Behind the U.S. Announcement of 59 Data Centers and 90 Gigawatts: What Critical Layer Are All Parties Competing For?

TechFlow Guide: As computing centers race against time to secure power, Behind-the-Meter generation is evolving from a past emergency backup option into the only lifeline for ensuring delivery schedules. The United States' announced 90-gigawatt plan for off-grid self-supply power plants marks a fundamental restructuring of computing infrastructure, moving away from the traditional public grid.

However, as generator sets and gas supply gradually homogenize into standardized off-the-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, but rather vying for the "system integration and electrical control layer" spanning from power generation to computing chips.

This article provides an in-depth analysis of the latest strategic alliances and commercial retreats by giants such as Schneider Electric and Caterpillar, revealing the ultimate decider in the transition of on-site power supply models from engineering packages to commercial implementation: whoever controls the control logic and single-line diagram architecture can build the deepest commercial moat over a computing infrastructure lifecycle of up to two decades.

According to the latest tracking data from Cleanview, 59 data center campuses across the US have cumulatively announced approximately 90 gigawatts (GW) of "Behind-the-Meter" autonomous generation capacity, with about 82 GW announced intensively from the beginning of 2025 to date.

However, the currently operational Behind-the-Meter capacity stands at only about 2 GW. [1] This massive gap precisely highlights how new this demand surge is, rather than indicating instability in project pipelines.

Prior to 2025, on-site self-supplied power was viewed solely as an emergency backup; today, it has become a rigid requirement for data centers to be powered and operational on schedule. Project timelines have become the harshest hard constraint: while the construction scale of North American data centers hits historical records, traditional grid connection queue reserves are being thoroughly depleted.

Energy research firm Enverus has issued a clear installation forecast: between 2026 and 2030, 25.5 GW of new industrial electricity demand nationwide will be met through the Behind-the-Meter self-supply model, with data centers accounting for 22.5 GW, roughly 36% of all new data center capacity added nationwide during that period. [2]

Supporting this massive load requires 31.6 GW of new on-site generation equipment, with gas-fired generation reaching 29.6 GW. Reciprocating internal combustion generators account for 7.1 GW, medium-to-small gas turbines for 6.2 GW, and fuel cells for 4.8 GW.

Figure: The 59 US data centers have announced approximately 90 GW of Behind-the-Meter installed capacity, with Enverus forecasting Behind-the-Meter load to reach 22.5 GW by 2030 (Source: Cleanview and Enverus Intelligence Research)

These figures depict piles of hardware equipment, but they do not explain: when an entire power plant fails to come online on schedule, who exactly does the client hold responsible?

The series of announcements by suppliers at the Yotta 2026 conference revolve precisely around this core question, which happens to be the most compelling aspect of the entire industry: the technological pathways for power generation equipment have essentially been settled, yet its commercial fulfillment architecture remains fraught with variables.

Avanza has previously systematically outlined the expansion scale of on-site power supply, the typical campuses behind it, and the five major technological routes driving its development. However, what remains unclear is the critical control layer built upon it: Who will be responsible for integrating all components? Who will control core customer relationships over the next two decades? And who is actively choosing to exit this layer?

Announced Partnership Models

On September 30, Wärtsilä, Schneider Electric, and Stanley Consultants jointly launched a full-process solution for US data center power, termed "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 power management; and Stanley Consultants oversees engineering design, permitting, multi-party coordination, construction management, and commissioning supervision.

Schneider has packaged this system into a concept called "AI Energy Park." Melton Chang, Executive Vice President of Power Systems at Schneider Electric, accurately defined it: "The AI Energy Park starts from computing loads, integrating power generation, electrical infrastructure, automation, and digital technologies into an organic system." [3]

Risto Paldanius, Vice President of Energy Americas at Wärtsilä, also clarified its original intent: "By integrating our complementary capabilities from the start, we can simplify project execution, reduce handover losses, and significantly shorten construction schedules."

Examining these verbs closely: coordinate, simplify, reduce handovers. Essentially, this is a concerted action agreement reached among three independent enterprises, aimed at appearing together in an alliance format. This is absolutely not a single-contract turnkey wrap project, nor does the announcement itself make such a commitment.

Unpublished Key Elements

The announcement contains no contract architecture, no megawatt capacity metrics, no specific customer names or pilot campuses disclosed, let alone any pricing. StorageReview, in its independent coverage of the partnership, pointed out that the release "provides no numbers, customers, or timeline." [4]

Public materials disclosed by Schneider to date have only covered the engineering and technical levels: namely, the division-of-labor logic for Generator-to-Chip, keynote speeches at the Yotta conference, and switchgear whitepapers.

As for the actual commercial terms—specifically, who will sign contracts with end customers and who will lead long-term commercial partnerships—the public record remains blank. If there were officially signed real customers, they would typically be publicly announced or at least disclose their installed capacity, but there has been no movement here.

The most logical interpretation is that the entire industry, including Schneider, is far faster at productizing on-site supply through "engineering frameworks" than in commercializing its business models, with the latter often proving far more difficult.

The Platform That Once Held These Questions Has Disappeared

For nearly seven years, Schneider once possessed an entity specifically designed to answer all the aforementioned commercial questions.

Founded in 2019, AlphaStruxure was a joint venture between Schneider and Carlyle Group. It was responsible for designing, financing, building, owning, operating, and maintaining on-site energy infrastructure, while providing fixed electricity rates with performance guarantees backed by contractual fulfillment.

It undertook an 11.4-megawatt microgrid project for the new Terminal 1 at New York's JFK Airport, built in phases until 2029; [5] and in April 2026, completed the largest bus depot microgrid in the US for Montgomery County, Maryland: featuring 4.8 MW of solar PV, 2 MW of storage, and capable of permanent islanded operation during grid outages. [6]

However, Schneider no longer describes itself as the parent company. As late as April 19, 2026, the AlphaStruxure official homepage still read "Our parents are Carlyle and Schneider Electric."

But by May 8, the same paragraph was quietly altered to the singular "Our parent is Carlyle," with Schneider's name completely erased. The surrounding context remained verbatim, indicating an extremely precise targeted edit, and this wording still remains on its official website today. [7]

This webpage copy revision was accompanied by no official transaction announcement, leaving no trace in press releases, SEC filings, or industry media reports. Carlyle's periodic financial reports also cannot verify this, as asset managers' 10-K forms only disclose management fees and AUM, without listing details of individual companies held by their funds.

Yet the indisputable objective fact is: Schneider is no longer presenting itself as a parent company, having exited this vehicle that once directly managed asset ownership and commercial contracting.

The "Energy-as-a-Service" (EaaS) model itself is unquestionable. Zero-CapEx on-site power supply represents an active, capital-rich competitive track: VoltaGrid raised $1 billion from Blackstone and Halliburton in May 2026, breaking through the $10 billion valuation mark, fueled precisely by holding and operating power plants while selling electricity directly to customers. [8]

However, precisely as on-site self-supply has evolved into the most fiercely contested focal track across the entire industry, Schneider has actively retreated from this heavy-asset investment layer.

Why This Might Be a Smart Move

In a positive direction, and one I consider entirely correct, this is a strategic decision regarding "which layer is truly worth long-term ownership."

What Schneider firmly retains is the electrical control and automation architecture layer, and the company is articulating its capital logic with hard cash: the control layer is the ultimate choke point of the entire chain.

Its software-defined medium-voltage switchgear released at the Yotta conference claims to be a global first, achieving complete decoupling of system intelligence from physical hardware. Schneider claims that compared to traditional designs, this solution reduces control wiring by 87%, terminal blocks by 91%, and copper usage by 85%, increases order-to-manufacture speed by up to 3 times, and cuts on-site commissioning time in half. [9]

Pilot projects will run through 2027 before full commercial rollout in 2028. Some technology stacks will land even earlier: 2 MW, 2.25 MW, and 2.5 MW prefabricated power modules based on the Galaxy VXL UPS have already gone on sale in Europe, with commitments to launch in the North American market by the end of 2026. [10]

Whoever takes the lead in drafting Single-Line Diagrams and control logic naturally becomes the default lead for equipment commissioning, maintenance services, and system upgrades throughout the asset's twenty-year lifecycle.

Generator sets will eventually be replaced and phased out, server racks updated comprehensively every three to five years; but underlying electrical topologies and relay protection schemes will stand the test of time, preserving unbreakable commercial ties with system operators alongside them.

Furthermore, a pure control system positioning allows for horizontal compatibility and sales to all projects on the market. Schneider had precedent for this back in March 2024, partnering with Mainspring Energy to package its proprietary control and integration systems around third-party linear generators. [11]

Partnering with Wärtsilä this time simply applies the same logic paired with higher-power engines. This playbook can similarly be seamlessly replicated onto Caterpillar plants, Bloom fuel cell units, or utility-invested units. Owning one energy park allows you to capture revenue from only that one park; whereas controlling the system integration layer lets you take a share of every energy park.

Strong financial performance fully validates the rationality of this bet. Schneider's H1 2026 revenue reached €21.2 billion, with 14.0% organic growth, while Q2 alone hit €11.5 billion, up 16.5% organically. [12]

Orders in the pure data center segment saw triple-digit surges in Q2, and actual sales in the broader data center and network endpoint markets maintained strong double-digit growth. The "System Business Model" (Systems), carrying prefab and EPC operations, accounted for 35% of Q2 revenue, achieving an impressive 28% organic growth rate.

Two Reasons to Withhold Judgment

First, the stunning metrics for the aforementioned switchgear originate entirely from Schneider's own Whitepaper No. 303, without any third-party independent audit, without verification from core clients like Equinix, and with zero pilot field-test data disclosed to the public. [13]

Until real benchmark customers step forward to endorse them, claims regarding "saving 87% wiring, 91% terminal blocks, and 85% copper," plus multi-fold speedups, should reasonably be viewed with caution as routine vendor marketing rhetoric.

Figure: Schneider Electric's claimed performance metrics for its software-defined medium-voltage switchgear (Source: Schneider Electric, unaudited by independents)

Tech publication The Register raised sharper questions, which Schneider's public materials have yet to address: "Executing erroneous or malicious software updates on critical electrical systems governing human and equipment safety carries catastrophic consequences fundamentally different from deploying a faulty desktop patch on an office computer." [13]

Migrating medium-voltage electrical protection and control logic to the software layer, while boosting efficiency, is essentially conducting a massive swap of operational risk.

A data center that chooses complete physical disconnection from the public grid to bypass connection queues simultaneously loses its final baseline safety protection. At this exact moment, this pure software protection scheme has become the sole barrier preventing instant load fluctuations from plunging the entire campus into total darkness.

Forces Are Converging on a Shared Keyword

High similarity in naming indicates a structural industry reenactment, rather than isolated corporate marketing gimmicks.

As early as June 2025, Eaton and Siemens Energy partnered to bundle 500-megawatt standard gas turbine units with energy storage systems, quantifying commercial returns by stating that advancing commissioning timelines by up to two years enables a typical 500 MW facility to generate an additional $2 billion to $3 billion in revenue. [14]

Vertiv and Caterpillar swiftly followed suit in November 2025, with Vertiv CEO Gio Albertazzi explicitly defining it as a "Grid-to-Chip" framework. [15]

Schneider and Wärtsilä then took the stage in September 2026, flying the banner of "Generator-to-Chip."

Within a mere fifteen months, three pairs of industry titans have densely tied marital knots, with two pairs independently adopting nearly identical "X-to-Chip" discourse. The entire industry has reached a consensus: the smallest unit of competition is no longer a single engine or switchgear cabinet, but the complete channel from source fuel directly to terminal computing power; the only unresolved question is who will lead the sales of this channel.

Figure: Six bundled integration models for on-site power supply and the role allocation of parties within the integration layer (Source: Avanza Energy Analysis)

Meanwhile, two business models exist entirely outside this bundling race, defining the industry's boundaries.

One involves utility power companies owned by investors directly and fully owning power plants, purchasing generator sets straight from OEM manufacturers. With the OEM retaining no software layer, the utility company sells electricity directly to data centers.

The other relies on project developers funding, constructing, owning, and operating power plants themselves, selling electricity behind the meter directly to customers. This is precisely VoltaGrid's flagship model, which AlphaStruxure once walked down.

Beneath these five paths lies the ultimate competitive baseline for all bundled solutions: where clients separately sign three fragmented procurement contracts and handle all integration work themselves. This yields six radically different combinations for deploying identical megawatt capacities, with the integration control layer ending up elsewhere in each mode. Only in Schneider's model does the integration layer remain securely handed to the software party writing the control logic.

The evolution of regulatory policy has made this commercial issue unprecedentedly urgent, with a driving force far exceeding the surface appearances shown at the Yotta conference. Under Section 206 of the Federal Power Act, the Federal Energy Regulatory Commission (FERC) issued a Show Cause Order on December 18, 2025, ruling that PJM interconnection rules lack sufficient clarity and consistency regarding rates and interconnection terms for utilizing newly added generator sets to serve Co-located Loads, forcibly imposing regulatory definitions.

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

The parts adopted and validated by FERC constitute direct, substantive benefits for energy parks: Interconnection applicants serving co-located loads can now apply for interconnection services based on actual loads lower than the generator's nameplate rated output. PJM must calculate system expansion and upgrade requirements based on their actual declared capacity rather than the machine's maximum theoretical output.

Clients can fully leverage existing acceleration clauses to directly skip subsequent complex evaluation phases under minimal grid impact, apply for temporary interconnection services to bring facilities online before full completion, and claim surplus capacity service rights on existing interconnection points.

It is precisely these four specific mechanisms that transform so-called "grid flexibility" from glossy brochure rhetoric into legally binding tariff rule clauses; this is also the underlying logic explaining why a computing campus choosing islanded self-supply today could legally resell peak-shaving flexibility to the main grid in the future.

Conversely, the parts rejected by FERC ruthlessly warn that rules continue to evolve amid rapid and turbulent shifts. FERC flatly vetoed PJM's attempt to substitute "point of interconnection" with "change of ownership point" in the co-located load definition, and dismissed PJM's proposal to modify the Behind-the-Meter self-supply application process on grounds of being beyond the scope of this hearing, ordering further remediation for compliance.

Deep-seated contradictions have been thrown entirely into the open: PJM's own Independent Market Monitor (IMM) has issued a warning that the current plan leaves completely undefined how the public grid should absorb massive power shocks accidentally injected backward from self-supply plants when large-scale co-located computing loads suddenly plummet overnight. Or how the grid system should plan fallback power supply if self-supply generators trip offline due to sudden faults.

FERC has temporarily categorized these challenges as beyond the scope of this case, deferring final rulings. Today, these monumental issues are wholly transferred to Special Case Files for Ultra-Large Loads, with FERC taking action: issuing a parallel Section 206 Show Cause Order on June 18, 2026, directed at America's six Regional Transmission Organizations (RTOs) and Independent System Operators (ISOs), commanding all parties to prove the rationality of their current rules concerning ultra-large loads and Behind-the-Meter self-generation, or face comprehensive restructuring. [17]

Following an August postponement, final formal responses from each region concluded by mid-November. This means: reliability standards that an originally islanded campus must meet to later interconnect and sell power to the grid are currently being hastily drafted across major grid regions, with nationwide conclusions still pending. Avanza previously provided an in-depth interpretation of the 60-day countdown window FERC left for the six grids, demonstrating why stringent environmental impact assessment permits—not mere FERC rule modifications—remain the absolute bottleneck restricting Behind-the-Meter gas generation.

Future Observation Focus

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

First, the first officially named customer and its contract structure. If a single general contracting agreement absorbing all responsibilities is ultimately signed, it means Schneider or its competitors have successfully captured the system integration layer, firmly locking in customer stickiness.

If clients continue signing three separate fragmented equipment and engineering contracts, it shows clients still play the role of ultimate system integrators. So-called ecological bundling is merely a dispatch convenience tool for expediting schedules, failing to drive genuine business model transformation.

Second, whether project financing arrives smoothly, and whose balance sheet ultimately shoulders this burden. Schneider has participated in syndicated financing multiple times, and it is worth savoring the roles of participating entities therein.

In September 2025, Schneider partnered with Microsoft and over twenty allies to launch the Accelerated Resilient Infrastructure Initiative, wherein six project developers secured $7.5 billion in dedicated funds for project delivery via Energy-as-a-Service contracts. Schneider explicitly positioned itself as a strategic integrator, responsible for matchmaking clients and developers while outputting technology, with one of the six developers anchoring this massive capital pool being none other than AlphaStruxure. [18]

Merely seven months later, Schneider's name was completely stripped from the parent company column of AlphaStruxure. Regardless of the equity change mechanisms behind it, its strategic direction remains unchanged: Schneider firmly secures its seat as an integrator, leaving the heavy responsibility of holding heavy assets for other giants' balance sheets to shoulder.

Massive external capital predators are accelerating their influx. In June 2026, Brookfield expanded its strategic cooperation with Bloom Energy on AI computing power electricity infrastructure by fivefold to $25 billion, exclusively designated for investing in and constructing on-site power supply infrastructure. [19] Data centers initially ventured into Behind-the-Meter self-supply simply because it was the fastest unique path to acquiring power; today, capital markets are crafting tailored astronomical fund pipelines for it, marking a distinct milestone toward this sub-sector's maturation.

Finally, whether bets placed on the electrical control layer can withstand the baptism of real-world trials. Schneider's accumulation of standardization in this field stretches far longer than implied by this Yotta conference announcement.

Back in 2023, it launched the EcoStruxure Microgrid Flex standardized microgrid solution, [20] continuously expanding its channels and financing allies throughout 2025.

In March 2026, it officially inaugurated a microgrid testing laboratory at its Andover global R&D center, with its Senior Vice President of Global Microgrids explicitly setting the goal as "building tested, verified, and documented reference architectures through more standardized methods to accelerate the deployment of highly robust microgrid control systems." [21] This demonstrates that a full half-year before the AI Energy Park launch, Schneider had already fully and openly commercialized its microgrid control architecture.

Figure: Public record timeline from 2019 to 2028, outlining key milestones for Schneider Electric and AlphaStruxure

Schneider has never directly linked the AI Energy Park to its existing microgrid business in public forums, given that specific product lines within a multinational giant often operate independently from the early departments that bred those capabilities. However, this clear chronological sequence holds immense reference value: it provides the strongest evidence that its comprehensive leap toward the control layer is a calculated, established strategy, absolutely not a speculative trend-following whim.

Schneider Electric Senior Vice President and CTO of Safety Power Business Jim Simonelli bluntly pierced the underlying objective reality constraints during the Nvidia GTC conference in April: "Three years ago, the industry's extreme challenge lay in how to exhaust heat from server racks; today, the true challenge is how to connect power into racks and route it to campuses." [22] Regarding schedulable on-site power meeting AI-scale demands, he was even more straightforward and candid: "Aside from natural gas, there are currently absolutely no viable alternative options."

Gas has already become a highly mature, undifferentiated commodity, while internal combustion engines are likewise priced-off-the-shelf industrial catalog products. The only segment within the entire chain capable of continuously generating compounding value through technological accumulation is strictly the system integration and electrical control layer.

And the supplier that ultimately facilitates real paying customers footing the bill for its integrated solutions in the future will formally unveil to the entire industry the true commercial ceiling and ultimate valuation of this critical control layer.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps