Chips & Compute

Samsung Electro-Mechanics expands Nvidia deal to substrates and MLCCs

Samsung Electro-Mechanics expanded its Nvidia agreement to advanced substrates and MLCCs, with Nvidia weighing more funding for its Sejong and Busan plants ahead of mass production in 2029.

Samsung Electro-Mechanics expands substrate, MLCC supply to Nvidia
Samsung Electro-Mechanics expands substrate, MLCC supply to Nvidiaanalogicus / pixabay
By Sophie Lindqvist5 min read

Why it matters

  • Samsung Electro-Mechanics expanded its Nvidia supply agreement on Wednesday to include advanced package substrates and MLCCs.
  • An AI server requires more than 10 times as many MLCCs as a conventional server; Samsung Electro-Mechanics holds an estimated 40% of the AI-server MLCC market.
  • Nvidia is considering additional funding for the Sejong and Busan plants, following an October commitment, with full-scale mass production targeted from 2029.
  • Nvidia interim CEO Colette Kress also announced substrate agreements with Kinsus and an advanced-packaging deal with SPIL.

Samsung Electro-Mechanics has expanded its supply agreement with Nvidia to include more advanced semiconductor package substrates and multilayer ceramic capacitors (MLCCs) for Nvidia's AI computing platforms, the Korean company announced Wednesday. Nvidia is also considering additional financial support for Samsung Electro-Mechanics' production facilities in Sejong and Busan, on top of funding it already committed in October.

The deal deepens a strategic partnership the two companies first announced just last month. Under that earlier agreement, Samsung Electro-Mechanics agreed to supply advanced package substrates for Nvidia's computing and high-speed networking products. Wednesday's revision broadens the scope considerably: more substrates for next-generation graphics processing units, money for Korean factories, and — for the first time — MLCCs in the supply agreement.

Neither company disclosed financial terms. But the structure of the deal tells its own story: Nvidia is not just buying components, it is helping to pay for the factories that make them.

Why does Nvidia want MLCCs in the deal?

MLCCs are the unglamorous workhorses of modern electronics. In high-density, high-performance AI computing systems, they ensure stable power delivery — a job that gets harder as chips draw more power and pack more silicon into each package.

The numbers explain why this market matters. An AI server requires more than 10 times as many MLCCs as a conventional server, according to Samsung Electro-Mechanics. And the company already holds an estimated 40 percent share of the MLCC market for AI servers.

That position gives it leverage. Samsung Electro-Mechanics is pursuing large-scale, long-term supply agreements for high-performance MLCCs designed to withstand high temperatures and voltages. Some customers have even begun making financial commitments to help the company expand production capacity — the same pattern Nvidia is now following.

  • An AI server needs more than 10x the MLCCs of a conventional server, per the company.
  • Samsung Electro-Mechanics holds an estimated 40% of the AI-server MLCC market.
  • It is chasing long-term supply deals for MLCCs rated for high temperatures and voltages.
  • Some customers are already funding its capacity expansion.

In short: the component shortage risk in AI hardware has shifted down the stack, and capacitor makers now have bargaining power once reserved for chip designers.

What is Nvidia's money actually buying?

The additional financial support under consideration would go toward expanding production infrastructure and introducing advanced manufacturing equipment at Samsung Electro-Mechanics' Sejong and Busan plants. That follows Nvidia's October commitment to fund dedicated manufacturing facilities at the same two sites.

The goal is scale with a deadline. Samsung Electro-Mechanics said the funding will let it establish a full-scale mass-production system starting in 2029.

The company says it will strengthen operations at both facilities along three tracks:

  • Improving production line efficiency.
  • Maintaining high quality standards.
  • Running flexible production systems that can respond to changes in demand.

That flexibility clause matters. AI hardware demand has swung hard in both directions over the past two years, and suppliers that cannot flex capacity up or down tend to get stuck with either shortages or write-downs.

Who said what?

Samsung Electro-Mechanics framed the expanded deal as a step toward becoming a comprehensive supplier of AI semiconductor components, covering both advanced package substrates and MLCCs.

"We will fully carry out our role as a total solutions provider by combining our world-class fine-pattern substrate technology with highly reliable MLCC technology to ensure Nvidia's AI platforms reach the global market on time with the highest performance and quality," Samsung Electro-Mechanics CEO Chang Duck-hyun said.

Nvidia confirmed the deal on its own timetable. Interim CEO Colette Kress, speaking during the company's earnings call on Tuesday local time, said: "We are announcing new agreements with Kinsus and Samsung Electro-Mechanics in substrates, and SPIL in advanced packaging, further strengthening our global supply chain capacity and resiliency."

Takeaway: Nvidia is lining up substrate capacity across multiple suppliers at once.

Why is Nvidia diversifying its substrate supply?

Kress's wording — "capacity and resiliency" — is the tell. Nvidia is simultaneously announcing substrate agreements with Kinsus, a Taiwanese substrate maker, and an advanced-packaging agreement with SPIL, also of Taiwan. Samsung Electro-Mechanics joins that roster as the Korean pillar.

Advanced package substrates are the physical foundation that connects a GPU die to the rest of the board. As Nvidia's GPUs have grown larger and faster, substrates have become a genuine bottleneck — the fine-pattern processing needed to route signals at high speeds is hard to manufacture at volume. Substrates are no longer a commodity purchase; they are strategic capacity.

By funding suppliers' factory expansions directly, Nvidia secures allocation without having to acquire the suppliers outright. Wednesday's announcement extends that playbook from substrates into power components.

  • Nvidia announced substrate deals with both Kinsus and Samsung Electro-Mechanics.
  • It also announced an advanced-packaging agreement with SPIL.
  • The deals span Korea and Taiwan, two of the three major substrate manufacturing bases.

That is a supply-chain hedge with real money behind it.

What happens next?

The near-term milestones are concrete. Samsung Electro-Mechanics will scale operations at Sejong and Busan, backed by Nvidia's October funding commitment plus whatever additional support gets finalized. The full-scale mass-production system comes online starting in 2029, according to the company.

The longer-term significance is broader. If AI server makers keep requiring 10 times the MLCCs of conventional servers, and if substrate complexity keeps climbing with each GPU generation, then component makers with both capabilities gain pricing power and strategic weight. Samsung Electro-Mechanics is betting that being a "total solutions provider" — one vendor for both the substrate and the capacitors sitting on it — makes it harder to displace.

For Nvidia, the bet runs the other way: paying upfront for capacity in Korea and Taiwan keeps its AI platform roadmap on schedule through the decade. <

Source: feed.koreatimes.co.kr

Share this article:

More from Sophie Lindqvist

Sophie Lindqvist

Show full bio

Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.

749 articles

Related articles

  1. Analysts: Nvidia's Wobbling Chip Dominance Hands Samsung an Opening
  2. Nvidia Case Puts AI Chip Supply Chains Under New Scrutiny
  3. Korea says Nvidia's 260,000-GPU pledge survives Huang's exit
  4. NVIDIA Locks In 3.5x More Substrate Capacity From Kinsus By 2028
  5. Microsoft Shifts AI Strategy Toward Homegrown Chips to Cut Nvidia Reliance

« Previous articleNext article »