Date: 26.05.2026
Topics: AI, Future telecoms
At Compound Semiconductor Applications (CSA) Catapult’s recent webinar – “Solid-State Transformers (SSTs): The Next Evolution in AI Data Centre Power Infrastructure, Power Grids and Electrification” – SSTs were shown to be not an incremental upgrade but something closer to a “Fosbury Flop” for power systems.
Conventional transformers remain highly efficient and reliable, but they are rooted in an architecture designed for a different era that had predictable loads, centralised generation and largely Alternating Current (AC) based distribution.
Now we need to get over a higher bar, and that needs new thinking.
AI-driven data centres, electrified transport, distributed renewables and storage are reshaping how power needs to flow, and it needs to be more dynamic, more flexible, and often in Direct Current (DC) native environments1.
SSTs respond to that shift by changing the architecture itself:
Like Fosbury’s technique, this isn’t just refining what exists, it’s not getting better at an old technique, it’s reframing how the whole system works.
Here’s where the analogy becomes richer.
Fosbury didn’t just invent a new technique and stop there. Athletes and coaches then spent decades refining it, they adjusted approach speed, body positioning, landing technique. The architectural shift created a new baseline, but performance still improved through marginal gains.
And that’s where a second sporting legend comes in; Sir Dave Brailsford.
Sir Dave Brailsford led the Sky Cycling Team and British Cycling to huge success by finding marginal gains. He focused on improving everything with incremental steps, one percent better nutrition, one percent better aerodynamics, one percent better recovery. Over time, those marginal gains built into a decisive advantage.
As discussed by the CSA Catapult experts in the webinar, SSTs enable a series of targeted improvements across the power chain:
Individually, some of these gains are modest, but together at AI Campus scale, with hundreds of kilowatts per rack the gains build to something remarkable.
This is the critical point for decision-makers:
SSTs are the architectural leap, and the marginal gains are how you extract the most value from that leap.
However, one of the most important messages, which is less visible, but equally critical is that the pace of innovation in power electronics is not just constrained by hardware.
It is constrained by confidence.
Part of the Brailsford strategy was the increase in accurate system level data in sport. Switching to American sporting success, and the story of the Oakland A’s 2002 season fictionalised in the film “Moneyball” highlights how a shift from instinct to data lays the groundwork for elusive marginal gains.
Development teams may have to rely conservative, previous product assumptions because they lack accurate, application-relevant data. Switching behaviour is complex, datasheets are static, and building in-house test capability is expensive and time-consuming.
The result is familiar across the sector:
This is the problem CSA Catapult is exploring with its new webinar on High-Fidelity Double Pulse Testing (DPT).
High-fidelity DPT is designed to give engineers rapid access to high-quality, real-world switching data, enabling better decisions earlier in the design process.
It provides:
What matters is not just the data, but what it enables.
With high-fidelity DPT, organisations can:
In environments like AI data centres, where efficiency, thermal performance and power density are tightly coupled, even small improvements, validated early, can have material impact at scale.
What emerges is not a single approach, but a unified one.
Together, they form a coherent strategy:
This is not about choosing between transformation and optimisation. It is about recognising that real progress requires both.
The shift underway in power systems is both significant and subtle.
It is not about waiting for one game-changing moment.
And it is not about incrementally improving legacy systems indefinitely.
It is about doing both at the same time.
As Dick Fosbury showed, sometimes you need to change the way you approach the bar entirely.
As Sir Dave Brailsford demonstrated, sometimes you win by improving everything else around it.
Architecture and optimisation, leap and marginal gain, working together.
For organisations willing to embrace that mindset, the gains may be incremental.
But the impact—like Fosbury’s—can redefine the entire field.
2.“The AI Data Centre Power Crisis: Why 1500 V DC Architectures Matter”
