SemiAnalysis Maps the 800VDC Datacenter Revolution Through Four Distinct Phases, Forecasts $24B Combined TAM by 2030
Deep dive published May 26, 2026 details equipment economics, supplier implications, and path from retrofit sidecars to solid-state transformers
SemiAnalysis has published a comprehensive technical and financial analysis of the 800VDC power transition underway across hyperscale datacenters, mapping adoption through four distinct architectural phases and sizing the combined market opportunity for power racks and solid-state transformers at approximately $24 billion by 2030. The report provides new detail on when specific equipment categories become obsolete, which suppliers are positioned to capture incremental content, and how facility-level electrical architectures will be rewritten as rack power densities approach 660 kilowatts with Nvidia's Kyber Ultra platform.
Phase One Begins Late 2026 as Google and Meta Lead Voluntary Future-Proofing
The first phase, already underway in pilot form, represents what SemiAnalysis calls the "white space retrofit" era. Google and Meta are leading, driven not by immediate necessity but by strategic positioning ahead of the forced transition. Current chip generations ramping in late 2026 and 2027, including Vera Rubin NVL72, top out at rack densities of 180 to 220 kilowatts, levels that three-phase AC can still deliver without hitting physical limits. Phase One is therefore voluntary future-proofing, and the architecture reflects that conservatism. New HVDC hardware layers on top of existing infrastructure rather than replacing it. The datacenter's electrical backbone stays intact, with the same transformers, UPS systems, switchgear, and automatic transfer switches.
The headline new equipment is a row-level cabinet called the HVDC power rack, a dedicated 42U enclosure that receives 415V AC from overhead busway and outputs 800VDC through cable to adjacent IT racks. Inside, it performs three jobs: rectification of AC to DC, battery backup unit modules for ride-through during outages, and optionally capacitor shelves for transient buffering during GPU load spikes. SemiAnalysis estimates the average selling price for these power racks at $400,000 to $500,000 per unit, roughly ten times the $40,000 ASP of standard AC power-rack equipment, or approximately $500,000 per megawatt of deployed capacity.
The concept originated with Microsoft's Mt Diablo project, and the OCP Diablo 400 specification, co-authored by Google, Meta, and Microsoft, has standardized it. But there is no one-size-fits-all design. Nvidia sits entirely outside the OCP framework and is developing a monopolar 800V reference design at 660 kilowatts, with air-cooled samples in mid-2026 and a liquid-cooled variant sampling late 2026. Within Diablo 400, the three co-authors diverge meaningfully on implementation details, with Meta running 600 to 800 kilowatts using 50 kilowatt HVDC output cables, Google pushing to 900 kilowatts by reallocating rack space from battery and supercapacitor slots to power supply units, and Amazon landing at 800 kilowatts on plus-minus 400V bipolar architecture.
Phase Two Arrives With 800VDC-Native Silicon, Inflection Expected 2027-2028
The real inflection comes with the arrival of 800VDC-native compute systems. At that point, 800VDC stops being a pilot and becomes a mandatory transition forced by physics and rack density. Operators electrifying the Kyber rack have no AC fallback at the rack inlet, and SemiAnalysis expects 800VDC penetration to spike sharply in this window. Because 800VDC-native silicon will land before facility-level 800VDC distribution is ready, the retrofit phase persists through Phase Two.
Architecturally, Phase Two looks very similar to Phase One. Both retrofit the white space with the HVDC power rack, both leave the grey space intact, and both rectify AC to DC in the row-level power rack. The key difference is where the voltage steps down to chip-usable levels. In Phase One, a power shelf inside the IT rack converts 800VDC to approximately 50VDC before it reaches the compute trays. In Phase Two, the 800VDC bus runs directly to the compute blade, and an on-blade power module handles the final step-down to 50V. Earlier Kyber designs shown at OCP depicted a DC-DC PSU sidecar adjacent to the compute rack, but SemiAnalysis now believes this approach is unlikely to be adopted at scale because a standalone sidecar consumes more aggregate floor and rack space than integrating the conversion stage into the blade itself.
Centralized UPS Systems Face Progressive Obsolescence
Traditional central UPS systems are probably the most contested piece of infrastructure in the 800VDC transition. In the new architecture, SemiAnalysis expects centralized low-voltage UPS systems to progressively lose their role and eventually become obsolete. In the retrofit era, the power rack sits directly on the 800VDC bus and houses BBU modules and supercapacitors. Both are natively DC-coupled. Battery backup units bridge seconds to minutes during outages, and supercapacitors absorb millisecond-scale GPU load transients. Together, they replace the centralized short-term battery storage and UPS ride-through function without the two to three percent conversion loss of an AC-DC-AC UPS pair.
Google and Meta already took this aggressive approach years ago, bypassing the central monolithic UPS with distributed architectures. In their implementation, AC power is distributed directly to the rack, the in-rack PSU handles AC-DC conversion, and rack-level lithium-ion battery backup units provide the short-duration bridge power. This removes the central UPS's AC-DC-AC conversion pair and improves efficiency, while also cutting in half the total battery capacity needed for the datacenter, since there is no longer a need for both an A-side and a B-side UPS.
That said, SemiAnalysis expects operators other than vertically integrated hyperscalers to keep the low-voltage UPS in place for redundancy and load fluctuation management, at least in the medium term. This is especially true for colocation providers, which prioritize flexibility and need to support mixed workloads including CPU racks, storage arrays, networking equipment, and older GPU racks that still run on AC.
Phase Three Rewrites Electrical Architecture With Centralized Rectification
Phase Three, expected late 2028 into 2029, changes the datacenter layout itself, and 800VDC becomes the building's electrical core. A dedicated upstream rectifier that sits in the grey space or outdoors converts 415V AC to 800VDC, distributing DC across the entire hall. These are mature units using silicon IGBTs or thyristors rated at 1200 to 1700 volts. The grey space splits in two. Medium-voltage transformers connecting the datacenter to the grid are unchanged. MV switchgear stays because the utility feed is still AC. LV transformers remain, stepping MV down to 415V AC for the upstream rectifier. But the 480V AC switchgear between LV transformers and PDUs has no role once 800VDC flows through the busway, and AC floor PDUs are eliminated along with it. In summary, everything above the AC-DC conversion point stays, while everything below it, designed for AC distribution, goes.
Once power is rectified, a DC busway replaces AC busway for hall-level 800VDC distribution. SemiAnalysis expects early 800VDC deployments to use feeder-only busway because tap-offs become more complex at 800VDC. Interrupting current under load creates a sustained arc that does not self-extinguish because DC has no zero-crossing point, while AC arcs naturally extinguish 100 to 120 times per second as the waveform crosses zero. DC-rated tap-off units with adequate arc interruption are physically larger, making them impractical today. Delta and ABB have publicly disclosed 800VDC busway programs.
To address arc interruption challenges, solid-state circuit breakers are now being adopted. SSCBs use silicon carbide or gallium nitride to interrupt fault current in microseconds. Because semiconductor switches can simply stop conducting with no physical contact separation, there is no arc to extinguish in the first place. ABB has the SACE Infinitus, a solid-state unit rated at 1000V and 2500A, with datacenter adaptation announced with Nvidia in October 2025. LS Electric has the first UL-certified DC molded case circuit breaker at 1500V, listed for datacenter applications.
In the white space, the power rack is replaced by what SemiAnalysis calls the battery rack. The battery rack shares most of the power rack's components and functions, but no longer performs AC-DC rectification because it receives 800VDC directly from the grey space. Three main components remain: DC distribution units that manage power distribution, switching, and monitoring across the 800VDC bus; BBU shelves that provide ride-through power during supply interruptions; and supercapacitors that absorb microsecond-to-millisecond transients that batteries are too slow to catch. SemiAnalysis expects content per megawatt for the battery rack to reach around $200,000 per megawatt.
Phase Four Brings Solid-State Transformers, the End-State Architecture
The final phase, expected beyond 2029, introduces the holy grail of DC power distribution: solid-state transformers. These are a new category of power electronic devices that replace conventional iron-core transformers with high-frequency, semiconductor-based converters. The SST replaces the LV AC-DC rectifier and low-voltage transformer with a single piece of equipment that converts directly from medium voltage to 800VDC. Phase Four and its datacenter layout is very similar to Phase Three, but the SST collapses additional conversion steps and brings new efficiency gains.
An SST does the same job as the massive iron-and-copper transformers in every datacenter's grey space: step voltage from utility-level medium voltage to a level IT equipment can use. A conventional transformer uses magnetic induction at grid frequency. An SST uses semiconductor switching stages to achieve the same conversion in a fraction of the volume. The datacenter SST is a three-stage device. The input stage converts AC to DC, handling medium voltage using silicon carbide MOSFETs rated at 3,300V or higher. The isolation stage uses a high-frequency transformer to step the voltage down while providing galvanic isolation. The output stage produces the final 800VDC that the distribution system needs, with no inverter required.
SSTs' core value proposition is energy efficiency, which translates directly to operating expense savings or unlocked compute capacity. By collapsing the medium-voltage transformer and rectifier into a single power-electronic stage, SSTs eliminate two conversion steps from the electrical chain. Vendors target up to fifteen percent total system efficiency improvement. SSTs are also dramatically smaller. A conventional transformer operates at 50 or 60 hertz and needs a massive iron core. An SST switches at 20,000 hertz or higher, shrinking the core by roughly ninety percent. That is where Infineon's claimed forty-times weight reduction and fourteen-times size reduction come from.
The best public SST benchmark comes from ETH Zurich: ninety-eight percent efficiency at 400 kilowatts in a 13.2 kilovolt AC-to-800VDC prototype presented at INTELEC 2025. Johann Kolar frames 98.0 to 98.5 percent as today's state of the art for full-scale SSTs, with ninety-nine percent as the next engineering target for datacenter units. Different vendors now converge on that 98.5 percent ceiling: DG Matrix's Interport platform claims up to 98.5 percent, Amperesand's third-generation system claims greater than 98.5 percent, and Heron Power's Heron Link targets 98.5 percent MV-to-rack efficiency.
DG Matrix, backed by ABB with an Infineon SiC supply deal, is shipping pre-certification units and targeting UL certification by end of second quarter 2026. It is the only SST included in Nvidia's MGX reference architecture. Amperesand targets 30 megawatts of commercial deployments in 2026. Heron Power is building a 40 gigawatt US manufacturing facility for its 4.2 megawatt Heron Link units. Novos Power claims a direct MV-to-800VDC SST with fifty percent smaller footprint and air cooling. Eaton acquired Resilient Power Systems in August 2025 for SST expertise. More than $320 million flowed into SST startups in the twelve months ending March 2026.
Combined TAM for Power Racks and SSTs Reaches $24 Billion by 2030
SemiAnalysis sizes the 800VDC equipment market by applying phase-by-phase adoption timelines to incremental datacenter capacity build with chip-by-chip SKU calculation. The firm expects sidecar power rack TAM to peak at approximately $11 billion in 2028 before declining as facility-level 800VDC takes share in Phase Three. The assumption is power rack content of $500,000 per megawatt. By 2030, SST TAM is expected to reach approximately $13 billion, capturing the demand displaced from the sidecar layer plus the incremental MV-to-800VDC conversion, based on content of $1.25 million per megawatt. A portion of this opportunity is contested by MV rectifiers, but SemiAnalysis expects SSTs to capture the majority share. Total incremental capacity powered by 800VDC is expected to reach approximately 39 gigawatts by 2030.
Total Electrical Content Remains Flat, But Mix Shifts Dramatically
Total electrical content per megawatt stays in a $3.6 to $4.8 million band across four of the five architectures SemiAnalysis models. The main headline is a content migration from grey space to white space, and the resulting change in equipment mix. Grey space content shrinks in Phase Two as the centralized UPS exits. White space peaks in Phase One because the HVDC power rack arrives. By Phase Four, total content climbs to $4.0 million as the SST replaces the LV transformer and rectifier. SemiAnalysis calculates the baseline AC power path at 82.0 percent cumulative efficiency across seven conversion stages. Phase One barely improves to an estimated 83.7 percent. The real jump comes in Phase Two at 86.5 percent when UPS elimination cuts the chain from seven stages to five. Phase Three pushes to 86.9 percent, and Phase Four reaches 87.4 percent. At one gigawatt of IT load, the Phase Two gain translates to roughly 58 megawatts of continuous grid power savings, with Phase Four extending that to 69 megawatts.
Regulatory and Safety Challenges Will Gate Adoption Speed
The National Electrical Code, published by NFPA on a three-year cycle, governs electrical installation in the United States. Full 800VDC code support targets NEC 2029. Pre-2029 deployments therefore require custom Authority Having Jurisdiction approvals and OEM-level UL certification for each site. This is workable for hyperscalers with in-house code engineering teams but could represent a meaningful barrier for colocation operators and smaller builders. SemiAnalysis thinks NEC 2029 will achieve partial provision, while full code maturity probably lands at NEC 2032 or 2035. Partial means the basic framework exists, but DC-specific arc flash personal protective equipment tables, busway standards, and stored energy maintenance protocols will likely be absent.
The biggest safety risk is arc flash. IEEE 1584 does not cover DC, and NFPA 70E has no PPE table for 600 to 1000 volts DC. UL Solutions has launched a Direct Current Safety Research Consortium to build the missing hazard models. At 800 volts, many rack-adjacent tasks that were routine at 48 volts likely require a qualified person under NFPA 70E, with arc-rated clothing, insulated gloves rated to 1000V, and a face shield. Capacitor banks and BBU modules retain dangerous charge after power-down, and standard lockout-tagout procedures for AC do not account for stored DC energy.
Cooling is the largest AC load in an 800VDC datacenter, and no vendor sells a DC-native cooling ecosystem. Danfoss's Turbocor compressors, dominant in datacenter chillers, run internally on DC at 700 to 813 volts. Delta unveiled a 2.4 megawatt in-row cooling distribution unit supporting 800VDC at GTC 2026, the first major cooling component engineered for native DC. But the full stack including chillers, compressors, pumps, and building controls remains AC-dependent. Beyond cooling, switchgear operating mechanisms, lighting, fire suppression pumps, building management sensors, and security systems all run on AC. Nvidia's reference architecture retains an AC auxiliary bus alongside the 800VDC compute distribution for exactly this reason.
Delta Electronics Emerges as Structural Winner With End-to-End Integration
SemiAnalysis identifies Delta Electronics as the structural winner in the 800VDC transition. Delta's core advantage is end-to-end integration: it can deliver a complete 800V solution across the power shelf, BBU, power conditioning system including supercapacitors, and liquid-cooling systems as one validated package. Power shelf ASPs jump from roughly $40,000 per rack in a standard AC-DC configuration to roughly $400,000 for an HVDC power rack, a ten-times increase driven by scope expansion. Delta's moat is vertical integration across the full power chain, spanning from grid to chip. No other player spans this full stack, and Delta is the only player that can credibly supply every major component from the utility interconnect to the voltage regulator on the GPU board.
SemiAnalysis expects Delta to be the main supplier for Nvidia, Meta, and Google as early adopters, with power racks set for volume shipments by end of 2026. In a scenario where a dedicated Kyber 800V-to-50V sidecar is eliminated, Delta could dominate ninety percent of this market because of its strong existing expertise in in-rack power supply units. The bear case for Delta is its limited presence in grey space. UPS and PDU have historically been dominated by Western incumbents such as Vertiv, Schneider Electric, Eaton, and ABB. Delta's UPS share in the Americas is minimal, and the bigger incremental grey-space opportunity, solid-state transformers, is further out.
Vertiv Positioned as Grey-Space Leader Pushing Into White Space
Vertiv is the top supplier of UPS for most major datacenter operators, and both AWS and Microsoft rely heavily on Vertiv infrastructure. That installed base should drive pull-through as operators upgrade power architectures to support higher rack densities. On 800V hyperscaler engagement, Vertiv is actively working with Meta, Google, and Microsoft. Vertiv won the Meta 800V HVDC power rack program alongside Delta, despite historically low content at Meta and Google. This win is a meaningful step-up in Vertiv content, from near-zero in whitespace to roughly $1 million per megawatt. Vertiv's existing UPS business is not cannibalized in the near term with the move to HVDC. In the whitespace retrofit use case, Vertiv gets additive content: the legacy UPS stays in place at approximately $1 million per megawatt grey space, and the new power rack at approximately $1 million per megawatt white space stacks on top.
The main limitation is that Vertiv does not participate in server-side white-space power electronics. It does not make PSUs, BBUs, or DC-DC converter modules for the IT rack, leaving it with effectively zero share of white-space power content in current GB200 racks. If the industry migrates toward power racks that blend grey-space and white-space functions, Vertiv may need to build, partner, or acquire rack-level conversion capabilities to avoid ceding the new profit pool.
Western Electrical Equipment Vendors Face Mixed Outlook
SemiAnalysis is somewhat neutral on the large electrical equipment vendors winning content in 800VDC. They sit closer to the assembly level, and near-term 800VDC demand is still concentrated in a narrower set of early deployments. The firm does not expect a broad-based step-function uplift for Western integrators in the immediate term. Schneider Electric looks structurally behind Delta and Vertiv in the 800V HVDC race. Schneider showcased its 800VDC sidecar during OCP 2025 capable of up to 1.2 megawatts per rack, and management noted that sidecar will ship well before Rubin Ultra's 2027 timeline. SemiAnalysis thinks this is largely dedicated towards the Oberon platform rather than Kyber. Separately, Schneider is the global leader in medium-voltage switchgear and distribution, a position that should remain secure through an 800V transition.
Eaton unveiled a reference architecture in October 2025 built in support of Nvidia's 800VDC architecture. The architecture features supercapacitor-based peak buffering using Eaton's XLHV modules. Eaton has little white-space server power business: no PSUs, no BBUs, no DC-DC converters. In current GB200 racks, Eaton captures zero white-space content. The 800VDC transition shifts spending from grey-space, where Eaton dominates, to white-space, where Eaton is absent. Eaton's acquisition of Resilient Power Systems brings real solid-state transformer IP in-house. If SSTs become the Phase Three standard for facility-level power delivery, Eaton is positioned ahead of peers with a head start on technology development.
ABB's datacenter business sits within its Electrification segment. The product portfolio is narrower than Eaton or Vertiv: LV and MV switchgear, breakers, power distribution, MV UPS, gensets, and prefab eHouse solutions. ABB no longer sells transformers; that business now sits with Hitachi Energy. ABB called out that its serviceable addressable market is approximately $2 million per megawatt, meaningfully below Eaton at $2.9 million per megawatt and Vertiv at $3 to $3.5 million per megawatt. On 800VDC, ABB was clear that current strong orders are for existing AC power architecture, and new 800-volt DC architecture with Nvidia is a post-2028 opportunity. ABB announced an Nvidia collaboration on 800V DC architecture in October 2025, but this is positioning, not revenue, and the partnership appears vaguer than Vertiv's direct Nvidia co-development on power racks.
Legrand Faces Material Displacement Risk Across Portfolio
Legrand is a major white-space equipment vendor across busbars, busway, rack PDUs, and IT racks. Datacenters represent about twenty-six percent of Legrand's fiscal 2025 revenue. Management argued on the second-half fiscal 2025 call that only PDUs and UPS, about twenty percent of datacenter segment revenue, face 800VDC displacement, and that any losses there will be offset elsewhere in the portfolio. SemiAnalysis thinks that materially understates the risk. By phases three and four of the transition, the firm estimates roughly fifty-five percent of datacenter revenue is exposed, including rack PDUs and busway, Legrand's highest-margin products. As architectures move toward sidecars and grid-to-chip DC, Legrand's legacy AC distribution risks being designed out and replaced by lower-value DC distribution downstream of a competitor's platform.
Management frames grid-to-chip as a post-2030 issue, but the direction is clear: once an SST converts AC to DC at the grid edge, power stays DC to the chip, eliminating the inefficient AC-DC-AC conversions in between. The problem for Legrand is that it does not yet appear to have a fully formed 800VDC product portfolio ready for that architecture and seems to trail behind competition. Legrand expects to ship DC busbar and busway products by end of 2026, ahead of ABB's 2027 timeline, though ABB is developing its DC busway directly with Nvidia while Legrand is not. Management has confirmed Legrand has no sidecar product, no development timeline, and no partnerships or acquisitions to close the gap. Unlike Delta, Vertiv, and Schneider, Legrand is not part of Nvidia's 800VDC ecosystem.
Panasonic Dominates Battery Backup With 80 Percent Market Share
Panasonic Energy claims approximately eighty percent market share in datacenter battery backup units as of fiscal 2025, having shipped over 600 million lithium-ion cells into datacenter applications without a critical safety incident. BBU development has progressed through OCP architectures from 33 kilowatts to 72 kilowatts, targeting 102 kilowatt next-generation shelves, with underlying cell output rising from 80 watts to a target exceeding 200 watts. Fiscal 2025 sales are expected in the upper 200 billion yen range, with a fiscal 2029 target of 800 billion yen at twenty percent-plus return on invested capital, with over eighty percent of fiscal 2029 sales already secured through customer design wins. BBU demand continues to be revised upward, expanding ahead of schedule even relative to December 2025 guidance.
Panasonic is developing two next-generation product categories directly tied to the 800VDC transition. First, capacitor backup units use a proprietary supercapacitor developed in-house. The CBU is designed to be form-factor compatible with existing BBU shelves, allowing operators to mix BBU and CBU modules for flexible customization of backup duration versus fluctuation absorption. Second, Panasonic is developing high-voltage BBUs specifically for 800V power racks. Panasonic continues to use NCA chemistry over LFP for datacenter, citing energy density requirements incompatible with LFP's lower volumetric density at the shelf sizes datacenter applications demand.
Musashi Seimitsu Holds Monopoly Position in Supercapacitors
Musashi Seimitsu, through its subsidiary Musashi Energy Solutions, holds essentially a monopoly in supercapacitors for datacenter applications. Historically a small-cap Japanese auto parts company, Musashi has been making a high-conviction pivot into AI datacenter energy storage. Musashi's HSC, or hybrid supercapacitor, is not a standard electric double layer capacitor. It is a hybrid device that combines EDLC positive electrode using activated carbon with a lithium-ion pre-doped negative electrode. It provides much higher capacitance and energy density than standard EDLC options that other players are building. Musashi has noted that HSC sales are likely approximately 10 billion yen, a low single-digit percentage of fiscal 2026 earnings. Musashi has already signed contracts with Flex in the US and Delta in Taiwan. At PowerGen 2026, Bloom Energy also highlighted the role of supercapacitors in its Bloom Energy Stamp architecture, indicating supercapacitor content in grey space as well as white space.