Signal Briefs

The first Signal Brief in this series focused on a market imbalance: AI investors have over-indexed on Compute while underappreciating Interface. That brief also surfaced an early signal of physical scarcity — Meta and Apple both flagged memory pricing as a binding input constraint, and HBM is sold out through 2026.

The second signal extends that thread.

AI is no longer just a software race. It is no longer just a model race. It is not even only a semiconductor race.

AI is becoming an industrial-scale infrastructure project — and every infrastructure project eventually hits the same wall: power.

Within the Four Forces of AI Power, Energy is the least abstract force. It does not live in code, models, or user interfaces. It lives in turbines, substations, switchgear, transformers, cooling systems, grid connections, and available megawatts.

This earnings cycle made that clear. Every major company in the energy-and-infrastructure layer beat and raised.

1. Energy Is Becoming the Hard Ceiling

Compute can be financed. GPUs can be ordered. Data centers can be planned.

But electricity cannot be wished into existence.

The next stage of AI deployment depends on whether power can be generated, delivered, conditioned, cooled, and connected fast enough to support the capacity the market is already demanding.

This is why Energy is no longer a secondary operating input. It is becoming a primary growth constraint — and the companies positioned to relieve it are now reporting like AI infrastructure beneficiaries, not utilities.

2. GE Vernova: Grid and Generation Visibility

GE Vernova's Q1 results showed how quickly AI and electrification demand are moving into the industrial layer.

The company reported $18.3 billion of orders, up 71% organically, and revenue of $9.3 billion, up 16%. Backlog grew by $13 billion sequentially to $163 billion, and management pulled forward its $200 billion total backlog target to 2027 from 2028. Gas Power equipment backlog and slot reservation agreements grew from 83 GW to 100 GW, with at least 110 GW expected by year-end 2026.

The most important signal was inside Electrification: GE Vernova booked $2.4 billion in equipment orders to support data centers in a single quarter, more than all of 2025.

That is not normal utility demand. That is AI demand becoming grid demand.

3. Bloom Energy: On-Site Power as AI Infrastructure

Bloom Energy showed the other side of the Energy force: speed.

Where grid-scale infrastructure is slow, on-site power becomes valuable because it can bypass the waiting line.

Bloom reported Q1 revenue of $751.1 million, up 130% year over year, with adjusted EPS of $0.44 versus expectations of roughly $0.12. The company raised full-year 2026 guidance to $3.4 to $3.8 billion in revenue and $1.85 to $2.25 in adjusted EPS — an 80% growth midpoint that lifted the low end of the new range above the top of the prior range.

The marquee proof point came from Oracle's Project Jupiter, where Bloom was selected as the sole power provider for up to 2.45 gigawatts of AI capacity, displacing planned gas turbines and diesel backup. Bloom delivered a fully operational fuel cell system to Oracle in 55 days. Management noted that more than half of the data center backlog comes from customers other than Oracle, including hyperscalers, neo-clouds, and colocation providers.

Bloom is not just selling power equipment. It is selling time-to-power.

In an AI buildout constrained by grid availability, that matters.

4. Vertiv: Power and Cooling Become the Data Center Nervous System

Energy is not only generation. It is also thermal management, power distribution, and uptime infrastructure.

Vertiv sits directly in that layer, and Q1 confirmed it.

Revenue reached $2.65 billion, up 30%, with adjusted EPS of $1.17 and adjusted operating margin expanding 430 basis points to 20.8%. Backlog stood at $15 billion, up 109% year over year. Q4 2025 organic orders had grown 252% — the strongest order quarter in company history — and the momentum carried directly into Q1. Management raised full-year 2026 revenue guidance to $13.5 to $14.0 billion.

That fits the Four Forces map cleanly.

If Compute is the engine, Vertiv supplies the system that prevents that engine from overheating, failing, or becoming unusable at scale. AI does not just require chips. It requires conditioned physical environments where those chips can operate continuously — and Vertiv's 30%+ incremental margins on its growing backlog show that this layer is monetizing AI demand at a different rate than the chips themselves.

5. Quanta Services: The Grid Buildout Layer Is Not a Footnote

Quanta sits in the part of the Energy force that builds and upgrades the grid itself: transmission, interconnection, substation work, and utility infrastructure.

This layer is often described as slow. The Q1 print said otherwise.

Revenue was $7.87 billion, up 26% year over year. Adjusted EPS was $2.68, adjusted EBITDA hit a record $686 million, and total backlog reached a record $48.5 billion. Management raised 2026 revenue guidance to $34.7 to $35.2 billion and now expects technology and load center revenue to grow 110%, up from a previous 70% expectation. The stock jumped over $71 to $699.90 on the print.

Transmission and substation work do not create the AI narrative headline. But this quarter, Quanta's numbers said that grid construction is no longer a quiet input to the AI buildout.

It is part of the AI supply chain.

6. Dominion: Data Centers Are Already Changing Utility Earnings

The signal is also showing up at the utility level.

Dominion Energy beat quarterly profit estimates, with operating EPS of $0.95 against an $0.86 expectation. Its Virginia segment generated operating earnings of $670 million versus $561 million a year prior, a 19.4% increase. Contracted data center capacity reached approximately 51 GW as of March 2026, up 2.5 GW since December and more than triple the 16.5 GW reported in July 2023.

CFO Steven Ridge described the demand as "accelerating and durable."

That matters because Virginia is one of the world's most important data center regions. When data center demand starts moving utility earnings, Energy is no longer a background assumption. It is a measurable AI growth factor.

7. Valuation Signal: Scarcity Moves From Chips to Power

The first phase of AI scarcity was Compute.

The next phase is Energy — and within Energy, scarcity is showing up in two layers simultaneously: components (memory, HBM, transformers, switchgear) and capacity (megawatts, grid interconnection, cooling).

Using a growth-and-scarcity lens, the energy layer is becoming increasingly important because substitution is difficult. A model can be swapped. A chip vendor can be diversified. But grid capacity, power delivery, thermal infrastructure, and the components that build them are location-bound, time-constrained, and physically scarce.

That creates a different kind of value capture.

The companies positioned closest to power availability, electrical equipment, cooling, and grid execution are no longer merely industrial suppliers. They are becoming enablers of AI deployment — and the market is now pricing them that way.

The Signal

The AI market has spent two years asking:

Who has the most compute?

The better question is becoming:

Who can power the compute?

Compute defines capability.Interface defines control.Energy defines the ceiling.

This earnings cycle showed that ceiling is no longer theoretical. GE Vernova, Bloom, Vertiv, Quanta, and Dominion all beat and raised in the same window — a synchronized signal that is rare in industrial cycles and almost unheard of in utility-adjacent ones.

AI's next bottleneck is not only intelligence.

It is electricity, the components that deliver it, and the time required to install them at scale.