Hyperscale Power: Solid-State Transformers for AI Data Centers

A critical assessment of the EUR 5M seed building SSTs that operate at higher frequencies for AI data center power delivery — led by World Fund and Vsquared Ventures.

ProofStory Research March 11, 2026

EUR 5M Seed (~$5.4M USD) — March 10, 2026

Building solid-state transformers (SSTs) at higher frequencies than conventional iron-core transformers. Founded by Daniel Rothmund (CEO, ETH Zurich PhD — 99.1% efficient SST design) and Sami Pettersson.

EUR 5M
Seed Funding
99.1%
Efficiency Claim
100kW+
Nvidia Rack Power
$330M+
Competitor Funding

Three Core Questions

01

Is the Tech Real?

ETH Zurich PhD with 99.1% efficient SST design. Higher frequency = smaller form factor. Technically credible but pre-prototype. Academic bench-scale ≠ commercial deployment.

02

What About Competitors?

EUR 5M vs. $200M+ in competitor capital. Heron Power ($140M, ex-Tesla VP + a16z), DG Matrix ($60M, ABB-backed), Amperesand (Temasek). Not a funding gap — a different category.

03

What’s the Realistic Path?

Pre-prototype to commercial deployment: 3–5 years minimum. 18–36 month qualification cycles per customer. European HQ in US-dominated market. EUR 5M cannot simultaneously fund prototype + US market development.

Key Finding: The AI data center power problem is real and urgent. Hyperscale Power’s approach is technically credible with ETH Zurich PhD validation. However, EUR 5M pre-prototype against competitors with $60M–$140M who are further along creates exceedingly difficult competitive position. This is an early technical bet, not a competitive position.

The Numbers

Founded
2025, Zurich, Switzerland
Founders
Daniel Rothmund (CEO, ETH Zurich PhD) & Sami Pettersson
Funding
EUR 5M (~$5.4M) seed led by World Fund & Vsquared Ventures
Technology
SSTs at “tens of kilohertz” — higher frequency enables smaller form factor
Differentiation
Higher frequency = smaller size; PhD-origin 99.1% efficiency; purpose-built for AI data center rack power
Stage
Pre-revenue, pre-prototype. Seed funds first prototype.
Target Market
AI data centers — Nvidia racks at 100kW+ (current) to 1MW (roadmap)
HQ
Zurich, Switzerland (ETH Zurich spinout)

Why SSTs Matter Now

Nvidia GB200 NVL72 racks consume ~120kW. Roadmap projects 1MW. At these densities, iron-core transformers are physically larger than the racks they power.

SST Advantages

01

Smaller Form Factor

10kHz SST theoretically 200x smaller than 50Hz iron-core of equivalent power. Critical when floor space is at a premium.

02

99.1% Efficiency

Maintains efficiency across load ranges via dynamic semiconductor control. Iron-core drops at partial load.

03

Dynamic Control

Adjusts output in microseconds; iron-core cannot. Critical for variable AI workloads with rapid power demand changes.

04

Single-Unit Power Quality

Reactive compensation + harmonic filtering in one unit vs. separate equipment currently required.

Technical foundation is credible. But gap between ETH Zurich PhD and commercially deployed data center component is substantial. Academic efficiency at bench-scale ≠ commercial efficiency at data-center scale. Prototype build is the first real test.

Competitive Capital Stack

Heron Power: $140M raised (Drew Baglino, ex-Tesla VP + a16z). Operational power electronics expertise, not academic.

DG Matrix: $60M, ABB-backed. ABB = world’s largest transformer manufacturer. Manufacturing, distribution, and qualification advantages built in.

Amperesand: Temasek-backed. Direct data center operator access in Asia-Pacific.

Combined $200–330M vs. EUR 5M. Not a funding gap — a different category of company.

World Fund

Berlin climate tech VC — no US hyperscale operator relationships

Vsquared

Munich early-stage deep-tech fund — European focus

European HQ Risk

US operators dominate; procurement cycles headquartered in US

Qualification Timeline

UL listing, IEC standards, utility approvals = 18–36 months per customer

Prototype-to-Deploy

3–5 year minimum path to first commercial deployment

$280M Sector

SST sector fundraising context; Hyperscale Power = smallest entrant

Weaknesses & Threat Vectors

Five structural risks that define the challenge ahead.

High

Massively Underfunded

EUR 5M = 1/25th Heron Power, 1/12th DG Matrix. Capital determines deployment speed, prototype iteration margin, and ability to absorb failures.

High

Pre-Prototype Stage

Competitors almost certainly have early prototypes given capital and team backgrounds. Starting prototype clock in 2026 puts them years behind.

High

ABB Manufacturing Moat

DG Matrix has ABB’s manufacturing, distribution, and regulatory fast-tracking. Hyperscale Power must build all three from scratch.

Medium

Academic-to-Commercial Risk

99.1% efficiency at bench-scale ≠ across operating temperatures, humidity, partial loads, transient response, vibration tolerance.

Medium

US Market Underfunded

EUR 5M cannot fund prototype + US market development. Without US presence, risks being European niche player in a US-dominated market.

Assessment Matrix

Hyperscale Power is a technically credible but significantly underfunded entry into a rapidly consolidating sector.

Technical Foundation
High
ETH Zurich PhD, 99.1% efficiency, specific frequency differentiation
Market Opportunity
High
AI data center power crisis is structurally real and accelerating
Competitive Position
Low
EUR 5M vs. $200M+ competitor stack; pre-prototype; no manufacturing partner
Capital Risk
Low
1/12 to 1/25 of competitor funding; hardware iteration margin minimal
Geographic Risk
High
European HQ in US-dominated market; investors lack hyperscale-operator network
Execution Stage
Pre-prototype; 3–5 year estimated path to commercial deployment
Investor Signal
Medium
Credible European deep-tech investors but lack hyperscale-operator network
Investor Thesis
Technical differentiation via higher-frequency SST; prototype validation before market consolidation

Hyperscale Power is a technically credible but significantly underfunded entry into a rapidly consolidating sector. Relevant as an acquisition candidate if prototype succeeds — the higher-frequency SST approach is a differentiated technical asset larger operators might want to acquire. Not a standalone competitive threat at current funding. The EUR 5M seed is a founder’s bet on prototype success before the window closes.

The AI data center power problem is real and urgent. Hyperscale Power’s approach is technically credible. But EUR 5M pre-prototype against $200M+ in competitor capital creates an exceedingly difficult competitive position.

Research Sources

Based on publicly available information, including the funding announcement of March 10, 2026.

  1. TechCrunch — “Hyperscale Power raises EUR 5M seed for solid-state transformers” (March 10, 2026)
  2. ETH Zurich research documentation — Rothmund PhD, 99.1% efficiency, higher-frequency design
  3. World Fund — lead investor, Berlin climate tech VC
  4. Vsquared Ventures — co-lead, Munich deep-tech fund
  5. Heron Power — $140M raise, Drew Baglino (ex-Tesla VP), a16z
  6. DG Matrix — $60M, ABB backing
  7. Amperesand — Temasek backing
  8. Nvidia data center roadmap — GB200 NVL72 (120kW), 1MW projections
  9. Industry analysis — iron-core transformer size vs. AI rack density
  10. SST sector funding analysis — $280M sector total