AI Server Capacitor Prices Rise 40%; KKR and SK Form $1.3B Green Grid Platform

AI server supply chain price shocks are spreading upstream to capacitors and power chips while KKR and SK build a $1.3B clean-energy platform to power Korea's AI build-out.

📊 AI Server Demand Pushes Capacitor and Power Chip Prices Up 10-40% From July

Decoded: The AI-driven price surge in the semiconductor supply chain is expanding beyond GPUs and memory chips to upstream components, creating new bottlenecks that could slow the global AI infrastructure buildout, according to South China Morning Post on June 30. Japan's Murata Manufacturing — the world's largest maker of multilayer ceramic capacitors (MLCCs) — will raise prices for products used in AI servers and high-end automotive electronics by 10 to 40 per cent from July, citing a fully loaded order book, per Shanghai Securities News. AI servers require three to ten times as many capacitors as traditional servers. Power semiconductor prices are also rising: makers including Germany's Infineon, US-based Texas Instruments (TXN), and China's Yangjie Technology have all raised prices for power chips that regulate electricity inside AI data centers, according to Sinolink Securities analyst Liu Gaochang. The price increases have spread from selected products to broader categories, extending to aluminium foil, chemical materials, printed circuit board laminates, industrial gases, and ceramic components used in chipmaking tools. (South China Morning Post, June 30, 2026)

Why it matters: The MLCC and power chip price cycle signals that AI infrastructure cost pressure is now running through every layer of the bill of materials — not just the headline GPU and memory line items. Murata's 10-40% price increase on MLCCs lands in July, meaning AI server builders face margin compression on components they cannot substitute in the short term. For investors tracking AI infrastructure spending, this is a second-order inflation signal: hyperscalers and ODMs who locked in server pricing earlier in 2026 will absorb these increases on delivery, compressing per-rack economics. Texas Instruments (TXN) and Infineon stand to benefit from the pricing power shift — but it also narrows the margin buffer for data center operators who projected flat upstream costs while GPU and memory prices stabilized.

💡 KKR and SK Launch $1.3B Korea Renewable Energy Platform to Power AI Infrastructure

Decoded: South Korean conglomerate SK Inc. and US investment firm KKR & Co. announced on July 1 the formation of a Korea-focused renewable energy joint venture valued at approximately 2 trillion won ($1.3 billion). SK is consolidating renewable energy assets and businesses — including solar, wind, and fuel cells — from subsidiaries SK Innovation, SK ecoplant, and SK eternix into the new platform. KKR is making an investment of undisclosed size from its Asia Pacific infrastructure fund. The announcement comes as South Korea pursues a national semiconductor and AI infrastructure buildout requiring 6.3 gigawatts of electricity and 650,000 tons of water for its southwestern chip fab corridor, with an additional 8 gigawatts earmarked for new AI data centers. (Bloomberg, KKR press release, July 1, 2026)

Why it matters: KKR and SK's $1.3 billion renewable platform is a direct response to the power procurement math behind Korea's AI infrastructure ambitions. Samsung and SK Hynix's combined $648 billion and $29 billion capital programs respectively demand gigawatt-scale clean power access — the kind of behind-the-meter or grid-adjacent capacity that renewable platforms can provide faster than utilities can expand transmission. For investors, the KKR partnership structure — private infrastructure capital combined with SK's existing renewable asset base — is a template now visible in every major AI compute geography. It also signals that private equity is increasingly stepping in where public utility investment cycles are too slow to match AI buildout timelines, creating new infrastructure asset classes that sit between traditional power utilities and hyperscaler capex.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team