However, as power generation units and gas supply gradually homogenize into standardized shelf commodities, the core competitive focus of the industrial chain has quietly shifted upward. Major electrical giants and energy solution providers are no longer merely competing on hardware equipment; instead, they are vying for control over the "system integration and electrical control layer" spanning from power generation to computing chips.

This article conducts an in-depth analysis of the latest strategic alliances and commercial retreats by industry titans such as Schneider Electric and Caterpillar, revealing the ultimate decisive factor in the transition of on-site power supply models from engineering packages to commercial deployments: whoever controls the control logic and single-line diagram architecture will establish the deepest commercial moat throughout the two-decade lifecycle of computing capacity.

Cleanview’s latest tracking data shows that 59 data center campuses across the U.S. have collectively announced approximately 90 gigawatts (GW) of “behind-the-meter” (BTM) self-generation capacity, with about 82 GW declared in concentrated bursts since early 2025.

Yet only around 2 GW of this BTM capacity is currently operational.[1] This massive gap precisely underscores the novelty of this demand surge, rather than instability in project pipelines.

Before 2025, on-site power generation was seen merely as an emergency backup solution; today, it has become a hard requirement for timely commissioning of data centers. Project timelines have emerged as the most stringent hard constraint: while new data center construction in North America is setting historical records, traditional grid interconnection queue channels are being completely exhausted.

Energy research firm Enverus provides clear installation forecasts: between 2026 and 2030, 25.5 GW of new industrial power demand across the U.S. will be met through BTM self-generation, with data centers alone accounting for 22.5 GW—approximately 36% of all new data center capacity added during this period.[2]

Sustaining this massive load requires 31.6 GW of additional on-site power generation equipment, including 29.6 GW of gas-fired generation. Reciprocating internal combustion engines represent 7.1 GW, small- and medium-sized gas turbines 6.2 GW, and fuel cells 4.8 GW.

Figure: 59 U.S. data center campuses have announced roughly 90 GW of BTM installed capacity; Enverus forecasts BTM load reaching 22.5 GW by 2030 (Source: Cleanview & Enverus Intelligence Research)

These figures depict vast quantities of hardware, but they do not answer: who should be held accountable when an entire power plant fails to start on schedule?

The series of announcements made by suppliers at Yotta 2026 directly address this central issue—the most compelling aspect of the entire industry: while the technical path for power generation has largely settled, its commercial execution framework remains highly uncertain.

Avanza previously systematically mapped the scale of on-site power expansion, typical campus profiles, and the five key technological drivers behind it. However, one critical layer remains unclear: who integrates all components? Who will control core customer relationships over the next twenty years? And who is actively exiting this layer?

Published Partnership Frameworks

On September 30, Wärtsilä, Schneider Electric, and Stanley Consultants jointly launched a full-cycle solution for U.S. data center power called “Generator-to-Chip.”[3]

The division of labor among the three parties is extremely clear: Wärtsilä provides modular gas engine power plants; Schneider Electric handles electrical architecture, automation systems, and digital energy management; Stanley Consultants oversees engineering design, permitting, multi-party coordination, construction management, and commissioning oversight.

Schneider packages this system under the concept of the “AI Energy Park.” Melton Chang, Executive Vice President of Schneider Electric’s Power Systems, clearly defined it: “The AI Energy Park starts from compute load, integrating generation, electrical infrastructure, automation, and digital technologies into a cohesive system.”[3]

Risto Paldanius, Vice President of Energy for Wärtsilä Americas, also revealed their intent: “By integrating our complementary capabilities from the outset, we can streamline project execution, reduce handoff losses, and significantly shorten construction timelines.”

Upon close examination of these verbs—collaborate, simplify, reduce handoffs—this represents a coordinated action agreement among three independent companies aiming to present themselves as a unified alliance. This is not a turnkey contract delivery; the announcement itself makes no such commitment.

Unpublished Critical Elements

The announcement contains no contractual structure, no megawatt capacity metrics, no disclosed client names or pilot campuses, and no pricing information. StorageReview, in an independent report on this collaboration, noted that the release “provided no numbers, clients, or timelines.”[4]

To date, Schneider’s publicly disclosed materials cover only technical engineering aspects: the generator-to-chip division of labor, keynote speeches at Yotta, and whitepapers on switchgear.

The actual commercial terms—specifically, who signs contracts with end customers and who leads long-term business relationships—remain entirely absent from public records. If this were a real, signed client, it would typically be announced or at least disclose the installed capacity, yet there has been no such movement.

The most logical interpretation is that the industry, including Schneider, is advancing the “engineering productization” of on-site power far faster than its “business modelization,” which is often significantly more challenging.

The Platform That Once Addressed These Issues No Longer Exists

For seven years, Schneider held a dedicated entity capable of resolving all these commercial challenges.

AlphaStruxure was founded in 2019 as a joint venture between Schneider and The Carlyle Group. It was responsible for designing, financing, building, owning, operating, and maintaining on-site energy infrastructure, offering fixed electricity rates and performance guarantees backed by contractual obligations.

It undertook the 11.4 MW microgrid project for New York’s JFK Airport’s new Terminal One, built in phases through 2029;[5] and in April 2026, completed the largest bus depot microgrid in the U.S. in Montgomery County, Maryland—featuring 4.8 MW solar PV, 2 MW storage, and permanent islanded operation during grid outages.[6]

Yet Schneider no longer describes itself as the parent company. As late as April 19, 2026, AlphaStruxure’s official homepage still stated, “Our parent company is Carlyle and Schneider Electric.”

But by May 8, the same sentence was subtly revised to the singular form: “Our parent company Carlyle,” with Schneider’s name entirely erased. The surrounding text remained unchanged word-for-word, indicating a highly targeted edit—and this statement remains on their website today.[7]

This webpage update was not accompanied by any official transaction announcement—neither press releases, SEC filings, nor industry media coverage exist. Carlyle’s quarterly reports cannot confirm this either, as asset management firms’ 10-K filings disclose only management fees and AUM, not individual holdings within their funds.

Yet one undeniable objective fact stands: Schneider no longer appears as a parent entity and has exited the vehicle previously responsible for asset ownership and commercial contracting.

The “Energy-as-a-Service” (EaaS) model is unquestionably valid. “Zero-CapEx” on-site power is an active and capital-rich competitive arena: VoltaGrid raised $1 billion from Blackstone and Halliburton in May 2026, achieving a valuation exceeding $10 billion, fueled by its ability to own and operate power plants and sell electricity directly to customers.[8]

Yet precisely at the moment when on-site self-generation has become the hottest battleground across the industry, Schneider has voluntarily retreated from this high-capital investment tier.

Why This Might Be a Prudent Move

From a positive and arguably correct perspective, this reflects a strategic decision about “which layer is worth owning long-term.”

What Schneider firmly retains is the electrical control and automation architecture layer—and the company is demonstrating its capital logic with real investment: the control layer is the ultimate choke point of the entire chain.

At Yotta 2026, Schneider unveiled its software-defined medium-voltage switchgear, claiming to be globally pioneering, fully decoupling system intelligence from physical hardware. Schneider asserts this solution reduces control wiring by 87%, terminal blocks by 91%, and copper usage by 85% compared to traditional designs, increases order-to-manufacture speed by up to three times, and cuts on-site commissioning time in half.[9]

Pilot projects will continue through 2027 and the product will be fully launched in 2028. Some technology stacks are already earlier to market: 2 MW, 2.25 MW, and 2.5 MW prefabricated power modules based on Galaxy VXL UPS have already been sold in Europe and are promised to enter the North American market before the end of 2026.[10]

Whoever leads the creation of the one-line electrical schematic and control logic naturally becomes the default leader for equipment commissioning, operations and maintenance, and system upgrades throughout the asset’s 20-year lifecycle.

Generators will eventually be replaced and server racks updated every three to five years; but underlying electrical topology and relay protection schemes endure, preserving deep and unbreakable commercial ties with system operators.

Moreover, a pure control systems positioning enables horizontal compatibility and sales across all projects in the market. Schneider has precedent: in March 2024, it partnered with Mainspring Energy, bundling its proprietary control and integration systems into third-party linear generator peripherals.[11]

This partnership with Wärtsilä simply applies the same logic to higher-power engines. This approach can be seamlessly replicated with Caterpillar power plants, Bloom fuel cell units, or utility-owned generators. Owning an energy park captures gains from a single site; controlling the system integration layer allows capturing value across all energy parks.

Strong financial performance fully validates this bet. Schneider’s first-half 2026 revenue reached €21.2 billion, with organic growth of 14.0%, and Q2 alone achieved €11.5 billion, growing organically by 16.5%.[12]

Q2 orders for the pure data center segment saw triple-digit growth, while broader data center and network edge markets maintained strong double-digit sales growth. The “Systems” business model—carrying prefabricated components and EPC services—accounted for 35% of Q2 revenue and delivered an impressive 28% organic growth.

Two Reasons to Hold Judgment

First, the impressive metrics for the switchgear come exclusively from Schneider’s own White Paper #303, without any third-party audit, lack of validation from core clients like Equinix, and no external pilot test data has been disclosed.[13]

Until a real benchmark client publicly endorses these claims, assertions of “87% fewer wiring, 91% fewer terminal blocks, 85% less copper,” and multi-fold speedups should be viewed cautiously—as standard vendor marketing rhetoric.

Figure: Schneider Electric’s claimed performance metrics for its software-defined MV switchgear (Source: Schneider Electric, un-audited)

Technology media The Register posed a sharper critique, which Schneider’s public materials have so far avoided: “A faulty or malicious software update on a critical electrical system affecting human safety and equipment integrity carries consequences vastly different from accidentally clicking a desktop patch on an office computer.”[13]

Migrating MV electrical protection and control logic to software brings efficiency gains but fundamentally entails a massive operational risk swap.

A data center choosing to go fully isolated from the public grid to bypass interconnection queues simultaneously loses its last line of defense. At this juncture, this purely software-based protection system becomes the sole barrier against instantaneous load surges and total blackout across the entire campus.

Converging Forces Around a Single Keyword

The striking similarity in naming signals a structural industry repetition, not a mere marketing gimmick by individual firms.

As early as June 2025, Eaton and Siemens Energy jointly bundled 500 MW standard gas turbine units with energy storage systems, quantifying returns: typical 500 MW facilities could generate an additional $2 to $3 billion in revenue by accelerating commissioning by up to two years.[14]

Vertiv and Caterpillar followed suit in November 2025, forming a partnership where Vertiv CEO Gio Albertazzi explicitly defined it as a “Grid-to-Chip” framework.[15]

Schneider and Wärtsilä entered the scene in September 2026, branding their initiative “Generator-to-Chip.”

In just fifteen months, three pairs of industry titans have formed dense alliances, with two even adopting near-identical “X-to-Chip” messaging frameworks. The industry has now converged on a consensus: the smallest unit of competition is no longer a single engine or switchgear, but the complete channel from fuel source to end-user computing power. What remains unresolved is who will lead the sales of this channel.

Figure: Six integrated packaging models for on-site power and role delineation across integration layers (Source: Avanza Energy Analysis)

Meanwhile, two business models remain entirely independent of this bundling race, defining the industry’s boundaries.

One is utility-scale power companies owned by investors, which directly own power plants, purchase generators from OEMs without retaining any software layer, and sell electricity directly to data centers.

The other is project developers who finance, build, own, and operate power plants on-site and sell electricity directly to clients—exactly the core model enabling VoltaGrid’s success, and the path once walked by AlphaStruxure.

Beneath these five pathways lies the fundamental competitive baseline for all bundling solutions: customers sign three fragmented procurement contracts independently and bear full integration responsibility. This results in six distinct combinations for deploying identical megawatt capacities, with integration control layers assigned to different parties in each model. Only in Schneider’s model is the integration layer firmly entrusted to the software provider writing the control logic.

Regulatory evolution has made this commercial issue exceptionally urgent, progressing far beyond what was evident at Yotta 2026. Under Section 206 of the Federal Power Act, the Federal Energy Regulatory Commission (FERC) issued a rebuttal order on December 18, 2025, determining that PJM’s grid procedures lacked sufficient clarity and consistency regarding rate structures and interconnection terms for serving “co-located loads” with newly installed generation capacity, and imposed a regulatory definition.

PJM submitted its compliance plan on January 20, 2026, and FERC formally ruled on April 16, 2026, approving some parts and rejecting others.[16]

The portions accepted by FERC provide direct, substantive benefits to energy parks: applicants seeking interconnection services for co-located loads may now apply based on actual load, below the generator’s nameplate rating, and PJM must calculate system upgrade requirements according to the reported actual capacity—not the machine’s theoretical maximum output.

Customers can fully leverage existing acceleration clauses, skipping subsequent complex assessments under minimal grid impact, applying for temporary interconnection service to commission power ahead of full facility completion, and asserting rights to excess capacity services at existing interconnection points.

It is precisely these four mechanisms that transform “grid flexibility” from a glossy brochure claim into legally enforceable tariff provisions—explaining why a data center campus opting for islanded self-generation today could, in the future, comply with regulations to sell peaking flexibility back to the bulk grid.

The rejected portions, however, deliver an unflinching warning: rules are still undergoing rapid and volatile evolution. FERC decisively denied PJM’s attempt to redefine “interconnection point” as “ownership change point” within the co-located load definition, and dismissed PJM’s proposed changes to the BTM self-generation application process on grounds of exceeding the current scope, ordering further remediation.

The deeper conflict is now laid bare: PJM’s independent market monitor (IMM) has warned that the current framework offers no clarity whatsoever: how should the public grid absorb massive, sudden power injections from self-generating plants when large-scale computing loads collapse instantaneously? Or how should the grid plan backup power when a self-generated generator suddenly trips offline?

FERC temporarily deferred these issues, ruling them outside the current case scope. Now, these century-defining challenges have been transferred to a dedicated super-load docket. FERC has acted: on June 18, 2026, it simultaneously issued Section 206 rebuttal orders to all six Regional Transmission Organizations (RTOs) and Independent System Operators (ISOs) nationwide, requiring them to demonstrate the rationality of their current procedures involving super-large loads and BTM self-generation, or face full reconstruction.[17]

After an August extension, final formal responses from each region were due by mid-November. This means: reliability standards for a formerly isolated campus wishing to sell power back to the grid are now being hastily drafted across regional grids—with no national consensus yet. Avanza previously deeply analyzed FERC’s 60-day countdown window for the six major grids and argued why strict project environmental review permits, not mere FERC procedural changes, constitute the ultimate bottleneck blocking BTM gas generation.

Key Observations for the Future

Three decisive indicators will reveal the ultimate outcome of this business model faster than any official press release:

First, the first officially named client and contract format. If the final agreement is a single, comprehensive turnkey contract covering all responsibilities, it indicates that Schneider or its competitor has successfully captured the system integration layer and locked in customer stickiness.

If clients still sign three separate, fragmented equipment and engineering contracts, it implies customers remain the ultimate system integrators—the so-called ecosystem bundling is merely a scheduling convenience to accelerate timelines, not a true business model transformation.

Second, whether project financing secures funding and who bears the burden on whose balance sheet. Schneider has participated in syndicated financings multiple times; notable is the role played by participants.

In September 2025, Schneider joined forces with Microsoft and over twenty allies to launch the Accelerated Resilience Infrastructure Initiative, securing $7.5 billion in dedicated funding for six project developers, delivered via EaaS contracts. Schneider explicitly positioned itself as a strategic integrator, facilitating client-developer matchmaking and technology delivery, while one of the six fund recipients—AlphaStruxure—was itself a former partner.[18]

Just seven months later, Schneider’s name had been completely removed from AlphaStruxure’s parent list. Regardless of the underlying equity mechanism, the strategic direction remains consistent: Schneider firmly secures the integrator seat, leaving heavy asset-holding responsibilities to other giants’ balance sheets.

Massive external capital is rapidly flowing in. In June 2026, Brookfield expanded its strategic partnership with Bloom Energy for AI computing power infrastructure fivefold to $25 billion, specifically targeting investments in on-site power infrastructure.[19] Data centers initially adopted BTM self-generation merely because it was the fastest route to power access; now, capital markets are creating massive funding pipelines tailored for this niche—marking a clear milestone in the sector’s maturation.

Finally, whether the bet on the electrical control layer can withstand real-world testing. Schneider’s standardization legacy in this domain is far more extensive than implied by the recent Yotta 2026 announcement.

Back in 2023, it launched the EcoStruxure Microgrid Flex standardized microgrid solution,[20] and continued expanding its channel and financing partners throughout 2025.

In March 2026, Schneider officially inaugurated a microgrid test lab at its global R&D center in Andover, where its Global Senior Vice President of Microgrids explicitly set the goal: “To accelerate deployment of robust microgrid control systems by developing tested, validated, and well-documented reference architectures using more standardized methods.”[21] This reveals that even half a year before the AI Energy Park launch, Schneider had already publicly commercialized microgrid control architecture.

Figure: Timeline from 2019 to 2028, mapping key milestones of Schneider Electric and AlphaStruxure

Schneider has consistently refrained from directly linking the AI Energy Park to its established microgrid business in public forums—after all, specific product lines within a multinational giant often operate independently from the early departments that nurtured their capabilities. Yet this clear chronological sequence holds significant reference value: it provides the strongest evidence that Schneider’s full transition toward the control layer is a deliberate, premeditated strategy—not a spontaneous speculative trend.

Jim Simonelli, Senior Vice President and Chief Technology Officer of Schneider Electric’s Safety Power Business, bluntly revealed the underlying reality at NVIDIA’s GTC conference in April: “Three years ago, the industry’s biggest challenge was evacuating heat from cabinets; today, the real challenge is getting power into the cabinet and delivering it to the campus.”[22] When discussing scalable on-site power for AI-grade demands, he was even more candid: “Besides natural gas, there is currently no viable alternative.”

Natural gas is already a highly mature undifferentiated commodity, and reciprocating engines are off-the-shelf industrial catalog items. Within the entire chain, only the system integration and electrical control layer can sustain compound value through technological accumulation.

And the supplier who first secures paying customers to implement their integration solution will definitively unveil the true commercial ceiling and ultimate valuation of this critical control layer for the entire industry.

Author: Christopher Johnson (Avanza Energy), Translated by DeepFlow TechFlow