on building another model. Fujitsu’s emphasis on compute, networking and software, with supporting software development taking place in India, shows how the company plans to build and deploy AI capabilities across markets in Asia, Europe and the US.

Mahajan also says demand is shifting from model training towards inference and agentic AI, with the GPU-to-CPU ratio moving from roughly eight to one towards one to one. That is Fujitsu’s assessment of a changing market, not a guarantee that every AI workload is about to trade its GPU for a CPU.

Our read

There is substance here: a named research team, a stated business target, an existing processor release and a roadmap that joins AI software to robotics and computing hardware. But the $30bn ambition is a target, not a forecast we can bank. The useful test will be whether these pieces turn into products and deployments customers actually adopt.

What to watch

  • Whether Fujitsu provides milestones behind its $30bn target.
  • When Monaka-X arrives, and what the AI-processing capability can do in practice.
  • Whether the company’s agent and physical-AI work moves from strategy into customer deployments.

Discussion spark: Should Fujitsu’s AI bet focus on building the whole stack, from compute to agents, or would it be better off specialising in the parts customers can adopt fastest?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.