AI is running out of power. NVIDIA just told the world how it plans to fix that — and which companies it needs to do it.

Think of it like a house that suddenly needs way more electricity. You can't just plug a factory into a wall outlet.

Imagine your house runs fine on normal wiring. Now imagine you want to plug in something that uses as much power as an entire factory. The wires in your walls would melt. You would need to rebuild the whole electrical system — thicker wires, a bigger panel, a totally different setup. That is exactly the problem NVIDIA's customers are facing with AI. Each rack of AI chips used to need about as much power as a few houses. The new ones need as much as a small neighborhood. The old data center wiring simply cannot deliver that much electricity safely or efficiently. So NVIDIA designed a new system called 800 VDC. The companies that build the parts for this new system are about to get a wave of orders — and that is the investment opportunity.

The old way had too many middle-men for the electricity. The new way cuts most of them out.

Picture electricity as water flowing from a city reservoir to your kitchen tap. In the old data center design, the water passed through about seven different pumps and valves before reaching the chips — and each one leaked a little. NVIDIA's new 800 VDC design removes most of those pumps and sends the power almost straight to the chips at a much higher "pressure" (voltage). Higher pressure means you can push the same amount of power through much thinner pipes — which saves a huge amount of copper, space, and money.

What Gets Thrown Out

The old design used a few big, expensive pieces of equipment that the new design simply doesn't need anymore. Companies that only make these old parts are in trouble:

• The AC battery-backup units (called "AC UPS") — giant boxes that kept power steady. Gone in the new design.
• The old power distribution cabinets (called "PDUs") — the boxes that split power to different parts. Gone.
• The thick low-voltage wiring — replaced by much thinner high-voltage wiring.

What Gets Bought Instead (The New Shopping List)

The new design needs new parts — and this is where the money flows:

• Solid-state transformers — a smarter, smaller kind of transformer that does the job of the old bulky one. Brand-new market.
• High-voltage power chips — special chips (made of a material called silicon carbide) that can handle 800 volts without melting.
• Batteries built right into the system — in the old design batteries were optional. In the new design they are required, to smooth out AI's spiky power demand.
• New cables and connectors — designed for high-voltage DC power.

Why Batteries Suddenly Matter So Much

AI computers don't use power smoothly. They spike — drawing a flood of electricity one second and much less the next. That kind of jerky demand can damage the power grid. So the new design requires big batteries sitting right at the front of the data center to act like a shock absorber, smoothing out the spikes. This turns battery storage from a "nice to have" into a "must have" — which is great news for battery companies.

NVIDIA published the full list. Here are all 28 — and what each one actually does.

This is the heart of the report. NVIDIA put out a slide titled "Who We're Working With" listing the 28 companies it is partnering with to build the 800 VDC power system. When NVIDIA — the most important company in AI — tells you exactly which suppliers it depends on, you pay attention. Below is every single company on that slide, grouped by what they make, explained in plain language. Not all of them are easy to buy as a US investor (some are based in Asia or Europe), but knowing the full list tells you where the money is going to flow.

Group 1 — The Big Power Equipment Makers

These companies make the large electrical gear — transformers, switchgear, and the systems that move power around the building.

Group 2 — The Power-Chip Makers (Silicon Carbide & Gallium Nitride)

These make the special chips that can handle high voltage without overheating. This is the "secret sauce" that makes 800 VDC physically possible.

Group 3 — The Analog & Power-Management Chip Makers

These make the smaller "brain" chips that manage and regulate the power flow at every step. Many already sell heavily to NVIDIA.

Group 4 — The Builders, Cables & Manufacturing Partners

These companies physically assemble the power supplies, make the cables, and put the whole thing together.

The partner list is the obvious play. But some stocks already on our radar benefit just as much.

The 28 NVIDIA partners are the direct beneficiaries. But the 800 VDC shift creates a bigger wave that lifts other boats too — and several of those are names we have already researched and follow. These are the ones worth re-examining specifically because of this power story. Here is why each one connects.

The Power-Generation Plays

The Power-Chip Play

The Connectivity & Infrastructure Plays

$50-80 billion of new spending. And that's just the power gear, not the AI chips.

Let's put real numbers on this. Industry estimates say the world will spend somewhere between $50 billion and $80 billion building out this new 800 VDC power infrastructure between now and 2030. To be clear: that is money spent ONLY on the power equipment — transformers, chips, cables, batteries. It does not include the hundreds of billions being spent on the actual NVIDIA AI chips. This is a separate, additional pile of money flowing to the companies on the partner list.

Where The $50-80 Billion Goes

Why The Spending Is Basically Guaranteed

Here is the part that makes this different from a typical "hot stock" story. The companies building AI data centers — Microsoft, Google, Amazon, Meta, Oracle — are spending around $350 billion combined in 2026 on AI infrastructure. They have already committed the money. They have already ordered the NVIDIA chips. Those chips are useless if the building can't power them. So the power upgrade isn't optional — it's the thing standing between these companies and the AI capacity they've already paid for. When spending is forced like that, the suppliers tend to do very well.

One Real Risk On The Numbers

The timeline matters. The big shift to 800 VDC happens around 2027, when NVIDIA's next-generation chips (called Rubin) arrive. If those chips are delayed, the power spending gets pushed back too. So while the direction is very likely, the exact timing could slip by a year or more. That's the main thing to watch.

Being on NVIDIA's list is a huge head start. But these companies still compete with each other.

Getting named by NVIDIA is like a restaurant getting recommended by the most famous food critic in the world — suddenly everyone wants in. But the named companies still have to compete with each other for the actual orders. Here's how the main battles shake out, in simple terms.

The Equipment Makers: Vertiv vs Eaton vs ABB vs Schneider

These four are the giants fighting for the big power-equipment orders. Vertiv is the most focused on data centers specifically and works hand-in-hand with NVIDIA. Eaton and ABB are bigger overall but spread across many industries. Schneider and Siemens are the European competitors. The simple takeaway: Vertiv is the purest data-center bet, Eaton is the steady giant, and there's enough spending coming that several of them can win at once.

The Power-Chip Makers: Infineon vs onsemi vs STMicro vs Wolfspeed

These make the high-voltage chips. Infineon is the biggest and most profitable — the safe choice. onsemi and STMicro are profitable mid-size players. Wolfspeed is the small, risky, pure-play that just came out of bankruptcy. The simple takeaway: if you want safety, go with the profitable giants; if you want the high-risk, high-reward lottery ticket, that's Wolfspeed.

The Threat From China

Chinese companies are building similar power chips at lower cost, backed by their government. For now, the American and European companies have a technology lead — especially on the most advanced chips needed for 800 volts. But over time, cheap Chinese competition could eat into the simpler parts of the market. This is the main long-term risk to watch for the chip makers.

Why this could work. And what could go wrong.

Three ways to invest, from safest to riskiest. Pick what fits you.

You don't need to bet everything on one stock. The smartest way to play a big trend like this is usually a small basket. Here are three approaches, sorted from lowest risk to highest.

The Safe Way — Big, Profitable Companies

Buy the established giants that are profitable today and named by NVIDIA: Eaton (ETN) for power equipment, Infineon (IFNNY) or onsemi (ON) for power chips, and Analog Devices (ADI) or Texas Instruments (TXN) for management chips. These won't double overnight, but they make real money now and the AI power trend adds to their growth. Lower risk, steady upside.

The Balanced Way — The Pure-Play Leaders

Vertiv (VRT) is the cleanest pure data-center power bet and works directly with NVIDIA. Monolithic Power (MPWR) already sells power chips into NVIDIA's systems and is an under-the-radar winner. These have more upside than the giants but also more ups and downs. A core position in VRT plus MPWR captures most of the thesis.

The Risky Way — The Small, High-Reward Bets

This is where the names we already follow come in. Wolfspeed (WOLF) — the high-voltage chip pure-play, just out of bankruptcy, could double or could fall hard. HYLN — the on-site power generator with native 800V output. FLNC — batteries that just became mandatory. NVTS — the small gallium-nitride chip maker with an NVIDIA deal. Keep these small — under 1% each — because they can swing wildly. Size them as lottery tickets, not core holdings.

A Simple Sample Basket

Catalyst Calendar — What To Watch

Every name in one table. NVIDIA's partners plus the ones we already follow.

Here is the whole opportunity on one screen — the easy-to-buy NVIDIA partners at the top, then the names we already follow that benefit from the same wave. Use this as your quick-reference cheat sheet.

AI's biggest hidden problem is power. NVIDIA just handed us the list of companies fixing it.

Here is everything in one paragraph. AI chips have gotten so powerful that the buildings holding them can't deliver enough electricity. NVIDIA designed a brand-new power system called 800 VDC to solve this, and it published the exact list of 28 companies it's working with to build it. That means $50-80 billion of new spending over five years, on top of the money already going to AI chips — and the spending is basically guaranteed because big tech has already bought the chips and now needs to power them. The cleanest ways to play it are the named partners like Vertiv, Eaton, and the power-chip makers. The higher-risk, higher-reward angle comes from names we already follow — Wolfspeed, HYLN, and Fluence — that fit the story perfectly. This isn't a quick trade. It's a multi-year shift in how AI gets powered, and we now have the roadmap.