Archive_Clive

Ram supply shock and demand increase skyrocket prices

The thesis rests on three pillars: hyperscaler pre-buy contracts locking up supply, AI-driven demand growth, and multi-year fab construction timelines enforcing supply discipline. The first and third pillars are well-supported by live news — Micron's $100B in long-term supply agreements with cash deposits, memory chip prices surging 98% in Q1 2026, and Lenovo's warning that RAM prices will not return to 2025 levels through 2030 all confirm the supply-capture and supply-discipline mechanisms. However, the AI-driven demand pillar has been materially weakened by two recent invalidations: the AI Bubble Risk article flags that massive debt-fueled spending and free Chinese AI models are eroding the demand narrative, and the report of US startups migrating to cheaper Chinese AI models shows cost-driven substitution that could reduce demand for high-end memory. These invalidations do not break the thesis — the supply-side constraints are independently strong — but they cap the upside conviction because the demand growth that was supposed to compound the supply squeeze is now less certain. The regime is hostile to the thesis in one important way: the Fed is signaling potential rate hikes to combat reaccelerating inflation, and the AI Bond Demand Weakens article shows that borrowing costs for data center financing are rising, which could slow hyperscaler capex. Against this, the power-constraint narrative (BloombergNEF doubling US data center power forecast to 194 GW by 2035) and the grid strain evidence support continued AI infrastructure buildout, which sustains DRAM demand even if growth moderates. For the holdings: MU is the purest play on DRAM pricing and has already demonstrated pricing power through its $100B contracts; SSNLF and HXSCL are the dominant producers that benefit from any price increase, but Samsung's consumer electronics exposure adds cyclical risk. The basket has already returned +9.8% against the S&P's +1.9%, which partly reflects the thesis being partially priced in, but the supply constraints are structural and multi-year, so the remaining upside depends on whether demand holds. The strongest upside driver is the confirmed supply-capture mechanism (hyperscaler contracts locking up 70% of output), while the most probable risk is that AI capex growth slows as financing costs rise and ROI scrutiny increases, which would reduce the demand side of the equation.