Power and Grid Infrastructure, Key to AI Revolution
Highlights:
Market resilience with Tech reasserting leadership: The S&P 500 and Nasdaq have surged to record highs, driven by a historic 18-session winning streak in the semiconductor index while ripping +38% on the month. If one wants to nit-pick, this rally is lacking broad participation; if the "Mag7" and semiconductor stocks are excluded, the broader equity market is actually lower than its mid-January levels.
Geopolitical and Macroeconomic Headwinds: While the stock market has largely "moved on" from the two-month U.S.-Iran war, significant risks persist, including WTI oil prices trading above $95/bbl and global crude inventories being drawn down at a record pace. These factors, combined with surging semiconductor prices, may impact future margins and guidance for major tech companies reporting earnings this week.
The AI-Infrastructure Bottleneck and Policy Response: The "AI revolution" is facing critical constraints in power and grid infrastructure, with transformer lead times now running three to four years. In response, the Defense Production Act (DPA) was invoked via Presidential Determination 2026-10 to prioritize grid equipment as essential to national defense, shifting the investment framework from climate policy to national security.
Full Commentary
The Tech sector, led by the Mag7 and the semiconductor sector, has been on a tear since the ceasefire was announced. The Nasdaq gained +1.5% last week, which added to its breathtaking +17.5% rally over the past four weeks. Since late March, the Nasdaq Composite has gone from being down nearly -10% on the year to +6.9% as of Friday's close – talk about an about-face. The S&P 500 has also vaulted ahead +12.0% in the past four weeks to fresh record highs and has now swung into the green to the tune of +4.5% on a YTD basis. Few things have been as impressive as the meteoric surge in the SOX index, which has not experienced a single down day in April and is up by an incredible +38% this month alone. The Semiconductor index has now risen in eighteen straight sessions, the longest winning streak since the composite was created back in 1994. From a technical perspective (see RSI readings in the chart below), the sector looks to be getting ahead of itself on a short-term basis, but given the surge in earnings we’re seeing, it's hard to say this move isn’t fundamentally justified.
It’s been impressive to see the market continue to add to all-time highs last week, despite WTI oil prices ripping 13% to $94, short-end bond yields rising 7 bps, and AI disruption names coming back under significant selling pressure. Other moves catching my attention are the inability of gold and Defense stocks to catch much of a bid during this conflict. The likes of Lockheed Martin, Raytheon, and Northrop Grumman posted great earnings results, yet the Dow Jones U.S. Select Aerospace & Defense Index is down nearly 15% since missiles started flying in the U.S. / Iran conflict – a rather meaningful underperformance relative to the S&P 500, gaining +3% over the same time frame.
The stock market has clearly moved on from the war. Headlines coming out of the weekend indicate that Iran has floated a pathway where the U.S. gets a reopening of the Strait in exchange for a deferred discussion on nuclear capabilities. This gives President Trump an off-ramp if the objective is simply to reopen trade flows and move on, but I would think this is a bit of a non-starter if the terms mean flows still run through the Iranian-controlled route. Not to mention, I don’t see Trump caving on the nuclear piece with markets at all-time highs and the economy holding steady. My sense is that neither side is experiencing enough pain locally to make them think they have to cave on their demands, which probably means the current status quo will be maintained until something materially shifts the balance.
Well, the same stock market that withstood the global pandemic, the Russian invasion of Ukraine, the huge inflation and rates cycle, the crisis in regional banks, and the tariff file, has now skated right through this two-month U.S.-Iran war and all the damage to global supply chains that has been left in its wake. We have the equity market at new highs, even though energy costs and market interest rates remain far higher than they were before the conflict in the Middle East began. Goldman Sachs strategists, meanwhile, just lifted their Q4 oil price forecast (for Brent) to $90 per barrel (from $80 per barrel) as global crude inventories are now being drawn down at a record clip (matching the previously upped projection by Morgan Stanley).
Whether or not the performance of equity markets makes any sense will ultimately be arbitrated in the fullness of time. As of now, the resilience of global economic growth and a stellar corporate earnings season (so far) are outweighing the risk of an estimated 10 million barrels per day of oil from the Persian Gulf being sequestered from the global oil market. Not to mention the intense pressure building in the jet fuel and petrochemicals markets. Although maybe the equity market is efficiently arbitrating the events of the last two months, where 118 S&P 500 stocks are still down more than 10% from their pre-war levels, these are the ones vulnerable to the rising costs for raw materials, fuels, and those dependent on sales to customers. On the other hand, I see more than 80 companies whose share prices are up more than +10% since the end of February, and these companies are concentrated in the large-cap AI-related space. All of that seems to check out with reality.
Digging a little deeper, and this data point is a bit more concerning, but if you strip out the Mag7, the U.S. equity market is down over the past two months. The S&P 500 equal-weight index, while off the bottom, is no higher today than it was in early February. So, the cumulative Advance-Decline line looks very constructive, but I sense the participation is not exactly that broad. In fact, strip out one subsector – the red-hot semiconductor group – and the stock market is still lower today than it was in mid-January.
As for the five Mag7 names reporting this week, I believe all of them will beat estimates for current Q1. Having said that, I could see the following issues with guidance for the above companies: 1) booked orders impacted at the end of Q1 by the Iran war, 2) the surge in semiconductor prices impacting future margins, 3) rising oil prices raising shipping costs, 4) falling consumer confidence slowing future ad budgets, 5) a pull forward in demand due to anticipated price increases resulting in less second half growth, and 6) rising AI spend impacting cash flow. As a reminder, despite stellar Current Q4 results reported in January, the five names reporting next week declined 1% on average in reaction to earnings the next day, with three of the five down. Given the 20% increase in Mag7 stocks since their March 30th bottom, which has driven the S&P to technically overbought levels, the risk-reward after this sharp rally is not great, and a shallow correction would not surprise me in the near term.
Meanwhile, sentiment and positioning readings have shifted meaningfully back to a bullish posture with the NAAIM exposure index pushing into the 90s, the CNN Fear and Greed index nearing extreme greed territory, call buying overwhelming put hedging, and the VIX index falling towards its low for the year. In a nutshell, I’m of the view that the risk/reward from here is a ‘pair of twos’ – not terrible, but not great either. What’s clear is that we’re in a bull market, and the primary trend is higher. That doesn’t mean there is nothing to worry about, but aggressively positioning for worst-case outcomes is proving costly when Mr. Market is clearly telling you that’s not his primary focus.
I have talked and written about how ‘power generation’is the most obvious intersection between several of our most convicted long-term investment themes – industrial renaissance (onshoring, grid infrastructure, electrification, supply chains as a matter of national security…) and energy / critical minerals as a chokepoint for the AI revolution. On April 20, the President signed the Presidential Determination 2026-10, a Defense Production Act (DPA)Section 303 determination for grid infrastructure, equipment, and supply-chain capacity.
“On January 20, 2025, I issued Executive Order 14156 (Declaring a National Energy Emergency), under the National Emergencies Act. That order found that America’s inadequate energy production, transportation, and infrastructure constitute an unusual and extraordinary threat to the Nation’s economy, national security, and foreign policy ...
The Nation’s capacity to design, produce, and deploy large-scale grid infrastructure, including transformers, high-voltage transmission components, advanced conductors, power electronics, substations, and grid-supporting manufacturing equipment, is dangerously limited.”
The text formally classifies transformers, transmission lines, conductors, substations, high-voltage circuit breakers, power control electronics, protective relay systems, capacitor banks, and electrical core steel as essential to national defense. The reality, though, is that even after years of upward revisions, the grid backlog only continues to grow. In 1Q26, the quarterly net addition to GE Vernova’s (GEV US) electrification segment backlog was nearly as large as the annual additions from 2022-2025. Growth is accelerating, not abating. Blackstone President Jon Gray shed further light on these constraints last week during a Bloomberg interview:
"The build-out of artificial intelligence infrastructure is the single-biggest driver for the company right now, as global demand for computing power outstrips supply. Eight of the top 10 investments for us in terms of performance in the quarter were around data centers, liquid natural gas and battery storage, and it’s really rippling through our firm. It's clearly the biggest driver for us today."
This is similar to the constraints we’re seeing in the semiconductor space, where AI platforms regularly hit compute limits with engineering teams effectively competing to consume as much compute as possible, where there is a real fear that under-spending = career risk.
Anthropic is now throttling Claude Pro/Max during peak hours (5 AM – 11 AM PT);
OpenAI is reported to be sunsetting older models (e.g., GPT-4o) to free up capacity, and has moved away from “unlimited” subscription and forced higher users to higher-priced buckets (ditto Anthropic).
Microsoft employees expect GPU wait times for cloud customers to persist through 2026 as it (and other cloud providers) keeps more GPUs for internal efforts, bigger customers.
Comes after the recent Anthropic Claude Mythos launch, which came with some speculation the firm is working to shift/scare compute allocation to highest-value uses, reports users hitting limits faster.
This all creates a perverse incentive to spend aggressively…even inefficiently. Constraints are now everywhere…power, CPUs, GPUs, copper, engineers, but transformers are a particularly acute bottleneck, especially given the historical reliance on Chinese producers. Friendlier suppliers are seeing record backlogs and have announced U.S. capacity expansions. This is where recent DPA Title III announcements aligns policy with need while converting a long-running complaint into an actionable plan. Large power transformer lead times are running three to four years, and distribution transformer backlogs sit at nearly a year.
The DPA invocation does three things, in order of importance: 1) it routes priority ratings ahead of other industrial uses, which compresses lead times for grid-end customers and extends them for everyone else, 2) it provides federal capital that crowds in private investment and protects existing U.S. manufacturers from being undercut by Chinese capacity additions, 3) it replaces a climate-policy frame with a national-security frame, and national-security framing survives administration transitions in a way that subsidies for renewable energy did not.
Right now, the U.S. relies heavily on four Korean industrial companies as the marginal suppliers for the transformer backlog. Combined, these four companies have an order backlog that sits at $23.9B as of Q3 2025, representing five to six years of work. No wonder the Korean stock market is one of the best performing markets in the world ytd – that, and its exposure to two dominant memory companies. The bottleneck is great enough to warrant a national security concern and DPA action. This has been and continues to be an investable theme that extends years, not months or quarters, and can be expressed in several different verticals: industrial grid infrastructure companies, critical metals like copper, aluminum, platinum, specialized steel, and rare earths; nuclear energy across the entire supply chain, natural gas turbines, renewable energy, and then there are the various tech AI adjacent themes.
You cannot have an AI revolution without energy / power to operate it, and all the other ancillary inputs that go into building out the necessary infrastructure.
The articles and opinions in "Capital Market Musings and Commentary" are for general information only, and not intended to provide specific investment advice. Performance, dividends and other figures have been obtained from sources believed reliable but have not been audited and cannot be guaranteed. Past performance does not ensure future results. Investing inherently contains risk including loss of principle. Advisory services offered through Casilio Leitch Investments, an SEC registered investment advisor. Copyright © 2023 Casilio Leitch Investments. All Rights Reserved.