Fujitsu (FJTSY) opens MONAKA server sales in November as global CPU rollout follows

A row of server racks in a data center aisle, lit by rows of blue, orange and red status lights, with two people walking away from the camera in the background.

AI-generated illustration

Key points

  • Fujitsu begins MONAKA server sales in November; global CPU rollout follows
  • 2nm compute die built on Fugaku technology
  • Fujitsu rose 7.4% in Tokyo on Monday
  • 24 companies already testing prototype servers

Fujitsu (FJTSY) will begin selling MONAKA servers in Japan and Europe in November 2026, while global sales of its standalone MONAKA processor are planned for the fourth quarter of its fiscal year ending March 2027. The company announced the rollout in a Monday release from Kawasaki.

Fujitsu shares closed at 4,026 yen in Tokyo on Monday, up 7.4% on the day, and its US-listed shares closed at $26.16, up 7.0%.

The server is the Fujitsu MONAKA Server, and it goes to data center operators, enterprises and the academic and high-performance computing sectors. Fujitsu says it will also sell it to the defense sector. It's manufactured at Fujitsu's Kasashima Plant. The processor, designed and developed in Japan, will be sold on its own to cloud and data center operators and to server vendors.

What is FUJITSU-MONAKA?

It's a central processor rather than a graphics chip, aimed at AI inference, the stage where a trained model answers a query instead of learning from data. The design comes out of the work behind Fugaku, the supercomputer Fujitsu built for Japan's government, and the technology behind it is based on results from a project subsidized by Japan's New Energy and Industrial Technology Development Organization.

Fujitsu puts the maximum operating frequency at 3.8GHz and memory transfer speed at 8800MT/s. The release says the chip delivers "twice the throughput for AI inference compared to other CPUs" and names no competitor and no benchmark. It also carries confidential computing, which encrypts data and applications in memory in hardware, using Arm (ARM) CCA.

Fujitsu presented the layout at Hot Chips in August. The chip is a stack of chiplets joined by die-to-die hybrid bonding. The core die is made on Taiwan Semiconductor's (TSM) N2P 2-nanometer process and the SRAM and I/O dies on its 5-nanometer N5, according to ServeTheHome, which put the count at 144 cores per chip and 12 channels of DDR5 memory. Two versions were shown. The high-performance part runs at 500 watts with a 2.9GHz base clock and the high-efficiency part at 350 watts and 2.1GHz. Arm says the design scales to 288 cores across dual sockets on Armv9, with support for PCIe 6.0 and CXL 3.0.

Fujitsu showed a successor in the same Hot Chips session. MONAKA-X is planned on a 1.4-nanometer process and will support the NVLink Fusion interconnect from Nvidia (NVDA). It's the processor slated for FugakuNEXT, the machine that replaces Fugaku.

What Fujitsu expects to sell

Prototype AI servers running MONAKA are in testing at 24 companies in finance, telecommunications and other sectors, and Fujitsu is in talks with about 40 more, Nikkei Asia reported on September 13. The company is targeting 250 billion yen ($1.63 billion) of cumulative sales of the chip to outside customers through fiscal 2030, alongside sales of 60,000 AI servers.

Nikkei reported that Fujitsu will begin exporting the chips to the United States and Asia "as early as next year." Monday's release puts a quarter on it. The global availability Fujitsu describes for that fourth quarter includes the US and Asia-Pacific.

November is the date sales open, not the date hardware arrives. Fujitsu will "begin selling" the 2U and 1U rack-mount servers in Japan and Europe starting November 2026, and shipments "will then commence sequentially from April 2027," according to the release. Both dates attach to the MONAKA Server. Before that, Fujitsu says it will offer the server to select customers in the financial, telecommunications and manufacturing sectors in the second half of fiscal 2026 to expand use cases. The release gives no shipping date for the standalone processor.

The 1U server uses composable disaggregated infrastructure and CXL, which lets memory and accelerators be allocated beyond the physical box. Fujitsu says it runs on air cooling in ambient temperatures up to 40 degrees Celsius and on water at up to 45 degrees, and that this cuts server cooling power by up to 80%.

Where Supermicro fits

Super Micro Computer (SMCI) agreed in October 2024 to develop and market a platform built around MONAKA, along with liquid-cooled systems for high-performance computing and generative AI. Charles Liang, president and chief executive of Supermicro, said in the announcement that the systems would be "optimized to support a broad range of workloads in AI, HPC, cloud and edge environments." Vivek Mahajan, corporate vice president, CTO and CPO at Fujitsu, called the arrangement "a groundbreaking initiative that will accelerate green computing innovation" in the same release.

Monday's release doesn't name Supermicro. The servers announced Monday are designed, developed and manufactured by Fujitsu in Japan. Supermicro closed at $36.74 on Monday, up 25.5% since the start of the year, after revenue landed at the low end of its own guidance in July.

Fujitsu has named no customer for MONAKA. Sales of the 2U and 1U rack-mount servers open in Japan and Europe in November.

Yen is converted at 153.42 yen to the dollar, the September 13, 2026 rate.

Frequently asked questions

When does Fujitsu start selling FUJITSU-MONAKA?

Sales of the Fujitsu MONAKA Server open in Japan and Europe in November 2026, and shipments commence sequentially from April 2027, according to Fujitsu's September 14, 2026 release. The standalone processor is planned for cloud and data center operators and server vendors in the fourth quarter of fiscal 2026, the quarter ending March 2027, with global availability including the US and Asia-Pacific. Fujitsu gives no shipping date for the standalone processor.

What is FUJITSU-MONAKA?

An Arm-based server CPU aimed at AI inference, built on technology from Fugaku, the supercomputer Fujitsu built for Japan's government. Fujitsu puts maximum operating frequency at 3.8GHz and memory transfer speed at 8800MT/s, and says it delivers twice the AI inference throughput of other CPUs, without naming a competitor or a benchmark. The technology is based on results from a project subsidized by Japan's New Energy and Industrial Technology Development Organization.

What process node is the MONAKA chip made on?

Two. The compute die is made on TSMC's N2P 2-nanometer process and the SRAM and I/O dies on its 5-nanometer N5, joined by die-to-die hybrid bonding, according to ServeTheHome's report from Hot Chips in August 2026. ServeTheHome put the count at 144 cores per chip and 12 channels of DDR5. Arm says the design scales to 288 cores across dual sockets on Armv9, with PCIe 6.0 and CXL 3.0.

How many companies are testing MONAKA servers?

Prototype AI servers are in testing at 24 companies in finance, telecommunications and other sectors, with talks under way with about 40 more, Nikkei Asia reported on September 13, 2026. Fujitsu is targeting 250 billion yen ($1.63 billion) of cumulative chip sales to outside customers through fiscal 2030, alongside 60,000 AI servers.

Is Supermicro building the MONAKA servers?

Super Micro Computer agreed in October 2024 to develop and market a platform built around MONAKA. Fujitsu's September 14, 2026 release does not name Supermicro, and describes the servers announced that day as designed, developed and manufactured by Fujitsu in Japan, at its Kasashima Plant. This is general information, not investment advice.

More on FJTSY and ARM

Dennis Singleton
Dennis Singleton

Dennis Singleton has spent years following the markets, but what keeps his attention is how AI is built. He writes about the companies behind the technology, from semiconductor designers and advanced packaging to photonics, memory, networking, and the hardware powering modern AI. His approach starts with filings, earnings, and industry research, then translates the important details into clear, straightforward analysis without unnecessary hype.