Week One

This is the first portfolio review. The three research pieces — transformers, copper, cooling — were published between April 9 and April 11. A week later, 9 of 10 positions are green. ETN leads at +7.1%. MOD is the only red at -0.4%, narrowing from -1.5% yesterday. The portfolio is working.

But the interesting thing about the first week isn't the price moves. It's the developments that confirm, complicate, or challenge the original theses. Price tells you what the market thinks today. Developments tell you whether the market will think differently tomorrow.

Transformers: thesis strengthening

The transformer bottleneck piece argued that AI's binding constraint is electrical infrastructure, not chips. One week later, the evidence has gotten harder to ignore. Nearly half of all planned US data center capacity has been canceled or delayed — not for lack of GPUs, but for lack of transformers and grid connections. Lead times have stretched further: 3 to 5 years for large power transformers, up from the 24-30 months I cited in the original piece.

The companies are responding at scale. Eaton announced a $340 million investment in a new South Carolina transformer manufacturing facility, with production starting in 2027. That brings their cumulative North American manufacturing investment past $1 billion since 2023. Quanta Services held an investor day on April 4 projecting a $2.4 trillion total addressable market and targeting $21.60-$26.75 EPS by 2030. They're doubling their 765 kV transformer facility in Pennsylvania.

GE Vernova reports Q1 earnings on April 22. Quanta reports April 30. These will be the first real test of whether backlog is converting to revenue at the rates guided. Watch the book-to-bill ratios.

One development I didn't anticipate: China controls roughly 60% of global transformer manufacturing capacity, and US imports surged to over 8,000 units in 2025 — up from fewer than 1,500 in 2022. Bloomberg reported that the US data center expansion relies heavily on Chinese electrical equipment. This adds a geopolitical dimension the original piece didn't address. If tariffs or trade restrictions hit transformer imports, the bottleneck tightens further. Good for domestic manufacturers (GEV, ETN), potentially disruptive for the buildout timeline.

No signal changes for the transformer positions. All three — BUY ETN, BUY GEV, HOLD PWR — hold.

Copper: thesis holding, execution risk rising

The copper thesis is structurally intact. Prices are at $6.07/lb, up 5% month-over-month and 30% year-over-year. Analysts project a 330,000-tonne deficit for 2026. The Grasberg mine in Indonesia — the world's second-largest — remains partially closed after a September mudslide, with the Block Cave section (70% of output) not expected to reopen until Q2. Every supply disruption tightens the thesis.

But two material developments complicate the picture.

First: Southern Copper's CEO Oscar Gonzalez Rocha died on April 7 after leading the company since 2004. Twenty-two years of institutional knowledge and industry relationships. No successor has been announced. The Tía María expansion project — $1.8 billion, 120,000 tonnes of annual capacity — is officially "under review, not canceled" per Peru's Mining Ministry. This is one of SCCO's key growth catalysts. A prolonged leadership transition could delay it further.

I'm moving SCCO from BUY to HOLD. The fundamental copper thesis hasn't changed, but the company-specific execution risk has. A 22-year CEO can't be replaced without friction, and Tía María was already politically contentious before this. Hold if you own it. Wait for the successor announcement and clarity on the expansion timeline before adding.

Second: on April 6, the Trump administration imposed new Section 232 tariffs on copper — 50% on pure copper and copper products, 25% on derivative articles. Domestic-content products get a lower 10% rate. This is structurally positive for US copper miners (FCX benefits from price support) but creates cost pressure on downstream copper users, including the very transformer manufacturers that the first thesis is built on. The tariff adds another feedback loop: copper tariffs raise transformer costs, which could slow data center construction, which is the demand driver for copper. The system is more interconnected than any single thesis captures.

FCX reports Q1 earnings on April 23. Analysts expect 100% year-over-year EPS growth. Watch for Grasberg recovery timeline commentary and any guidance on tariff effects.

Signal changes: SCCO moves from BUY to HOLD. FCX, COPX, BHP unchanged.

Cooling: thesis confirmed

The cooling thesis was the most contrarian of the three — arguing that the January CES sell-off (Modine -21% in a day) was a misreading of what Jensen Huang actually said. One week later, the thesis is confirmed on multiple fronts.

Nvidia's Vera Rubin architecture, when it ships in H2 2026, will require 100% liquid cooling at 600 watts per rack. That's not "water chillers are unnecessary" — it's "direct-to-chip liquid cooling is mandatory." Every existing data center that wants to run Rubin needs to retrofit. Cooling infrastructure lead times are 12-18 months. The window for ordering is now, not later.

The M&A activity validates the thesis from the buy side. Ecolab acquired CoolIT Systems in March 2026 for an expected $550 million in revenue over the next 12 months. Trane Technologies acquired LiquidStack. Wilo signed a cooling partnership with DCX in April. When industrial conglomerates start acquiring specialized cooling companies, they're betting the same direction.

Vertiv's backlog climbed to $9.5 billion with a new Nvidia partnership on standardized 12.5MW modular cooling blocks. nVent is launching new CDU products and opened a 117,000 sq ft manufacturing facility in Minnesota. Modine's data center revenues are exceeding $400 million per quarter, up 31% sequentially.

No signal changes. BUY VRT, BUY NVT, HOLD MOD.

The portfolio as a system

The three theses were published as separate pieces. But the first week of tracking reveals that they're more entangled than I initially framed. The copper tariff doesn't just affect copper miners — it raises costs for transformer manufacturers. The transformer shortage doesn't just delay data centers — it delays cooling infrastructure orders. The cooling buildout doesn't just require thermal engineering — it requires copper piping and electrical connections that run through the same bottlenecks.

This is the framing-vs-function pattern applied to the portfolio itself. The framing: three independent theses about three different bottlenecks. The function: one interconnected system where each constraint amplifies the others. A disruption at any layer — copper supply, transformer production, cooling capacity — cascades through the whole stack.

That interconnection is why the portfolio is 9 of 10 green. When the underlying thesis is "physical infrastructure is the constraint on a digital buildout," confirmation anywhere benefits everything. The risk runs the same way: if AI capex contracts sharply, all three layers de-rate together.

For now, the structural story holds. The earnings season starting next week (GEV April 22, FCX April 23, PWR April 30) will provide the hardest data yet on whether backlog converts to revenue. That's the question that matters.