{"schema_version":"2026-09-05.topic-graph-v1","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors","topic":{"slug":"power-semiconductors","name":"The Power Chain","description":"A sourced reference collection on the semiconductors that carry electricity rather than information — why AI data centres are rebuilding their power architecture around 800 volts of direct current, why silicon carbide and gallium nitride are what make that possible, and why the problem that decides whether it works is servicing a live rack rather than saving a percentage point.","coverage_notes":"Twelfth packet, and the thinnest source base in the index so far, which is stated at the top rather than buried. Nine sources were consulted: five were retrieved and read, and four were surfaced and set aside. One read returned only navigation furniture to the automated fetcher and was opened in a browser instead. Two of the set-asides returned HTTP 403, one of them — the Open Compute Project's Solid State Transformer design specification — to a browser as well as to the fetcher, which is a hard block rather than a fetcher restriction and was not circumvented. That is the central limitation and it shapes everything else: an architecture designed by several companies through a standards body is described here through the announcements of individual participants, four of the five reads being suppliers or a customer describing their own parts. There is no independent measurement of anything in this collection. Accordingly, every efficiency ratio published by every participant — 98% per conversion stage, 97% and 98% product ratings, 99.3% power-factor-correction peak, up to 5% end-to-end, 25 to 40% loss reduction, 99% at 10 kV, 50% smaller cooling systems — is recorded as that company's claim and asserted nowhere, on the specific grounds that per-stage and end-to-end efficiency are different quantities, the stage counts differ between the architectures being compared, and no source states its baseline configuration. What is asserted is device physics, product specifications, stated architecture and stated intent. One source is a press release filed as an exhibit to a Form 8-K, which does not make its claims independent but does mean they were filed with a securities regulator. Two sources are shared with other collections by design — NVIDIA's 800 VDC engineering post, read for the copper topic, and Google's OCP EMEA post, read for liquid cooling — and their retrieval notes are carried verbatim so the register agrees with itself across the index. Named gaps, in order of value: the OCP Solid State Transformer specification, blocked; the Mt Diablo power specification, not retrieved, which is why the relationship between the plus/minus 400 V and 800 V approaches is described only in one participant's terms; independent trade reporting on the standardisation effort, blocked; and the other named suppliers in the ecosystem — onsemi, STMicroelectronics, Vicor, Microchip, Delta, Eaton — none of whose material was retrieved, so this page should not be read as a survey of who supplies what. Not yet editor-reviewed; every assertion reads as reported.","primer":"Electricity does not reach a processor in the form the grid delivers it. It steps down through a chain of conversions, each built from power semiconductors and each losing a percentage. That chain is being rebuilt, because the voltage it was designed around has run out.\n\nThe proposal is to convert medium-voltage alternating current — 13.8 kV, or 13.8 to 35 kV depending on whose account you read — directly to 800 volts of direct current at the data-centre perimeter, removing intermediate stages, and to distribute at that voltage to racks of a megawatt and beyond. Google describes the same shift in its own terms: from the 48 VDC it championed a decade ago to plus or minus 400 VDC supporting up to 1 MW per rack, developed with Meta and Microsoft under a project called Mt Diablo. The detail worth keeping is why that nominal voltage: to draw on the supply chain electric vehicles already established. The power architecture can be rebuilt this fast because somebody else paid to industrialise the parts.\n\nSilicon carbide and gallium nitride are doing the work, and the reason is switching speed rather than any headline efficiency claim. A device that switches faster needs smaller magnetics to move the same power, which is why a solid-state transformer can be a fraction of the size of the iron-and-copper unit it replaces. A silicon-carbide manufacturer puts the gap starkly: SiC devices operate above 10,000 hertz where conventional silicon IGBTs rated at 6500 volts are generally limited to a few hundred.\n\nEvery participant publishes an efficiency figure and none can be checked here — 98% per conversion stage, product ratings of 97 and 98%, power-factor correction peaking at 99.3%, an end-to-end projection of up to 5%. All are recorded as claims and asserted nowhere, for a specific reason: per-stage and end-to-end efficiency are different quantities, the number of stages differs between the architectures being compared, and no source read states its baseline. What can be said without one is structural. Removing a conversion stage removes its losses, and at a megawatt a rack, one percentage point is ten kilowatts.\n\nThe problem that actually decides whether 800 volts is deployable gets almost no coverage, and it is not efficiency. It is whether a technician can pull a board from a live rack. At 54 volts that is routine; at 800 VDC connecting or disconnecting on a live bus needs the board's capacitance charged and discharged under control, or the event is a fault. Infineon's stated contribution is a hot-swap controller on silicon-carbide JFET technology that lets boards be exchanged while the rest of the rack keeps running — and the economics are stated plainly alongside it: an AI server costs as much as thirty times a traditional server, so a rack powered down for service is a rack that stops earning.\n\nThe demand curve the components are being specified to is published by the people building them. Infineon puts AI rack power at around 120 kilowatts now, rising to 500, and to 1 megawatt by the end of the decade. Google projects more than 500 kW per rack before 2030. Navitas describes supporting 1 MW racks and beyond. Three companies with different interests landing on the same order of magnitude is meaningful corroboration — but they remain projections, and this entire architecture is being built in anticipation of a load rather than in response to a measured one.\n\nWhat is missing is named and it is substantial. The multi-vendor specification behind all of this returned HTTP 403 to both an automated fetcher and a browser and was not circumvented, so an architecture designed by several companies through a standards body is described here through the announcements of individual participants — four of the five sources read being suppliers or a customer describing their own parts.","primer_note":"Manufacturing.ai's own editorial synthesis, not evidence. Every factual statement in it is separately asserted and sourced in this response's concepts and events.","reviewed_through_date":null,"source_count":5,"inspected_source_count":5,"consulted_reference_count":6,"blocked_reference_count":3,"set_aside_reference_count":3},"figures":[{"kind":"time-series","title":"The rack the power chain is being built for","caption":"AI server rack power as the suppliers building the conversion equipment state it. These are the figures the components are specified against, and all three are projections.","sourceNote":"Infineon, October 2025: rack power estimated to increase from around 120 kilowatts to 500 kilowatts, and to 1 megawatt by the end of the decade. Corroborated in order of magnitude by Google's projection of more than 500 kW per IT rack before 2030 and by Navitas describing support for 1 MW racks and beyond. Nothing here is a measured load.","unit":"AI server rack power, kilowatts","points":[{"label":"Today","value":120,"display":"~120 kW"},{"label":"Next step","value":500,"display":"500 kW"},{"label":"End of the decade","value":1000,"display":"1 MW"}]}],"blocked_references":[{"title":"800 VDC Architecture for AI Data Centers","publisher":"NVIDIA","url":"https://www.nvidia.com/en-us/data-center/technologies/800-vdc-architecture/","source_type":"company_website","retrieval_status":"retrieved_no_content","content_inspected":false,"published_at":null},{"title":"OCP Pushes 800 VDC Toward an Open Standard for Megawatt AI Racks","publisher":"Converge Digest","url":"https://convergedigest.com/ocp-800-vdc-standard-megawatt-ai-racks/","source_type":"journalism","retrieval_status":"blocked_403","content_inspected":false,"published_at":null},{"title":"OCP Solid State Transformer (SST) Design Specification v0.3","publisher":"Open Compute Project","url":"https://www.opencompute.org/documents/ocp-sst-design-specification-v0-3-final-pdf","source_type":"other","retrieval_status":"blocked_403","content_inspected":false,"published_at":null}],"set_aside_references":[{"title":"OCP 800 VDC whitepaper and LVDC Solid-State Transformer Specification","publisher":"Open Compute Project","url":"https://www.opencompute.org/","source_type":"other","retrieval_status":"search_result_only","content_inspected":false},{"title":"OCP Mt Diablo power specification","publisher":"Open Compute Project","url":"https://www.opencompute.org/projects","source_type":"other","retrieval_status":"search_result_only","content_inspected":false},{"title":"Other power-semiconductor suppliers to the 800 VDC ecosystem","publisher":"Various","url":"https://www.onsemi.com/","source_type":"company_website","retrieval_status":"search_result_only","content_inspected":false}],"concepts":[{"slug":"ai-rack-power-density","name":"AI Rack Power Density","category":"component","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#ai-rack-power-density","assertions":[]},{"slug":"power-usage-effectiveness","name":"Power Usage Effectiveness (PUE)","category":"process","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#power-usage-effectiveness","assertions":[]},{"slug":"rack-power-distribution","name":"Rack Power Distribution","category":"component","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#rack-power-distribution","assertions":[]},{"slug":"serviceability-at-high-voltage","name":"Serviceability at 800 Volts","category":"component","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#serviceability-at-high-voltage","assertions":[{"id":"83334f83-e2b9-4871-ace7-015a1908c94a","predicate":"concept.description","statement":"Suppliers and operators converge on the same trajectory: Infineon puts AI rack power at around 120 kW rising to 500 kW and to 1 MW by the end of the decade, Google projects more than 500 kW per rack before 2030 with an architecture for up to 1 MW, and Navitas describes supporting 1 MW racks and beyond — all projections, not measurements.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"59891562-f21e-45cc-a2bf-0aa0c94b2aea","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Infineon advances leading-edge 800 Volt AI data center power architecture enabling better efficiency and serviceability","publisher":"Infineon Technologies AG","url":"https://www.infineon.com/press-release/2025/INFXX202510-003","source_type":"press_release","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-10-13","documents":[]}},{"id":"7ed5973a-bdd5-4c68-9222-2d400c39eb79","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Enabling 1 MW IT racks and liquid cooling at OCP EMEA Summit","publisher":"Google Cloud","url":"https://cloud.google.com/blog/topics/systems/enabling-1-mw-it-racks-and-liquid-cooling-at-ocp-emea-summit","source_type":"company_website","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-04-29","documents":[]}},{"id":"bbf78f06-6280-469f-9ab3-9d45feabb642","stance":"supports","is_primary_for_assertion":false,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Navitas Semiconductor Form 8-K Exhibit 99.1: collaboration with NVIDIA on 800 V HVDC architecture","publisher":"U.S. Securities and Exchange Commission (EDGAR)","url":"https://www.sec.gov/Archives/edgar/data/1821769/000162828025027705/ex9912025-05x21prrenvidiac.htm","source_type":"regulatory_filing","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-05-21","documents":[]}}]},{"id":"85113b1a-ccb2-4d77-99e8-9dbe01fcc709","predicate":"concept.description","statement":"The binding problem at 800 VDC is serviceability rather than efficiency: exchanging a board on a live bus needs controlled pre-charge and discharge, and Infineon's stated contribution is a silicon-carbide JFET hot-swap controller allowing board replacement while the rest of the rack runs — significant because an AI server costs as much as thirty times a traditional one, so uptime dominates.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"61862435-1ad3-4027-8d1d-0cad56dcdc68","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Infineon advances leading-edge 800 Volt AI data center power architecture enabling better efficiency and serviceability","publisher":"Infineon Technologies AG","url":"https://www.infineon.com/press-release/2025/INFXX202510-003","source_type":"press_release","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-10-13","documents":[]}}]}]},{"slug":"power-conversion-chain","name":"The Power Conversion Chain","category":"component","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#power-conversion-chain","assertions":[{"id":"307fb87a-9498-489d-88d7-9f4506998817","predicate":"concept.description","statement":"The power chain is being rebuilt to convert medium-voltage AC directly to 800 VDC at the data-centre perimeter, removing intermediate stages and serving megawatt racks; Google frames the same move as 48 VDC to plus/minus 400 VDC with Meta and Microsoft under the Mt Diablo project, chosen at that voltage to reuse the electric-vehicle supply chain.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"473a7c68-d2cb-402d-acdb-225e782cf586","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Enabling 1 MW IT racks and liquid cooling at OCP EMEA Summit","publisher":"Google Cloud","url":"https://cloud.google.com/blog/topics/systems/enabling-1-mw-it-racks-and-liquid-cooling-at-ocp-emea-summit","source_type":"company_website","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-04-29","documents":[]}},{"id":"58c059b5-f245-4763-b92a-59eeecd577e8","stance":"supports","is_primary_for_assertion":false,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Navitas Semiconductor Form 8-K Exhibit 99.1: collaboration with NVIDIA on 800 V HVDC architecture","publisher":"U.S. Securities and Exchange Commission (EDGAR)","url":"https://www.sec.gov/Archives/edgar/data/1821769/000162828025027705/ex9912025-05x21prrenvidiac.htm","source_type":"regulatory_filing","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-05-21","documents":[]}},{"id":"f0c488c4-473e-4e7a-af36-1aabc8dad845","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"NVIDIA 800 VDC Architecture Will Power the Next Generation of AI Factories","publisher":"NVIDIA","url":"https://developer.nvidia.com/blog/nvidia-800-v-hvdc-architecture-will-power-the-next-generation-of-ai-factories/","source_type":"company_website","retrieval_status":"fetched","content_inspected":true,"published_at":null,"documents":[]}}]},{"id":"40797cb1-d506-4995-ae11-f70b2c7fb222","predicate":"concept.description","statement":"Participants publish per-stage efficiencies as high as 98%, product ratings of 97-98% and power-factor-correction peaks of 99.3%, and an end-to-end projection of up to 5% — all recorded as claims and asserted nowhere, because per-stage and end-to-end figures are different quantities and no source states its baseline configuration.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"0a21a45a-24f9-47d5-b7a8-84622cdcb1a8","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Navitas Semiconductor Form 8-K Exhibit 99.1: collaboration with NVIDIA on 800 V HVDC architecture","publisher":"U.S. Securities and Exchange Commission (EDGAR)","url":"https://www.sec.gov/Archives/edgar/data/1821769/000162828025027705/ex9912025-05x21prrenvidiac.htm","source_type":"regulatory_filing","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-05-21","documents":[]}},{"id":"f24caf2a-75c3-4e82-a577-8e21c6dc4758","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Infineon advances leading-edge 800 Volt AI data center power architecture enabling better efficiency and serviceability","publisher":"Infineon Technologies AG","url":"https://www.infineon.com/press-release/2025/INFXX202510-003","source_type":"press_release","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-10-13","documents":[]}}]}]},{"slug":"wide-bandgap-power-semiconductors","name":"Wide-Bandgap Power Semiconductors","category":"component","canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#wide-bandgap-power-semiconductors","assertions":[{"id":"4b179941-6c36-4974-b84c-5998d9c0b9b6","predicate":"concept.description","statement":"Silicon carbide and gallium nitride matter for switching speed rather than headline efficiency: SiC devices switch above 10,000 Hz where 6500 V silicon IGBTs are generally limited to a few hundred hertz, and higher frequency shrinks the magnetics — which is what makes a compact solid-state transformer possible.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"25a38e91-0145-4e0c-877a-ec3c2034a31d","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Infineon advances leading-edge 800 Volt AI data center power architecture enabling better efficiency and serviceability","publisher":"Infineon Technologies AG","url":"https://www.infineon.com/press-release/2025/INFXX202510-003","source_type":"press_release","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-10-13","documents":[]}},{"id":"57f7c1b0-b4ae-41ed-ab5a-5b631944a616","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Powering AI with reliable silicon carbide-based solid-state transformers","publisher":"Wolfspeed","url":"https://www.wolfspeed.com/knowledge-center/article/powering-ai-with-reliable-silicon-carbide-based-solid-state-transformers/","source_type":"company_website","retrieval_status":"fetched","content_inspected":true,"published_at":null,"documents":[]}}]}]}],"events":[{"slug":"infineon-hot-swap-800vdc-2025","title":"Someone Solves the Live-Rack Problem","category":"technology_generation_milestone","event_date":"2025-10-13","date_precision":"day","country":"Germany","region":null,"canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#infineon-hot-swap-800vdc-2025","assertions":[{"id":"095211ac-1d83-4be4-a2e0-1c042191925a","predicate":"event.description","statement":"Infineon announced on 13 October 2025 a silicon-carbide JFET hot-swap controller allowing server boards to be exchanged on a live 800 VDC bus while the rest of the rack operates — addressing serviceability rather than efficiency, and framed against an AI server costing as much as thirty times a traditional one.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"21383b4f-3c71-4fe9-89fd-781c4fa5c040","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Infineon advances leading-edge 800 Volt AI data center power architecture enabling better efficiency and serviceability","publisher":"Infineon Technologies AG","url":"https://www.infineon.com/press-release/2025/INFXX202510-003","source_type":"press_release","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-10-13","documents":[]}}]}]},{"slug":"eight-hundred-volt-architecture-announced-2025","title":"The Rack Power Architecture Moves to 800 Volts","category":"technology_generation_milestone","event_date":"2025-05-21","date_precision":"day","country":"United States","region":null,"canonical_url":"https://www.manufacturing.ai/topics/power-semiconductors#eight-hundred-volt-architecture-announced-2025","assertions":[{"id":"d4638125-666c-480f-9b2d-881a4d7ed90a","predicate":"event.description","statement":"Navitas announced a collaboration with NVIDIA on an 800 V HVDC data-centre architecture in a release filed with the SEC on 21 May 2025: 13.8 kV AC converted directly to 800 V at the perimeter using solid-state transformers and rectifiers, supporting 1 MW racks and beyond for Kyber systems and Rubin Ultra accelerators.","assessment":"reported","editor_reviewed":false,"reference_only_source_count":0,"evidence":[{"id":"0d8c8dfb-9772-4f41-ae98-66714a5714b2","stance":"supports","is_primary_for_assertion":true,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"Navitas Semiconductor Form 8-K Exhibit 99.1: collaboration with NVIDIA on 800 V HVDC architecture","publisher":"U.S. Securities and Exchange Commission (EDGAR)","url":"https://www.sec.gov/Archives/edgar/data/1821769/000162828025027705/ex9912025-05x21prrenvidiac.htm","source_type":"regulatory_filing","retrieval_status":"fetched","content_inspected":true,"published_at":"2025-05-21","documents":[]}},{"id":"da09e9a8-743e-4cd0-8b01-0b4e4c9a56e5","stance":"supports","is_primary_for_assertion":false,"origin_independence":"unknown","note":null,"quote":null,"source":{"title":"NVIDIA 800 VDC Architecture Will Power the Next Generation of AI Factories","publisher":"NVIDIA","url":"https://developer.nvidia.com/blog/nvidia-800-v-hvdc-architecture-will-power-the-next-generation-of-ai-factories/","source_type":"company_website","retrieval_status":"fetched","content_inspected":true,"published_at":null,"documents":[]}}]}]}],"policy":{"publication_model":"transparency","evidence_entries_may_share_a_source":true,"linked_documents_do_not_add_verification":true,"correction_intake_available":false,"assessment_meanings":{"primary-source confirmed":"Editor-reviewed against a first-party source.","cross-source corroborated":"Editor-confirmed across two independent origins.","reported":"Supported by the attached sources; not yet editor-reviewed.","disputed":"Sources conflict; competing accounts are both attached.","superseded":"Replaced by a newer assertion.","unverified":"No supporting evidence attached."},"evidence_eligibility":"`evidence[]` contains only sources whose content was actually retrieved and read (`content_inspected: true`). Sources that could not be retrieved are counted per-assertion in `reference_only_source_count`, listed for the whole collection in `consulted_references`, and never presented as evidence. `source_count` counts only sources an assertion actually rests on — it deliberately excludes `consulted_references`, so it is never inflated by material nobody could read.","topic_scope":"Assertions and source counts are limited to the topic's current packet lineage. Reusing a concept in another topic does not silently import that other topic's assertions.","rights":"Statements are Manufacturing.ai's own prose. Source excerpts are not redistributed; `quote` is present only where explicitly cleared.","attribution":"Cite as Manufacturing.ai, https://www.manufacturing.ai/topics/power-semiconductors","correction_url":"https://www.manufacturing.ai/methodology#corrections"}}