Musashi Seimitsu is a higher-torque play on the AI power and 800V theme than MLCC makers like Murata and Taiyo Yuden. Its supercapacitors smooth AI workload power spikes, a role that grows as architectures move to 800V; Skeleton's product suits peak shaving and is more complementary than competitive. The supercapacitor market could top $15B within years at likely 20%-30% operating margins versus 5% in the base business, which could lift operating profit over 10x. Target: a 2x in 12 months. x.com  … Show more

Musashi Seimitsu is the only maker mass-producing hybrid supercapacitors for AI data centers, and that supply gap could become a structural constraint like CoWoS in 2023. Nvidia built them into power delivery from GB300, about 300 per NVL72 rack, implying demand on the order of ten million units in 2026 versus Musashi capacity of 6.5 million by Q3 2026; Ming-Chi Kuo notes the count may be cut. Rubin should use more, Skeleton and Jianghai trail, and the 800V HVDC shift needs requalification. … Show more

Handling giant voltage and power swings is a big and arguably underappreciated opportunity over the next two to three years for battery storage and supercapacitor makers. Musashi Seimitsu and VinaTech are the preferred ways to play supercapacitor demand. x.com  x.com 

Musashi Seimitsu's hybrid supercapacitors look more necessary as Nvidia's Vera Rubin NVL72 racks arrive faster than expected. Dell delivered the first rack to CoreWeave on May 31, 2026, while the market had modeled Rubin as a late-2026 or early-2027 story. When the rack's 72 unified GPUs jump from idle to full inference load, the power surge hits the local grid like a shockwave, so data centers must deploy energy buffers to keep voltage sags from shutting down clusters. x.com 

Meta's unmet AI demand pays its bottleneck suppliers. Super Micro's $19.5B quarterly outlook and under 8x forward earnings point to a double. CoreWeave's Meta deals total $35.2B; it should easily top $100. Builders say nearly all new sites use Dell or Super Micro; Dell could easily double. Micron's demand is off the charts at 5x forward earnings. Sources say Vertiv's gear is sold out for 2-3 years. VivoPower has 2GW+ potential in 5-7 years. Digi Power X has a $1.1B Cerebras deal. x.com  x.com  … Show more

Networking, power and compute look more durable than cyclical memory in the AI buildout. Lumentum's revenue compounded year after year across the 2010s and 2020s, unlike Micron's; AWS revenue also compounded. Lumentum, Nebius, Bloom Energy and Credo should heavily outperform SanDisk and Micron from here, though AI itself remains heavily doubted. x.com  x.com 

Inference's move toward edge devices and dense, memory-hungry racks should drive demand for specialized chips, links, power delivery and cooling. McKinsey projects inference at 60% of AI computing demand by 2030. Ambarella makes on-device AI chips; Astera Labs links GPUs, memory and networks; Credo supplies short copper connections; Rambus earns from memory interfaces; Navitas reduces power waste; Modine cools dense racks; and Super Micro builds inference-specific server racks at volume. x.com  … Show more

Power delivery is the least crowded way to own AI capex, as capacitance moves onto the chip package. AI chips run at about 1V while drawing 800W to over 1kW, with current spiking from 100A to 900A instantly, too fast for distant MLCCs. AP Memory, the purest silicon capacitor play, posted record Q2 revenue up 78% with SiCap up 273% to 27% of sales, and one design needs nearly 100 caps per substrate with visibility into 2028. Samsung Electro-Mechanics signed a $1B SiCap supply contract. x.com 

Magnachip Semiconductor is a small pure-play power semiconductor maker that could re-rate violently if 800V data-center power becomes a full supercycle. Nvidia says the legacy 54V standard has become a bottleneck, and its 800 VDC partners are all large, discovered names. Magnachip has $50 million of net cash in a $240 million market cap and trades at 0.92x revenue. Management guides to $300 million revenue at 30% gross margin within three years, with design wins up 61%. It is higher risk. x.com 

Panasonic's rack-level backup architecture addresses the AI buildout's biggest near-term constraints, grid interconnection queues and power quality at 100kW-plus densities, an asymmetric play on power as AI's bottleneck. Hyperscalers can add protected kWh per rack instead of overbuilding central UPS, and peak shaving flattens spiky loads. Supercapacitor CBUs, targeted for FY27, handle fast transients for the 800VDC shift. Over 80% of FY29 is locked in design wins. x.com