Sunday Special Jones Report
______
Introduction:
It is where NVTS steps in with two key technologies: Silicon Carbide (SiC) for the high-voltage front end and Gallium Nitride (GaN) for the fast rack-level converters. Together, they form the backbone of a new revolution in power delivery -- one that will save Billions each year in wasted energy.
Why SiC and GaN Matter
Traditional silicon power devices are hitting their limits. SiC and GaN push far beyond, both switching faster and running cooler.
SiC handles very high voltages and current — perfect for converting grid-level AC into the 800-volt DC that feeds the building.
GaN excels at ultra-fast switching at lower voltages — ideal for converting that 800-volt DC bus down to 48 volts at the server racks.
The result: smaller equipment, less heat, and much higher overall efficiency.
It’s the high-speed switching itself — turning current on and off thousands of times per second — that performs the real work of converting AC into smooth, usable DC power.
Figure 1: Power Flow Using NVTS SiC + GaN Architecture
At the front end of every AI data center, incoming 13.8-kilovolt AC power from the grid is stepped down and rectified into DC.
This is where Silicon Carbide (SiC) devices take charge.
They handle high voltage and high current, switching cleanly and fast enough to convert that grid power into a stable 800-volt DC bus that feeds the rest of the system.
The rapid switching feeds and shapes the current through the rectifier and filter stages, turning raw AC into the steady DC bus that powers everything downstream.
Each 1-megawatt power block uses about forty SiC devices, and a 300-megawatt campus will need roughly twelve thousand of them.
The result is less heat, smaller hardware, and higher overall efficiency -- the essential first step in making AI’s enormous power demand sustainable.
How it works:
Incoming 13.8 kV AC passes through protection breakers and a step-down transformer.
From there, rows of SiC power switches rapidly turn current on and off to perform the AC-to-DC conversion.
These high-speed switches — known technically as MOSFETs (metal-oxide semiconductor field-effect transistors) — are the key to high-efficiency power transfer.
DC link capacitors and bus bars stabilize the current, while a liquid-cooling loop keeps the system running efficiently under heavy load.
Once power reaches the data hall, Gallium Nitride (GaN) devices take over.
Each rack’s converters step the 800-volt DC bus down to a regulated 48-volt DC supply that feeds the GPU and CPU power modules.
GaN switches operate with less loss and much higher switching speed than legacy silicon.
A single 80-kilowatt rack uses around forty to fifty GaN devices, which means roughly 180,000 NVTS GaN chips across a 300-megawatt campus.
Higher frequency means smaller transformers, less copper, and far less heat — all translating to lower power bills and lighter cooling loads.
48 volts DC is the sweet spot — high enough for real efficiency, low enough to stay safely serviceable. It’s classed as SELV (Safety Extra-Low Voltage) under IEC standards, meaning technicians can work around it without full high-voltage PPE. Compared to today’s 12-volt rack systems, the shift to 48 V slashes distribution losses and heat — a necessary upgrade as AI loads grow.
NVTS SiC devices handle the heavy lift at the 800-volt front end, while NVTS GaN chips drive the high-frequency 800-to-48-volt step at each rack. Together they form the silicon backbone of this quiet power revolution.
Across that same 300-megawatt AI campus:
About 12,000 SiC devices at the front end and 180,000 GaN devices at the racks. That’s nearly 200,000 NVTS power semiconductors per site, with each one quietly working to turn waste into efficiency.
Multiply that by the wave of new AI campuses on the planning boards, and the numbers rise into the tens of millions — and billions of dollars in energy savings every year.
Every watt that feeds the AI revolution must first pass through semiconductors like these.
And as thousands of new megawatts come online, so will the demand for hundreds of thousands of NVTS power chips per site.
This is the quiet foundation under the entire AI buildout — silicon that doesn’t compute, but makes the computing possible.
When you look at that scale, it’s clear where this story leads.
If NVIDIA designs the brains of AI, then Navitas builds the power behind it — the Navitas of Power.
With so much at stake, and NVIDIA already the one to recognize what’s unfolding, the question isn’t if the market wakes up — it’s when.
__
______________
More later so ....Stay tuned, if you dare!
For now, we close by noting that any view on the market and stocks on any particular day may change in the days to come. That is why we watch and see how our views match up with reality. Looking ahead a few months may be a way to do things - but thinking too deeply about world events and the recent alliances forming, can make projecting ahead a dicey endeavor.
All in all - we use the word maybe "some", not "too much" and play it accordingly. Never get arrogant in our notions because things do change - and individual stocks are subject to many factors outside our control. So, we try to -stay aware.
With all the above caveats and attempted prognostications, I will close this post. Stay tuned for more opining on the market and stocks to watch.
ALL in my humble opinion, scroll down and read more.This site does NOT make Buy / Sell recommendations.
________
__
______________
More later so ....Stay tuned, if you dare!
For now, we close by noting that any view on the market and stocks on any particular day may change in the days to come. That is why we watch and see how our views match up with reality. Looking ahead a few months may be a way to do things - but thinking too deeply about world events and the recent alliances forming, can make projecting ahead a dicey endeavor.
All in all - we use the word maybe "some", not "too much" and play it accordingly. Never get arrogant in our notions because things do change - and individual stocks are subject to many factors outside our control. So, we try to -stay aware.
With all the above caveats and attempted prognostications, I will close this post. Stay tuned for more opining on the market and stocks to watch.
________
