Behind the 90 GW Announcement Across 59 U.S. Data Centers: What Critical Layer Are Stakeholders Competing For?

Author: Christopher Johnson (Avanza Energy)

Compiled by: TechFlow

TechFlow Editor's Note:In the current race by compute centers to secure power, behind-the-meter self-generation is evolving from a former emergency backup option into the sole lifeline for protecting delivery schedules. The 90 GW of off-grid self-generation plant plans already announced in the United States marks a foundational restructuring of compute infrastructure away from traditional public utility grids.

However, as generator sets and gas supply gradually become commoditized, standardized off-the-shelf products, the core competitive focus in the value chain has quietly shifted upstream. Major electrical giants and energy solution providers are no longer competing solely on hardware; they are competing for the system integration and electrical control layer spanning from the generation side to compute chips.

This article provides an in-depth analysis of the latest strategic alliances and commercial pullbacks by giants such as Schneider Electric and Caterpillar, revealing the ultimate differentiator as onsite power shifts from engineering packages to commercial execution: whoever controls the control logic and single-line diagram architecture can build the deepest commercial moat over a twenty-year compute lifecycle.

Cleanview's latest tracking data shows that 59 U.S. data center campuses have cumulatively announced approximately 90 gigawatts (GW) of behind-the-meter self-generation capacity, with about 82 GW announced in a concentrated wave from the start of 2025 to date.

However, only about 2 GW of behind-the-meter capacity is currently in commercial operation. [1] This huge gap highlights how new the demand surge is, rather than any instability in the project pipeline.

Before 2025, onsite self-generated power was viewed only as an emergency backup option; today, it has become a mandatory ticket for data centers to energize and commission on schedule. Schedule has become the hardest constraint: while the North American data center construction backlog hits record levels, the traditional utility interconnection queue pipeline is being completely exhausted.

Energy research firm Enverus has provided a clear capacity forecast: between 2026 and 2030, 25.5 GW of new industrial power demand in the United States will be met through behind-the-meter self-generation, with data centers accounting for 22.5 GW, or about 36% of all new U.S. data center capacity during that period. [2]

Supporting this massive load requires 31.6 GW of new onsite generation equipment, of which gas-fired generation accounts for 29.6 GW. Reciprocating internal combustion generators account for 7.1 GW, small and mid-size gas turbines for 6.2 GW, and fuel cells for 4.8 GW.

Chart: 59 U.S. data centers have announced about 90 GW of behind-the-meter capacity; Enverus forecasts behind-the-meter load will reach 22.5 GW by 2030 (Source: Cleanview and Enverus Intelligence Research)

These numbers describe piles of hardware, but they do not answer: when an entire power plant fails to start on time, who should the customer hold accountable?

Suppliers' series of announcements at the Yotta 2026 conference are centered on this core question, which is also the industry's most compelling dynamic: the technology pathways for generation equipment have largely been settled, but the commercial delivery framework remains full of uncertainty.

Avanza has previously systematically mapped the scale of onsite power expansion, the typical campuses behind it, and the five technology pathways driving its development. What has not yet been clearly defined is the critical control layer built on top of it: who is responsible for integrating all components? Who will control the core customer relationships over the next twenty years? And who is actively choosing to exit this layer?

Announced Partnerships

On September 30, Wärtsilä, Schneider Electric, and Stanley Consultants jointly launched an end-to-end solution for U.S. data center power, calling it 'Generator-to-Chip'. [3]

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

Schneider packaged this system into a concept called "AI Energy Park." Melton Chang, Executive Vice President of Schneider Electric’s Power Systems business, defined it precisely: "An AI Energy Park starts from compute load and integrates generation, electrical infrastructure, automation, and digital technologies into one organic system."[3]

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

Read the verbs closely: coordinate, simplify, reduce handover. In essence, this is a concerted-action agreement among three independent companies to appear together as an alliance. This is not a single-contract Turnkey Wrap project, nor does the announcement itself make such a commitment.

Key Elements Not Disclosed

The announcement contains no contract structure, no megawatt capacity metric, no specific customer names or pilot campuses, and no pricing. In its independent coverage of the collaboration, StorageReview noted that the release "provided no numbers, customers, or timelines."[4]

To date, Schneider’s public disclosures cover only the engineering and technology layer: the generator-to-chip division of labor, the keynote at the Yotta Conference, and a switchgear white paper.

The actual commercial terms—who signs the contract with the end customer, and who leads the long-term commercial relationship—remain entirely absent from public records. If this were a formally contracted real customer, the company would normally be announced or at least its installed capacity disclosed; this time, there has been no such move.

The most logical interpretation is that the entire industry, including Schneider, is turning onsite power into an engineered product far faster than it is turning it into a business model, and the latter is often far more difficult than the former.

The Platform That Once Addressed These Questions No Longer Exists

For seven years, Schneider owned a dedicated entity designed to answer all of the commercial questions above.

AlphaStruxure was founded in 2019 as a joint venture between Schneider and the Carlyle Group. It designed, financed, built, owned, operated, and maintained onsite energy infrastructure, while providing fixed power prices and performance guarantees backed by enforceable commitments.

It took on the 11.4 MW microgrid project for the new Terminal One at John F. Kennedy International Airport (JFK) in New York, built in phases through 2029;[5] and in April 2026 completed the largest bus-depot microgrid in the United States for Montgomery County, Maryland: 4.8 MW of solar PV, 2 MW of energy storage, and permanent islanding capability during grid outages.[6]

However, Schneider no longer describes itself as a parent company. As recently as April 19, 2026, AlphaStruxure’s official homepage still stated, "Our parent companies are Carlyle and Schneider Electric."

But by May 8, the same paragraph had quietly been changed to the singular "Our parent company is Carlyle," with Schneider’s name removed entirely. The surrounding text remained unchanged word for word, indicating a highly precise, targeted edit, and that wording remains on its website today.[7]

This website copy change was not accompanied by any official deal announcement—no press release, SEC filing, or trade-media coverage. Carlyle’s periodic financial reports also cannot confirm this, because asset managers’ 10-K filings disclose only management fees and AUM, not individual portfolio-company holdings within their funds.

But the indisputable objective fact is that Schneider no longer presents itself as a parent company and has stepped away from the vehicle that once directly owned assets and underwrote commercial mandates.

The "Energy-as-a-Service" (EaaS) model itself is beyond doubt. "Zero-CapEx" onsite power is an active, well-capitalized competitive field: in May 2026, VoltaGrid raised $1 billion from Blackstone and Halliburton, surpassing the $10 billion valuation mark, with its strength derived from owning and operating power plants and selling electricity directly to customers.[8]

Yet, just as onsite self-generation became the hottest competitive battleground across the industry, Schneider deliberately stepped back from this asset-heavy investment layer.

Why This May Be a Wise Move

Viewed positively—and, in my view, entirely correctly—this is a strategic decision about which layer is truly worth owning over the long term.

Schneider Electric is firmly retaining the electrical control and automation architecture layer, and the company is putting real capital behind its logic: the control layer is the ultimate chokepoint of the entire chain.

The software-defined medium-voltage switchgear unveiled at Yotta is billed as a world first, achieving complete decoupling of system intelligence from physical hardware. Schneider Electric claims that, compared with traditional designs, the solution reduces control wiring by 87%, terminal blocks by 91%, and copper usage by 85%, improves order-to-manufacturing speed by up to 3x, and cuts on-site commissioning time by half. [9]

Pilot projects will run through 2027, with a full commercial rollout in 2028. Some parts of the technology stack will land even earlier: prefabricated power modules built on the Galaxy VXL UPS in 2 MW, 2.25 MW, and 2.5 MW configurations are already available in Europe and are committed to entering the North American market by the end of 2026. [10]

Whoever leads the development of the one-line electrical diagram (One-line Diagram) and control logic naturally becomes the default leader for equipment commissioning, operations and maintenance services, and system upgrades across the asset's twenty-year lifecycle.

Generator sets will eventually be replaced and retired, and server racks will be completely refreshed every three to five years; but the underlying electrical topology and relay protection schemes will become more entrenched over time, along with the ironclad commercial lock-in with system operators.

In addition, a pure control-system positioning can be cross-compatible and sold to every project in the market. Schneider Electric has precedent: in March 2024, it partnered with Mainspring Energy to bundle its proprietary control and integration system around third-party linear generators. [11]

This partnership with Wärtsilä is simply the same logic applied to a higher-power engine. The playbook can be seamlessly replicated at Caterpillar power plants, Bloom fuel cell units, or utility-financed units. Owning one energy campus lets you capture only the returns from that campus; controlling the system integration layer lets you take a share of every energy campus.

Strong financial performance fully validates the bet. Schneider Electric's first-half 2026 revenue reached EUR 21.2 billion, with 14.0% organic growth; second-quarter revenue was EUR 11.5 billion, with 16.5% organic growth. [12]

Orders in the pure data center segment recorded a triple-digit surge in the second quarter, while actual sales in the broader data center and network end markets maintained strong double-digit growth. The 'Systems' business model, which carries prefabricated and turnkey engineering offerings, accounted for 35% of second-quarter revenue and delivered impressive 28% organic growth.

Two reasons to hold off on judgment

First, the impressive switchgear metrics come entirely from Schneider Electric's own White Paper No. 303, with no independent third-party audit, no confirmation from key customers such as Equinix, and no pilot test data disclosed publicly. [13]

Until real reference customers step forward to endorse it, claims such as saving 87% of wiring, 91% of terminal blocks, and 85% of copper, as well as several-fold speed gains, should be viewed cautiously as standard vendor marketing.

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

Technology outlet The Register raised sharper questions that Schneider Electric's public materials have so far avoided: 'Executing an erroneous or malicious software update on a critical electrical system affecting personnel and equipment safety has catastrophic consequences entirely unlike a botched desktop patch on an office computer.' [13]

Migrating medium-voltage electrical protection and control logic to the software layer improves efficiency, but it is essentially a large operational risk swap.

A data center that physically disconnects itself from the public grid to bypass grid interconnection queues also loses its last baseline safety protection. At this point, the purely software-based protection scheme has become the only barrier between instantaneous load fluctuations and a total blackout across the entire campus.

Forces are converging on the same keyword.

The strong similarity in naming indicates a structural industry pattern, not merely a marketing gimmick by any single company.

As early as June 2025, Eaton and Siemens Energy joined forces to bundle standard 500 MW gas turbine units with energy storage systems and quantified the commercial return as follows: by bringing commercial operation forward by up to two years, a typical 500 MW facility could generate an additional $2 billion to $3 billion in revenue. [14]

Vertiv and Caterpillar followed soon after, forming an alliance in November 2025; Vertiv CEO Gio Albertazzi explicitly defined it as a “Grid-to-Chip” framework. [15]

Schneider and Wärtsilä then entered in September 2026, flying the “Generator-to-Chip” banner.

Within just 15 months, three pairings of industry giants formed closely spaced alliances; two of them even adopted, in a remarkable convergence of views, nearly identical “X-to-Chip” narratives. The industry has reached a consensus: the smallest unit of competition is no longer a single engine or switchgear, but the complete pathway from source fuel to end-user compute; the only unresolved question is who will lead the sale of that pathway.

Figure: Six bundled integration models for on-site power and the division of roles among parties in the integration layer (Source: Avanza Energy analysis)

Meanwhile, two other business models sit entirely outside this bundling race and define the industry’s boundaries.

One is an investor-owned utility that directly and wholly owns a power plant, purchases generator sets directly from OEMs, with the OEM retaining no software layer, and sells power directly to data centers.

The other is a project developer that self-finances, builds, owns, and operates the plant, selling power directly to customers behind the meter; this is VoltaGrid’s flagship model and the path once taken by AlphaStruxure.

Beneath the five paths above lies the lowest-level competitive baseline for all bundled solutions: the customer signs three fragmented procurement contracts and assumes all integration work itself. This produces six distinctly different combinations for deploying the same megawatts, with the integration control layer falling to a different party in each model. Only in Schneider’s model is the integration layer firmly held by the software party that writes the control logic.

Regulatory developments have made this commercial issue more urgent than ever, advancing far beyond the surface suggested by 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, finding that PJM’s tariff lacked sufficient clarity and consistency regarding rates and interconnection terms for serving “Co-located Load” with new generating units, and imposed a regulatory definition.

PJM submitted its compliance filing on January 20, 2026, and FERC issued its formal order on April 16, 2026, approving it in part and rejecting it in part. [16]

The portions adopted and endorsed by FERC provide a direct and material benefit to energy parks: interconnection applicants serving “Co-located Load” may now request interconnection service based on actual load below the generator’s nameplate rating, and PJM must calculate system expansion and upgrade requirements based on the applicant’s actual requested capacity rather than the machine’s maximum theoretical output.

Customers can fully leverage existing accelerated provisions to bypass subsequent complex evaluation stages when grid impact is minimal, request temporary interconnection service to energize and bring the facility online before full completion, and claim excess-capacity service rights at existing points of interconnection.

It is precisely these four concrete mechanisms that turn so-called “grid flexibility” from marketing brochure language into legally effective tariff provisions; this is also the underlying logic for why a compute campus that today chooses islanded self-generation could, in the future, compliantly sell peak-shaving flexibility back to the bulk grid.

The portions rejected by FERC serve as a stark warning that the rules remain in rapid flux. FERC flatly rejected PJM’s attempt to substitute “point of interconnection” with “point of ownership transfer” in the “Co-located Load” definition and, on the grounds that it exceeded the scope of the proceeding, rejected PJM’s proposed changes to the behind-the-meter self-generation application process, directing it to take further steps to comply.

The deep-rooted contradictions have been fully exposed: PJM's own Independent Market Monitor (IMM) warned that the current proposal leaves completely unclear how the public grid would absorb the massive power surge from unexpected reverse injections by behind-the-meter generation when a large-scale same-site compute load suddenly and instantaneously collapses, or how the grid system would plan for backup supply if a behind-the-meter generator unexpectedly trips offline.

FERC temporarily classified these difficulties as beyond the scope of the present proceeding and issued no final ruling. Now these once-in-a-generation challenges have all been transferred to the very large load docket, and FERC has acted: on June 18, 2026, it simultaneously issued Section 206 show-cause orders to the six U.S. regional transmission organizations (RTOs) and independent system operators (ISOs), requiring each party to demonstrate the reasonableness of its existing rules covering very large loads and behind-the-meter generation, or otherwise comprehensively redesign them. [17]

After the August extension, the regions' final formal responses were due in mid-November. This means that the reliability standards a campus originally operating in isolation must meet if it later wants to interconnect with the grid and sell electricity are currently being hastily drafted within major grid regions, with no national consensus. Avanza previously provided an in-depth analysis of the 60-day countdown window FERC left to the six grids, and argued why stringent project environmental impact assessment permits, rather than FERC rule changes alone, are the most fundamental bottleneck stalling behind-the-meter gas generation.

Future Focus

Three decisive bellwethers will reveal the endgame of this business model faster than any official press release:

First, the first publicly named official customer and its contract structure. If the final deal is a single turnkey contract assuming full responsibility, it means Schneider or its competitors have successfully captured the system-integration layer and firmly locked in customer stickiness.

If the customer still signs three fragmented equipment and engineering contracts separately, it indicates that the customer remains the ultimate system integrator, and the so-called ecosystem bundle is merely a scheduling convenience tool for meeting compressed deadlines, not a true business-model transformation.

Second, whether project financing can be secured smoothly, and whose balance sheet ultimately bears this burden. Schneider has participated in syndicated financing on multiple occasions, and the roles of the participating parties are worth noting.

In September 2025, Schneider, Microsoft, and more than 20 partners launched the Accelerating Resilient Infrastructure Initiative, in which six project developers secured $7.5 billion in dedicated funding to deliver projects through Energy-as-a-Service contracts. Schneider explicitly positioned itself as the strategic integrator, matching customers with developers and providing technology, while one of the six developers carrying this massive funding was AlphaStruxure. [18]

Just seven months later, Schneider's name was completely removed from AlphaStruxure's parent-company listing. Regardless of the equity-change mechanism behind it, its strategic direction has remained consistent: Schneider is determined to hold the integrator's seat, while leaving the heavy responsibility of owning asset-intensive holdings to the balance sheets of other giants.

Huge external capital is accelerating into the sector. In June 2026, Brookfield expanded its strategic partnership with Bloom Energy on AI compute power infrastructure fivefold to $25 billion, specifically for investing in and building onsite power infrastructure. [19] Data centers initially entered behind-the-meter self-generation simply because it was the only fast path to securing power; now capital markets are creating massive dedicated funding pipelines for it, a clear milestone indicating that this subsector is maturing.

Finally, whether the bet on the electrical control layer can withstand the real-world battle test. Schneider's standardization foundation in this area is far longer than this Yotta conference announcement suggests.

As early as 2023, it launched the EcoStruxure Microgrid Flex standardized microgrid solution, [20] and continued to expand its channel and financing partners throughout 2025.

In March 2026, Schneider officially opened a microgrid test laboratory at its Andover global R&D center. Its Senior Vice President of Global Microgrids explicitly set the goal of “building tested, validated, and fully documented reference architectures through more standardized methods to accelerate the deployment of highly robust microgrid control systems.” [21] This indicates that a full six months before the launch of the AI energy campus, Schneider was already publicly commercializing its microgrid control architecture.

Figure: A public-record timeline from 2019 to 2028 mapping key milestones for Schneider Electric and AlphaStruxure

Schneider has consistently avoided directly linking the AI energy campus to its existing microgrid business in public, since specific product lines within a multinational giant are often independent of the early departments that incubated those capabilities. Yet this clear chronological progression is highly informative: it provides the strongest evidence that its comprehensive leap into the control layer was a deliberate, established strategy, not an impulsive speculative bandwagon.

Jim Simonelli, Senior Vice President and Chief Technology Officer of Schneider Electric's Secure Power business, cut through to the underlying physical constraints during NVIDIA's GTC conference in April: “Three years ago, the industry's limiting challenge was how to remove heat from the rack; today, the real challenge is how to bring power into the rack and to the campus.” [22] When discussing dispatchable on-site power to meet AI-scale demand, he was even more candid: “Apart from natural gas, there is simply no viable alternative today.”

Natural gas is already a highly mature, undifferentiated commodity, and internal combustion engines are off-the-shelf industrial catalog products with transparent pricing. Across the entire chain, the only elements that can continuously generate compounding value through accumulated technical expertise are system integration and the electrical control layer.

The supplier that first convinces real paying customers to purchase and deploy its integrated solution will formally reveal to the entire industry the true commercial ceiling and ultimate valuation of this critical control layer.

Source: 深潮 TechFlow