Home Projects Short Term Strategy Long Term Strategy About Contact

Deal Origination · Real Assets · Switzerland

Assets that have to move.
Owners who have to sell.

Real Assets. The shovels of every gold rush.

CHF 5–150M
Deal range
5
Core pillars
Switzerland
Domicile · Swiss law

The Story

A name is not branding.
It is an explanation.

What was missing in Kamenica was never goodwill. It was infrastructure priced, built and distributed wrongly — and the capital that could have fixed it went somewhere else. Everything this firm does follows from that one observation.

The Story of Kamenica
Intro · I · II · III · Epilog
Open

Investment Philosophy

Preserve value.
Do not predict markets.

What carries value through a cycle is not yield. It is control over something that stays mechanically necessary — and the discipline to buy it from someone who has to let it go.

Buy what cannot be replicated and what others will need. That logic has not changed in centuries.
Filter one

Entry price

Buy from sellers under structural pressure or from holders who are indifferent. Hold across decades, not quarters. The return sits in the counterparty's constraint, not in the market's direction.

Filter two

Irreversibility

Buy lasting mechanical necessity that the market prices as a cycle. Reject anything a signature can undo — that line separates a structural position from a bet on politics.

Five Pillars

Where we look.

Kamenica Capital is not a fund, not an asset manager and not a bank. We originate real asset transactions between CHF 5M and 150M — too specialised for the large banks, too complex for retail platforms.

01

Energy & Critical Resources

Grid assets, generation, energy parks with secured offtake. Baseload capacity is the wall every other pillar eventually runs into.
02

Water Infrastructure

Treatment, distribution, and cooling-water dependency in power generation and compute. Judged by mechanism and by unit, not by scarcity talk.
03

Digital Infrastructure

Compute sites where the location cannot be copied and grid access is secured, plus the software layer that runs on top of them. Power access binds, not floor space.
04

Land & Development

Building land, demolition and conversion objects in ZH, SZ, ZG and LU. Off-market, straight to the owner. The current search brief sits here.
05

Hard Assets & Precious Metals

Physical gold in Swiss non-bank custody. Central banks have bought net since 2010 — a buyer whose mandate does not care about the price.
Pipeline

Deal Pipeline

What we are working on.

Across the five pillars, plus two situations that sit outside them. Counterparties, volumes and terms follow a mutual non-disclosure agreement — they do not precede it.

Pillar 01

Generation and grid

Energy parks with secured offtake, grid connection and permits already in place. Actively sourcing.

NDA required.
Geography
Switzerland / Europe
Pillar 02

Water treatment and distribution

Treatment and distribution, and the cooling-water dependency of power generation and compute. Whether that dependency is a running withdrawal or a volume bound once changes what is scarce — and what it is worth.

NDA required.
Geography
Switzerland / EU
Pillar 03

Compute sites and the layer above

Hyperscale and edge sites where the location cannot be copied and grid access is secured. Alongside them, work on power- and performance-efficient computing — which lowers what a site draws from the grid instead of adding to it. Our own analysis software belongs to this pillar.

NDA required.
Geography
Switzerland / EU
Pillar 04

Land and development

Building land, demolition and conversion objects, sourced off-market and straight from the owner. Development and general planning capacity on the delivery side.

NDA required.
Geography
ZH · SZ · ZG · LU
Pillar 05

Physical gold

LBMA-certified gold in Swiss non-bank vaults. Central banks have bought net since 2010 — a source of demand whose mandate does not depend on the price. Accumulating.

NDA required.
Geography
Switzerland

Outside the pillars

Situations that fit no pillar but pass the same screening.

Environmental technology

Air filtration

Durable filter technology for particulate removal, with a multi-year Swiss operating record and expansion into the Gulf. What drives it is tightening air quality regulation, not voluntary demand.

Medical technology

Wound treatment

A wound care device in clinical validation, EU class IIb, patented in the US and EU, with a defined trade-sale exit. The revenue case rests on the reimbursement pathway, not on consumer adoption.

Current search brief

One active brief inside pillar 04. It is deliberately narrow: anything outside these criteria gets a decline at first contact instead of a process that ends in a decline anyway.

  • VolumeCHF 5–150M
  • TypeBuilding land, demolition, conversion
  • GeographyZürich · Schwyz · Zug · Luzern
  • AccessOff-market, straight to the owner
  • SellerUnder structural pressure, or indifferent
  • ExitBuyer can be named, jurisdiction is liquid

The exit criterion is binary. A project that cannot name its buyer does not enter the process, whatever the return looks like.

Submit a project

Short Term Strategy

Who has to act,
and how far it travels.

Run on the firm's own book. Two questions carry the whole approach: who is under an obligation to trade regardless of what they believe, and where does that obligation show up next. The first is measured in the option chains, the second along the chain of production.

Measures

Accumulation and distribution

Reading a chart estimates accumulation and distribution from price and volume. Gamma exposure computes the same thing from the option chains. Both look at dealer inventory — one infers it, the other calculates it.

Selects

Regime picks the product

A Markov regime classifier does not answer up or down. It answers which market is worth looking at at all. Only once a product is selected does direction get calculated — and it gets calculated from supply and demand, not from the chart.

Holds

Until it is priced in

How long a position lives depends on pricing status, not on a calendar. It lives as long as the constraint is unpriced. Once priced, it is over — whether or not it is in profit.

Stops

Breakers on the book

Kill switches act on the book, not on the single trade. A stop protects a position; a breaker protects the ability to keep trading at all.

What a regime rules out

A worked example of the first step, because it is the one that decides everything after it. In an environment of high rates and sticky inflation, the question is not whether a stock goes up. It is which category still has a buyer who has no choice.

01
Ruled out

Aspirational discretionary

The USD 500 ticket and the USD 1,000 bag do not disappear because taste changed. They disappear because fixed costs ate the income they were paid from. Income stayed, housing, debt service and food took the room.
400M → 330Mglobal luxury customer base, 2022 to early 2026 · Bain & Company · measured
Out
→
Where it went

Down, not away

The spending did not vanish, it traded down. Resale, value formats, deferred big-ticket purchases. Anyone shorting the category outright misreads a substitution as a collapse.
60%of US and European consumers buying second-hand luxury · J.P. Morgan survey, September 2025 · measured
Shifted
02
Stays in

Cashflows that carry the index

Not real assets as a class — that is too coarse. What hedges is the contract underneath: CPI-linked concessions, regulatory resets, rent escalation. Someone is obliged to pay more when the index moves, and the obligation does not ask whether they want to.
+5% / +7%listed infrastructure over global equities p.a., at 3–4% and above 4% inflation · First Sentier, 15-year series · derived from an index comparison
In
03
Counter-check

Not everything physical qualifies

The same quarter that rewarded contracted infrastructure punished rate-sensitive property vehicles. The dividing line is not physical against financial. It runs between what can pass the index through and what only gets discounted by it.
REITs and TIPS weakestwithin a real assets strategy, Q1 2026 · State Street · measured
Split

So the regime does not say buy infrastructure and sell luxury. It says the buyer under obligation has moved from one category to the other, and it names where to look for the next calculation. Direction still has to be derived from supply and demand once the product is chosen.

One caveat that belongs next to the numbers: the discretionary pullback is a split, not a collapse. The top of the income distribution kept buying, and forecasts for 2026 still show low single-digit growth for personal luxury goods. What broke away was the aspirational tier, and that is the part driven by fixed costs. Status as at 28.07.2026.

The chain

A shock does not stay where it lands. It travels, and it takes time doing so. The model keeps three things apart that usually get lumped together — and that separation is where most of the work sits.

Link

The material changes. Raw material, intermediate, finished good, consumption. Four of them.

Connector

Same material, new place and later. Carries freight, cover, delay. Sits between every pair, which is why logistics shows up wherever it bites.

Event

Changes nothing itself. It hits a named point in the chain, and whether it can be undone by a signature decides everything downstream.

Worked example: Hormuz

The trigger was not forecast, and we do not claim it was. What was calculated is what follows once it happened. Below is that calculation against what actually occurred — including where it went wrong.

◆
Event

Strait closure

Hits the transport section between raw material and intermediate.
15 / 88transits per day · 19.07.2026 against the pre-crisis norm · IMF PortWatch · measured
Reversible
01
Link

Raw material

Production kept running. The barrel was never scarce — the passage was.
output heldGulf producers at or above quota through Q2 2026 · derived from production reporting, no independent volume series · lag 0
Held
→
Connector

Freight and cover

Lag under one week · pressure forward
Cover is compulsory, the passage has no substitute, and repricing lags because underwriting moves slower than diplomacy. This is where the shock actually landed — and the only place it held.
0.25% → 7.5–10%war-risk premium as a share of hull value · pre-war against 22.07.2026 · Marsh via Platts · measured. Freight spread USD 469k/day with transit against 220–230k without · measured.
Held
02
Link

Intermediate — crude price

Modelled in April as a pair of scenarios. Wrong quantity to model: this level is set at a negotiating table, not by an obligation, and it turned three times.
120.88 → 71 → 100+ → 88Brent USD/bbl · peak 30.04. · 06.07. · 23.07. · 27.07.2026 · measured. The April forecast around it was assumed and is labelled as such.
Missed
03
Link

Finished good — refining

Crack spreads widen when input costs rise against output sold under contract. Direction confirmed, size not derived independently.
jet fuel spike, July 2026direction measured, magnitude not derived · lag roughly four months from the event
Partly
→
Connector

Last mile

Lag not established · pressure forward
Not measured separately. Named anyway, because leaving it out would quietly load its cost onto whichever link sits next to it.
not measuredlowers coverage instead of being filled with a plausible number
Open
04
Link

Consumption

Showed up as a guidance revision rather than a price tag — which is where pass-through becomes visible first.
US airline guidance cutJuly 2026 · derived from company guidance, not from a price series · lag roughly four months
Delayed

Coverage: three of seven positions carry a measured value with a source and a date — the event, the freight connector and the crude price. Two are derived from reporting rather than from an independent series. Two are not measured at all. Coverage therefore stands at 43%, and it is stated rather than closed with plausible numbers. Total lag from event to consumption: roughly four months.

Dated record

Highlighted rows are our own dated working files. The rest is public record.

02.03.2026Strait closed · war-risk cover pulled
08.04.2026Ceasefire · Brent 110 → 96
12.04.2026Own scenario pair: 60% ceasefire holds → 85, 40% breaks → 110
30.04.2026Brent peaks at 120.88
17.06.2026Memorandum · exports partly restored
06.07.2026Own review: nine corrections · nothing survives as a tradeable conviction
23.07.2026Brent back above 100
27.07.2026Brent ~88, minus 8% in a day, talks resume

What the miss shows

Whoever sat on either branch of the April pair would have been stopped out — the June branch by July, the July branch inside a day. The mistake was not the weighting. It was modelling the one quantity that gets decided at a negotiating table.

The structural error sat one level deeper. In the earlier version, freight and insurance ran as a link, as though they transformed something. They do not — the same barrel goes in and comes out, only later and dearer. Pulled out as a connector, the picture reads differently: the shock never hit a link at all. It hit the transport section between two of them, which is also the one place the effect held.

Three of the nine corrections from 6 July belong to this chain: the price level, the normalisation timeline, and the war-risk figure, where a pre-attack warning level had been carried forward as if it were the wartime premium.

What held were the places where someone has no choice: the insurer who must cover, the charterer with no second route. Direction was opinion. Obligation could be measured.

The programme

Everything above runs by hand today. We are building it into our own software: the chain with its links, connectors and events, the regime classifier that decides admissibility, and the gamma calculation from the option chains. It is in beta and not publicly available. Access opens from Q4 2026 — participants use it and feed back into it, and that feedback shapes the next version.

Join the beta

Long Term Strategy

Hold what stays necessary.

The short-term work asks who has to trade this week. This one asks what still has to exist in twenty years — and who will be forced to sell it before then. Preservation, not prediction: the future cannot be forecast, so the portfolio is built to survive each of the four most likely environments rather than to bet on one.

Branch A

Base case

The compute build-out holds. Energy and grid infrastructure reach their strongest pricing power.

Branch B

Adjacent

Capital spending moderates. Yield with something physical behind it beats growth stories.

Branch C

Contraction

Capital spending contracts. Diversified energy infrastructure and gold carry the book.

Branch D

Shock

Geopolitical rupture. Switzerland reprices as a destination for capital that is shopping for a jurisdiction.

What survives a signature

Force alone is not enough. An obligation that a negotiated agreement can lift is a direction, not a structure — and the Hormuz chain showed exactly that, three times over. So every candidate runs through four questions before anything else is looked at.

01

Restart time

Does it run again the moment there is an agreement, or does something have to be physically rebuilt?

02

Storability

Does the stock survive the interruption, or does time itself destroy it?

03

Substitution

Is there another input that does the same job?

04

Lead time

How long from decision to delivery for new capacity?

Two candidates, neither a position

Both were run through the same procedure. Both stopped, for different reasons — which is why both are shown.

Stops at the unit

Water for compute

Chain
Efficiency opens new use cases, use cases add sites, sites add cooling. Better hardware does not shrink the demand — it grows it.
Correction
A closed loop does not remove the dependency, it changes its unit. Filled once, circulating, bound — and short of a change in cooling technology it barely comes free again. Stock, not flow.
Open
Size. The bound volume grows with every new site, while water rights are priced as an annual flow. Whether the bound stock is large enough to move that pricing has not been derived — the mechanism holds, the magnitude does not yet.
Stops at the pricing

Helium

Chain
It comes out of LNG processing as a by-product, so output stops dead when the plant does. Liquid stock boils off in 35 to 48 days. In wafer fabrication nothing else does the job.
Passes
Force and irreversibility. Damaged trains need a physical rebuild, not a signature.
Open
The pricing. Spot has already doubled and only the contract side lags. And part of the outage is a plant waiting for the route to reopen — reversible, and not to be counted with the rest.

Where this is going

Origination is the entry point, not the destination. Finding an asset and passing it on is a fee. Holding it, developing it and keeping control of it through a cycle is a position — and that is the direction of travel.

Today
Origination and structuring. We find assets, run them through the screening, and match them to capital. What we own ourselves is limited and stated as such.
Building
Our own analysis software, currently in beta. It is not a side product — it is the first piece of digital infrastructure we own outright, and it belongs to pillar 03 for that reason. Until it stands, we work with external builders rather than pretending to do it alone.
Ahead
Infrastructure development. Not brokering someone else's asset but building and holding the thing itself — energy, water, compute, land. That takes balance sheet, partners and years, and none of it is claimed as present capability.

Founder

Bardh Shabani

Founder of Kamenica Capital. Sole proprietorship, Döttingen, Aargau. The firm was set up in February 2026.

My first contact with institutional real assets was a solar park. What it settled was not a strategy but a direction — the sense of standing in the right place. The strategy came later, out of watching which claims survive scrutiny and which do not.

What I put in front of someone, holds. Where it does not, the correction is published next to it.

What the firm is not

  • NotA fund or collective investment scheme
  • NotAn asset manager — no third-party money is managed
  • NotA bank, and not FINMA-licensed
  • IsA deal originator under Swiss financial market law

How We Evaluate

Three laws, and what
every number has to carry.

Every figure we produce passes the same three constraints. A number that breaks one of them does not get published in a softer form — it gets marked as missing.

01

Traceability

Every figure carries its full derivation. Being quoted somewhere does not verify a number — only the source's own derivation does. If the source shows none, you are holding a citation, not a fact.

02

Mechanical force

Movement comes out of balance sheets and obligations, not out of opinion. The question is never what someone expects. It is what someone has to do anyway.

03

Degree of freedom

Deterministic formulas run as code. Open questions run as language models with grounding required. The two never share an output field.

Three states, not two

Unknown inputs lower the coverage ratio. They do not get filled with plausible values — a plausible filler is invisible downstream and survives every check except tracing it back to source.

Measured

Value, unit, source, date, and the basis the source itself used.

Derived

Named inputs, named operation. Takes on the status of its weakest input.

Assumed

The value chosen, why it was chosen, and the range in which the conclusion still holds.

Eight-layer due diligence

Whatever clears the binary criteria enters the eight-layer process. Whatever fails at layer I goes no further.

I

Filter

Minimum volume, secured land rights, permits in hand.
II

Structure

Legal form, ownership, capital structure.
III

Revenue

Cashflow, offtakers, contracts.
IV

Technical

Feasibility, technology, what the sponsor has actually delivered before.
V

Regulatory

Jurisdiction, tax, FINMA relevance.
VI

Exit

Named buyer, secondary market, repayment — settled before entry.
VII

Fit

Which pillar, what fee, what NDA status.
VIII

Internal view

Confidential. Shared under NDA.
Contact

Contact

Write directly.

For an object that fits the brief, for capital looking at Swiss real assets, or for beta access.

bardh@kamenicacapital.ch

Kamenica Capital · Bardh Shabani
Döttingen, Aargau, Switzerland

What happens next

  • 01An answer within two working days — including a decline, with the reason.
  • 02Screening against the binary criteria. Most enquiries end here, and they end fast.
  • 03A mutual NDA before any figure, name or document moves in either direction.
  • 04An engagement letter setting out scope, fee and term before any work starts.

Legal

Privacy Policy

Controller: Kamenica Capital, Bardh Shabani, sole proprietorship, Döttingen, Aargau, Switzerland, bardh@kamenicacapital.ch. Personal data is processed under the Swiss Federal Act on Data Protection and, where it applies, the GDPR.

What is collected

Server logs via the hosting provider: IP address, browser and operating system, referrer, time of access, pages viewed — used for security and operation and nothing else. What you send by email or when asking for beta access: name, email address, the content of your message and any attachments, processed to handle your enquiry or to administer beta access. Documents submitted under a non-disclosure agreement are used to assess and manage the business relationship and are treated as confidential.

Cookies and local storage

The browser's local storage is used for exactly one thing: remembering which language you picked. No analytics, no tracking, no profiling, no third-party cookies.

Transfer and retention

Data is not sold, traded or handed to third parties for marketing. Transfers outside Switzerland or the EEA happen only where adequate protection is in place. Data is kept only as long as the stated purpose or the law requires.

Your rights

  • Information about the data held on you
  • Correction of anything inaccurate
  • Deletion, subject to statutory retention duties
  • Restriction of, and objection to, processing
  • Release or transfer of your data

To exercise any of these, write to bardh@kamenicacapital.ch. You can also complain to the Swiss Federal Data Protection and Information Commissioner. Transmission is encrypted over TLS. Swiss law applies; the courts of the Canton of Aargau have exclusive jurisdiction.

Last updated: July 2026.