AIToday
RoboticsTop Companies' AI MovesAI Business & IndustryTop Companies AIPublished: Nov 17, 2026, 06:31 JST

KDDI Matsuda: Sakai AI center enters inference phase, physical AI in next mid-term plan

Send on LINE
KDDI Matsuda: Sakai AI center enters inference phase, physical AI in next mid-term plan

3 Key Points

  1. What happened

    At KDDI SUMMIT 2026, Matsuda said the Sakai data center is adding equipment to shift from AI training to inference, and that a new company, ELYZA Robotics, was established.

  2. Why it matters

    Matsuda said the shift may force a rethink of what network quality means — AI processing speed over always-fast speeds — and that ELYZA Robotics, with ELYZA's domestic AI models, uses Sakai GPUs and KDDI's network as a test field.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Matsuda's remarks place KDDI's AI strategy on two tracks that meet in the network. The Sakai data center, now entering what he calls the inference phase, is paired with a planned proof-of-concept with APN, the next-generation all-optical network technology. That pairing matters because Matsuda is questioning the metric the industry has used to sell connectivity. Where once the value was a call that did not drop, and later raw speed, he told staff the "optimization function" may change so that AI processing speed counts for more than a constant 500Mbps.

The physical AI effort is described as deliberately patient. Matsuda said he kept the term out of the current mid-term plan because he wants execution and concrete social implementation rather than following a buzzword, and he framed the next three years as a preparation period. The ELYZA Robotics establishment fits that framing: it is a way to test where a robot's own edge processing, with a brain of hundreds of billions of parameters, is necessary and where heavy processing should go to Sakai.

Tying these together is Matsuda's view of KDDI's role as a combiner rather than a self-contained builder. He cited CDMA and SpaceX's Starlink as cases where KDDI connected technologies that had not reached Japan's last mile, and he applied the same logic to physical AI, where he sees a Japanese advantage in craftsmanship and unspoken on-site know-how. That know-how, he suggested, could be packaged like a swappable cassette on top of outside AI models.

FAQ
What is KDDI doing with its Sakai data center?
Matsuda said new equipment is being installed so the facility enters the inference phase, running AI in actual services rather than just training.
Why is physical AI not in KDDI's current mid-term plan?
Matsuda said full adoption is still some way off and he wants concrete social implementation before promoting it; he expects it to feature more prominently in the next mid-term plan.
What does KDDI aim to do with ELYZA Robotics?
Matsuda said the new company can use ELYZA's domestic AI models, Sakai's GPUs and KDDI's network to determine how much processing robots handle at the edge versus in the data center.
Top Companies AIRead Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleBerkshire's $377.4 billion cash could target ASML Holding next