Menu

Case studies.

120+Use cases tested

Each case below is one complete analysis, from the question being decided to the advice it ends with. Validatus runs many such analyses, then compares all the cases as one portfolio, on their financials, futures and gates.

See how cases are comparedAll eleven decisions

Case study 01 · June 2026

Energy storage

NORVIA: battery storage for AI data centres

A utility-scale battery storage company, and EUR 150 million for AI data centres

An analysis by Validatus, a strategic evaluation platform for the decision economy.

The question

Should EUR 150 million go into battery storage for AI data centres, or stay with the utility‑scale business?

The grid wait

4+ years

for a grid connection in primary markets

Data centres now wait for power.

Loyal buyers

71 in 100

renewed their energy‑management provider

In a survey of 146 large manufacturers, most kept the provider they had.

Early adoption

25–30%

of data centres have them

Batteries on site are still early.

The plan against the world

13

places of disagreement

Outside evidence disagreed with the company's plan in 13 places.

The difference

37

of 100 for standing out from rivals

The extra price for the longer‑lasting product is not proven.

The deadline to learn

90 days

to sign a paid pilot

A paid pilot settles the first unknown.

What the analysis concluded

Aligned

Demand looks large: the biggest cloud companies need power for AI.

Uncertain

Whether renting storage costs operators little enough, over its life, to make them change.

Constrained

On these scores, the company does not look ready to sell yet. Its product, brand and experience score 33, 30 and 28 out of 100.

The first step

The first money would fund the cheapest test that shows one group of buyers will take part.

The aim

This is a search for a small market it can win. Each step is for learning, not revenue.

1 · The futures

Two unknowns decide which future comes. Together, the two moves below win in two futures, hold in one and lose in one.

How do buyers want to pay? And does the wait for grid power make them act now?

Neither question is answered yet; the 90-day pilot test answers both. A and B are the two moves below: each card says what each would do in that future, and the sign to watch.

They act nowThey wait and watch

They buy it as a service

01WinsLook here first

Buyers sign fast

Buyers short of grid power sign service contracts quickly, and demand turns into deals.

A
NORVIA offers one owner a paid service pilot, then releases the next part of the money.
B
NORVIA signs that first customer quickly, before a rival does.
Sign to watch
A pilot is signed inside 90 days.
02WinsLook here first

One anchor first

Buyers like the service but move slowly, so one reference customer decides everything.

A
NORVIA builds the total-cost case with the buyer's finance team first.
B
NORVIA puts most effort here, and keeps spending small until the pilot is live.
Sign to watch
The cost case is accepted, but the pilot start keeps moving.

They want to own it

03Holds

Buyers buy hardware

Buyers act now but buy the batteries themselves, mostly from named rivals.

A
In this future, the service pilot would lose its point, because buyers want to own.
B
Still win one first customer. It helps even when buyers own.
Sign to watch
Tenders ask to buy the asset.
04Loses

Buyers stay put

Buyers neither switch nor own. They keep buying power from the utility.

A
In this future NORVIA would spend on neither move, because neither wins here.
B
NORVIA would stop the search for a first customer here.
Sign to watch
No paid pilot is signed in two quarters.
  1. A

    Wins in two futures on its own

    Sell storage as a service, not as hardware.

    Data-centre operators prefer not to own power equipment. No buyer has been asked yet whether a service contract would win the deal.

  2. B

    Wins in three futures on its own

    Win one data-centre owner first, as a customer others follow.

    One reference customer may decide every later sale. It works whether buyers rent or own.

  3. ↑

    Where to look first

    We would look at 01 and 02 first.

    Both keep the service model alive, and the evidence cannot yet tell them apart.

  4. !

    A risk outside these futures

    99% of refined gallium comes from China.

    A supply shock could change each of the futures above.

2 · The gates

The money is released one gate at a time.

The most you can lose of the EUR 150M if you stop at each gate, as a running total

EUR 3M

EUR 15M

EUR 50M

EUR 100M

First gate

A paid pilot is signed.

A data-centre owner signs a paid pilot. The analysis rates a signed pilot within 90 days as unlikely by the direct route, so two quarters is the real limit.

If no owner signs a paid pilot, the case for this service would be much weaker.

Second gate

Finance signs off.

The customer's finance team accepts the service model and what it costs over its life.

If finance is not convinced by the full cost, the service model may not work for these buyers.

Third gate

It runs on site.

The system meets the customer's own technical targets in their building.

If the product misses the pilot's targets, it may not be ready for the market yet.

Released if all three pass

The amounts are illustrative: they show how the EUR 150M could be released, not a fixed plan.

3 · The tension

Buyers need power urgently, but they change suppliers slowly and prefer not to own.

The marketThe planThe tension

Data centres wait over four years for a grid connection. And in a survey of large manufacturers, 71 in 100 kept their energy-management provider.

Outside evidence disagreed with the plan in 13 places. It doubts how strong demand is, and how soon it comes.

The first company to sign a contract tends to keep the account.

Data-centre operators prefer not to own power equipment, so they buy power through contracts.

The plan prices a sale of batteries, and no buyer has checked its numbers.

Buyers want to pay for power as a service, yet the plan sells batteries to own.

The first battery deals at data centres use 2-hour systems.

The product is built for 8 to 12 hours, and no one outside the company has tested it.

The market buys 2-hour batteries. The product runs longer, costs more and is untested.

4 · The evidence

What the sources say, what the analysis recommends, and patterns seen before.

5Domains scored
→
28Factors analysed
→
184Evidence items
→
111Sources
→
4Futures mapped
The outcomeApprove with conditionsWe suggest a 90-day test. If no paid pilot is signed within two quarters, the advice would change to not this year.

Facts from the sources

  • Big Tech plans to spend over $300bn on AI in 2025.

    Financial Times, 2025

  • AI is set to drive a 160% increase in data-centre power demand.

    University of California paper, 2024

  • Data-centre electricity use grew 3% a year from 2005 to 2015, and 10% a year from 2015 to 2024.

    IEA, Energy and AI, 2025

  • Only 25–30% of data centres use lithium-ion batteries for backup power.

    BCG, 2025

  • Operators prefer not to own power assets. They buy clean power through purchase agreements instead.

    IEA, Energy and AI, 2025

  • Grid waits are long and rising: one 2024 source puts hyperscale connections at 1–3 years, another at 7–10, and a 2026 outlook puts primary markets beyond four years.

    US DOE, 2024 · Brookings, 2024 · JLL, 2026

What the analysis recommended

Three recommendations

  1. NORVIA sells storage as a service, not as hardware.
  2. It partners first with the biggest cloud companies that try new technology early.
  3. It rolls out in phases, starting where policy favours it.

Two risks

  • Intense competition, and margins worn down.
  • Buyers are slow to take up new technology.

Patterns seen before

Service keeps the customer

Ongoing service keeps the customer with the seller.

Access is not adoption

A quote does not become a purchase when friction after it stalls the buyer.

Product as a service

Customers pay for use and never own the product.

Credit tightening

Hardware projects stall when credit tightens.

Supplier concentration

The plan depends on one supplier or one region, which concentrates the risk.

5 · Who says yes

Three parties decide whether the service is bought.

The operator

A data-centre owner, or a big cloud company, short of grid power

Holds
The site, and the power it cannot wait for.
Weighs
Operators prefer not to own power assets. They buy clean power through purchase agreements.

Says yes when

A paid pilot shows the service works on their site.

Their finance team

Finance or purchasing at the operator

Holds
The total-cost decision.
Weighs
In one study, 124 procurement managers rated long-term cost savings above environmental benefits.

Says yes when

They accept the service model and its full cost.

The early adopter

One or two operators who adopt new technology early

Holds
The reference that every later sale may follow.
Weighs
Batteries on site are still new: 25–30% of data centres have them.

Says yes when

They get better terms in return for being a public reference.

6 · What it cannot answer

The scores show a direction, and they are not forecasts. The evidence behind them is moderate to weak.

  • How much does the service save an operator over its life?

    If the saving is clear, selling can start. If it is marginal, the service model is in doubt. A customer's finance team could test it during pilot talks.

  • Which risks do operators fear most?

    Knowing the top one or two could shorten the sales cycle. Interviews with five to ten operations directors would show them.

  • Can the company supply at the pace data centres need?

    If it can, it becomes a lasting partner. If not, it stays in small niche work whatever the pilot shows. A study of owning or partnering for key parts would answer it.

  • Is 8 to 12 hours the right size?

    The first deals at data centres use 2-hour systems. If buyers want short ones, the product may be too large. The pilot customer's own targets would show it.

Sources: the analysis's own documents and research · JLL, Global Data Center Outlook, 2026 · Energy Policy, 2019 · IEA, Energy and AI. The company's name is changed.

Published in The Decision Layer · Updated · The case in two pages (PDF)

The world as we see it

We live in a decision economy. Companies now compete on how well they choose where the money goes. Most industries have changed how they work, but we still decide in slides. A slide is static and slow to build, and it falls behind when the market moves fast. In five years, we see strategic decisions as comparable as stocks on a Bloomberg Terminal. Every team puts its decision on the same footing. The people who evaluate them can see them as they change, and act sooner.

Bring a real case you are deciding on.

Bring us a real case