Agentic AI hype is now pushing into CPU substrate demand, giving Japan’s Ibiden a useful server-side bump.
Demand for CPUs for agentic AI computing appears to be nudging the substrate supply chain. Analysts shifted attention towards CPUs after Intel’s earnings report earlier this year, when Chipzilla claimed agentic AI would revive interest in server processors.
Data from Japanese printed circuit board maker Ibiden suggests that demand is already being felt upstream. The company now expects stronger demand for general-purpose server products in 2026.
Ibiden serves the AI market through its electronics segment, which manufactures integrated-circuit package substrates. These connect chips to printed circuit boards and sit in the less glamorous plumbing of the AI supply chain.
The company’s substrates are used for AI GPUs, CPUs and application-specific integrated circuits. ASICs are custom AI chips developed by outfits such as Broadcom, and they often work alongside Nvidia’s high-performance gear.
Ibiden’s results for the fiscal year ending 31 March give a peek at the latest AI supply chain mood. They show demand forecasts across different AI chip types, with the company expecting CPU demand to grow.
For fiscal 2026, ending in March 2027, Ibiden now expects 330 billion yen (€1.78 billion) in electronics segment sales. That is up from its earlier forecast of 310 billion yen (€1.68 billion).
Operating profit expectations have jumped from 57 billion yen (€308 million) to 75 billion yen (€406 million). The extra cash is being driven by general-purpose server products and switching IC products.
The general-purpose products feed the server CPU market. That is where Chipzilla and its friends are hoping agentic AI will create fresh demand, rather than leaving all the glory to accelerator cards.
Ibiden said that due to “transitions from training to intelligence,” the “demand for CPUs for general-purpose servers is expected to increase.”
To handle that demand, Ibiden expects to expand production capacity. Its 2026 production load is expected to reach 1.8 times its 2024 level.
By 2028, the load is expected to hit 2.4 times the 2024 level. Growth is expected to come from ASICs, AI servers and server CPUs, while PC demand is expected to droop.
That points to a shift in the AI hardware food chain. GPUs still get the headlines, ASICs get the hyperscaler love and CPUs are being invited back to the party.
