The Tables at the Back of the Book

Infrastructure is a physical record of the assumptions we made about how the world would behave. Those assumptions have expired.

When I was an engineering student, I took a business engineering class. We all did. It was required.

I do not remember the name of the textbook. I do not remember every formula we were taught. I probably could not solve some of the problems today without a generous amount of time and a patient tutor. But I remember the tables at the back of the book. The professor would direct us to them whenever we needed to make an assumption.

Turn to the back. Find the relevant table. Choose the number.

The tables contained the accumulated experience of the people who had designed things before us. Historical river flows. Rainfall probabilities. Expected demand. Material performance. Construction costs. Operating conditions.

The numbers had been developed years before I opened the book. Some were based on observations gathered decades before that. This was not considered a weakness. It was the whole point. The past provided the data. The table organized the past. We used the table to design the future.

Then there was another table.

This one listed the expected cost overruns for different kinds of major infrastructure.

Roads. Railways. Bridges. Ports.

As I remember it, the expected overruns ranged from somewhere around 40 percent to numbers approaching 100 percent, depending on the project.

That table confused me.

If the cost overrun was expected, was it really an overrun?

If experience told us that a certain kind of project regularly cost 40 percent more than the original estimate, why did we continue calling the lower number the estimate? Why not call the higher number the cost?

Over time, I came to understand that large projects contain enormous uncertainty. Supply chains break. Labour becomes scarce. Commodity prices move. Ground conditions surprise us. Designs change. Contractors fail. Governments intervene.

Large projects are difficult because reality contains more variables than the model.

That was true then. It is true now.

But something more fundamental has changed.

The old approach assumed that while individual projects contained uncertainty, the underlying operating environment remained reasonably stable.

The bridge might cost more than expected, but the river would remain within a familiar range.

The railway might take longer to build, but the ground beneath it would behave roughly as the historical record suggested.

The power plant might experience delays, but the water required to cool it would be available.

The cost table allowed for mistakes around the edges.

The engineering table still defined the centre.

Climate change is now moving the centre.

The river in the model

This week, Romania used military explosives to remove rock from a branch of the Danube. The purpose was not mining. It was not navigation. It was to redirect enough water toward the cooling channel of the Cernavoda nuclear plant to keep one of its reactors operating.

One reactor had already been shut down. The second was at risk.

In neighbouring Hungary, the Danube fell so low that the Paks nuclear plant, ordinarily responsible for nearly half the country’s electricity production, was reduced to just over 10 percent of capacity.

Households and companies were asked to conserve electricity. Freight trains stopped running during peak hours. Decorative lighting was turned off. Factories reduced production. Electricity imports surged. The lights did not go out.

Engineers intervened. Consumers responded. Neighbouring countries provided support. A complete failure was avoided.

But we should not take too much comfort from that.

The system did not come close to failing because the technology suddenly stopped working. It came close to failing because one of its operating assumptions did. The river was still there. There was simply not enough of it in the right place, at the right depth, at the moment it was needed.

The asset remained physically intact. Its operating margin disappeared.

This is what climate risk looks like inside infrastructure.

Not always a hurricane removing a bridge. Not always a fire consuming a transmission line. Sometimes it is a few centimetres of water. A cooling intake that can no longer reach the river. A waterway too shallow for a fully loaded barge. A reservoir that no longer provides the storage, electricity or irrigation around which an entire region was organized.

The infrastructure remains visible. The service it was expected to provide becomes unreliable. And once reliability disappears, value follows.

The smoke outside the window

For those of us living in North America, this can feel like a distant story.

The Danube. Romania. Hungary. A crisis unfolding somewhere else. But the same failure of historical assumptions is visible much closer to home.

In July, Canada had hundreds of active wildfires. Communities in British Columbia evacuated...destroyed. Northern Ontario communities were evacuated or preparing to evacuate. Rail operations were suspended near active fires. Smoke travelled hundreds of kilometres and settled over Toronto, pushing air quality into the very high risk category. One Toronto hospital network reported an 80 percent increase in emergency visits related to respiratory and air-quality concerns over a two-week period.

The fire was in northern Ontario. The operating effects were not.

Schools, hospitals, construction sites, transportation systems, outdoor workers and public events all became part of the same system.

The smoke crossed the distance between a forest and a balance sheet.

Around the world, similar lines are being crossed. Fires in France forced mass evacuations and exposed a substantial gap between total economic losses and the losses covered by insurance. Wildfires spread through Greece and Albania as heat and drought simultaneously strained agriculture and electricity systems across Europe.

The fire does not need to consume an infrastructure asset to impair it.

It can remove the people required to operate it. It can interrupt the railway serving it. It can make outdoor work unsafe. It can increase hospital demand. It can make insurance unaffordable. It can force a community to leave.

The event occurs in one location. The consequences travel through the system.

The river we allocated on paper

The Colorado River offers an even cleaner illustration of what happens when historical data becomes embedded in infrastructure, law and capital.

The 1922 Colorado River Compact divided the river between the Upper and Lower Basins, allocating 7.5 million acre-feet annually to each. The allocation was based on hydrological measurements collected during what we now know was an unusually wet period.

In other words, the table was wrong.

Not because the people who created it were careless. Because they believed the observations available to them described the system. A century later, farms, cities, dams, canals, power systems and industries have been built around those allocations.

Phoenix. Las Vegas. Los Angeles. Agriculture across the American Southwest. Semiconductor plants. Data centres. Hydroelectric generation.

Millions of people and trillions of dollars of economic activity now depend on water rights that exist more reliably on paper than in the river.

The seven Basin states have been unable to agree on how the diminished supply should be shared after the current operating rules expire.

The federal government has now advanced a plan under which Arizona, California and Nevada could face reductions of as much as three million acre-feet annually in the driest conditions, roughly 40 percent of current Lower Basin use.

Arizona may bear the largest burden. The Upper Basin states resist mandatory reductions, arguing that nature has already reduced what they can withdraw.

Cities, farms, industry, power generation, Indigenous nations and ecosystems are now being asked to divide a river that was allocated during a climate that no longer exists.

This is more than a drought-management dispute. It is a hundred-year-old infrastructure and allocation model encountering physical reality. We used the past to determine how much water would exist. We converted that assumption into legal rights. We converted the rights into canals, reservoirs, cities and cash flows.

Then the river changed. The table remained.

What institutional investors thought they were buying

For pension plans, sovereign wealth funds, insurers and other institutional investors, this is not principally an environmental issue. It is an underwriting issue.

Infrastructure has traditionally offered precisely what these investors need. Long-duration assets. Essential services. Stable or regulated cash flows. Inflation protection. High barriers to entry. Returns that can be matched against liabilities extending decades into the future.

A toll road could generate revenue for decades. An airport occupied a strategic location that could not easily be replicated. A port benefited from growing trade and physical scarcity. A power plant produced an essential service. A water utility enjoyed local monopoly characteristics.

The permanence of the physical asset appeared to support the durability of the financial return. But physical permanence is not the same thing as operational resilience.

A port cannot move when the water moves. A hydroelectric plant cannot relocate when precipitation patterns change. A railway cannot easily avoid a floodplain after the track has been laid. A thermal power plant cannot negotiate with a river. A forest fire does not respect the boundaries of a concession agreement. And a thirty-year contract does not become safe merely because it lasts thirty years.

Infrastructure is a physical record of the assumptions we made about how the world would behave. Those assumptions have expired.

That sentence is the investment thesis.

Long duration once reduced uncertainty by giving investors more time to collect stable cash flows.

Under climate volatility, long duration can create more time for the assumptions beneath those cash flows to fail.

The old investment model was straightforward. Buy an essential asset. Use historical variability to estimate the downside. Add leverage.

Assume insurance remains available. Assume the operating inputs remain accessible. Assume a concession or regulatory agreement protects the revenue. Collect the yield.

How institutional investors used to invest in infrastructure is over. Not because infrastructure is less important. Because the methods used to value its reliability are no longer reliable.

The efficiency trap

For decades, investors asked engineers and operators to remove waste from systems.

Reduce excess capacity. Lower inventories. Consolidate suppliers. Increase asset utilization. Avoid redundant infrastructure. Extract greater efficiency from every dollar of capital.

This produced enormous economic value. It also produced systems with very little room for surprise.

One water intake. One transmission route. One supplier. One port. One narrowly defined operating range.

Redundancy looked inefficient because the model treated disruption as a temporary exception. But what appears inefficient during stable conditions can become extraordinarily valuable under volatile ones.

A second grid connection. An alternative water source. On-site storage. Distributed generation. The ability to shift demand. The ability to substitute one material for another. The ability to operate at lower capacity without shutting down completely.

The cheapest system on an average day may be the most expensive system on the day that matters.

A new infrastructure diligence

Institutional investors need to conduct physical dependency diligence with the same seriousness they apply to financial, legal and commercial diligence. It is not enough to ask whether an asset has stable historical cash flows.

Investors need to ask what physical conditions made those cash flows possible.

How much water does the asset require? At what temperature? From which source? What happens if the source falls below a critical level? Can the asset operate at reduced capacity? Can another input be substituted? How long can it withstand a disruption? Which infrastructure does it depend on beyond its own perimeter? Where are the single points of failure? Who must pay to adapt it? Will the regulator allow those costs to be recovered? Will insurance remain available? Does the debt structure provide room for interruption, or does a temporary operating failure quickly become a covenant failure?

These are not sustainability questions sitting beside the investment model. They are the investment model.

A plant without cooling water does not have an ESG problem. It has a revenue problem. A port that cannot accommodate fully loaded ships has a throughput problem. A railway repeatedly closed by fire or flood has an availability problem. A community that cannot obtain insurance has a property-value and municipal-finance problem.

The distance between the river and the discount rate is shorter than investors think.

What should long-term capital own? The answer is not simply more climate infrastructure. The category is too broad. The better question is:

Which assets and technologies make prosperity less dependent on the world behaving exactly as it used to?

Water recycling that allows industry to reuse rather than continuously withdraw water. Cooling systems that reduce freshwater requirements. Storage that allows electricity to move through time. Transmission that allows it to move through geography. Distributed generation that prevents one failure from becoming everybody’s failure. Demand-response systems that create capacity without building another power plant. Wildfire detection and forest-management technologies that identify risks before they become emergencies. Industrial systems able to substitute inputs. Agricultural technologies that preserve productivity across a wider range of soil and water conditions. Sensors and controls that detect stress before failure.

These are not merely environmental investments. They are investments in uptime.

Continuity. Resource productivity. Avoided capital expenditure. Reduced volatility.

Institutional investors have historically paid a premium for predictable cash flows. They should now pay for the systems that make cash flows predictable.

The fiduciary question

Pension plans occupy a unique position. Their liabilities are intergenerational. Their beneficiaries will still require payments long after today’s investment committee members have retired. That should make pension plans the natural owners of resilient infrastructure.

But only if long-term investing means more than owning an asset for a long time.

A long holding period is not the same thing as long-term thinking.

Long-term thinking requires investors to question whether the physical conditions supporting an asset today will persist throughout its intended life. It requires them to challenge engineering standards based entirely on historical observations. It requires them to recognize that climate-adjusted capital expenditure is not necessarily an additional cost.

It may simply be the actual cost of owning the asset.

And it requires them to value resilience before a disruption reveals what its absence costs.

Back to the tables

I keep thinking about those two tables from school. The first told us what conditions to expect. The second told us how wrong the project estimate was likely to be.

In retrospect, the second may have been the more honest one.

It acknowledged that the model would eventually encounter reality. But even that table assumed we understood the general shape of the uncertainty.

Start with the base case.

Add a contingency.

Proceed.

That is no longer enough.

You cannot solve for a moving probability distribution by adding 20 percent to yesterday’s estimate. You cannot make an asset resilient by placing a larger contingency beside the same old assumptions. And you cannot meet a pension liability with infrastructure cash flows that exist only when the physical world cooperates.

The Danube has not disappeared.

The Colorado River has not disappeared.

Canada’s forests have not disappeared.

The lights remain on.

But Romania is blasting rock from a riverbed to protect a power plant. The United States is reconsidering century-old water allocations supporting major cities and industries. Canadian communities are evacuating while smoke closes the distance between northern forests and Toronto hospitals.

These are not separate environmental stories. They are the sound of the old tables failing.

The question for institutional investors is not whether climate change belongs in infrastructure underwriting. It is whether infrastructure underwriting that does not begin with climate volatility can still be considered underwriting at all.

The old way of investing is over.

Not because infrastructure matters less. Because it matters too much to continue designing, financing and valuing it for a world that will not return.

The future cannot be found at the back of the book.

We have to build for it.

---

To read more, you can purchase The Gigacorn Hunter: Seven Principles for a Climate Investor here.