Myuex

Long term energy storage solutions

The thesis that AI-driven power demand will drive adoption of long-duration energy storage is strongly supported by the regime and live forces: AI data center electricity demand is projected to double by 2030, grid connection bottlenecks are severe, and the BloombergNEF forecast that data centers will use 20% of US electricity by 2035 directly validates the core premise. However, two supporting pillars were recently invalidated: the 'Green Energy Shift' pillar weakened as oil prices stayed below $150, reducing the economic urgency for rapid storage adoption, and the 'AI Expansion Boosts Energy Use' pillar was undermined by the Fed signaling a potential rate hike, which could tighten capital for AI infrastructure. The remaining evidence—grid operator warnings, FERC orders, and battery giant pivots to data center storage—still supports the thesis, but the invalidations cap conviction. The basket fit is strong for GWH, FLNC, EOSE, IESVF, and NRGV, which directly produce long-duration storage, but BEP is a renewable generator with only 0.30 exposure to storage, diluting the basket's purity. The regime's reaccelerating inflation and hawkish Fed tilt are hostile to rate-sensitive growth stocks like these, which have already underperformed the S&P by 11.6% since posting. The scenario arithmetic leans positive but the invalidations and macro headwinds justify a score below the anchor, as the thesis still holds but with reduced probability of outperformance over the horizon.

GaN is the future of semiconductors

The thesis rests on a structurally sound physical advantage for GaN in power efficiency, and the regime strongly supports it: AI data center power demand is projected to double by 2030, creating an urgent need for efficiency gains, while China's gallium export controls tighten supply for non-Chinese producers. The most supportive evidence comes from the AI power demand surge (multiple events showing data center electricity use reaching 20% of US supply by 2035) and China's critical mineral export controls, which directly benefit the basket's gallium supply chain (AA, FPLSF). However, the thesis is undercut by two powerful counter-forces: the Fed's hawkish hold with reaccelerating inflation raises borrowing costs and could delay data center buildout, and the resumption of Chinese gallium exports to Japan in May 2026 eases the supply scarcity that was central to the author's argument. The basket's realized -20.5% return since posting reflects these headwinds, and the author's track record provides no corroboration. The scenario arithmetic leans slightly positive (anchor) but the live news—especially the gallium export resumption and Fed rate-hike risk—push conviction below that anchor. The strongest upside path is through NVTS design wins in AI data centers, but the slow enterprise procurement cycle (pressure #4) and regulatory uncertainty (pressure #3) create timeline risk that the thesis's 2027 volume production target may slip. The score sits below the anchor because the gallium supply invalidation is a live, confirmed event, not a hypothetical, and the macro regime of high rates and inflation directly contradicts the thesis's timeline for data center buildout acceleration. Key drivers are the AI power demand surge (supportive), China's gallium export resumption (contradictory), and the Fed's hawkish hold (contradictory). The basket's gallium supply chain (AA extracting from bauxite, FPLSF refining) is a credible passthrough for the supply scarcity thesis, but the resumption of Chinese exports weakens that link. NVTS remains the core holding, and its design wins are the most direct transmission path, but the competitive threat from ON Semiconductor's GaN portfolio launch and Wolfspeed's patent lawsuit add execution risk. The regime of reaccelerating inflation and a hawkish Fed is hostile to a thesis that depends on large-scale capital investment in data center efficiency upgrades, and this contradiction is a primary reason for the below-anchor score.