The physical economy, repriced
Infrastructure is what the economy calls everything it cannot afford to notice until it fails. A pipe bursts. A grid strains. A data center waits for power. Then the hidden machinery of ordinary life becomes the only story in town. McKinsey partners Alastair Green and Adrian Kwok would like investors to notice sooner. In a conversation with Inside the Strategy Room host Sean Brown, they put a number on the next wave of construction, repair and reinvention: $106 trillion of global infrastructure investment through 2040. It is not a forecast for bridges alone. Their definition stretches from airports and railways to fiber, satellites, cooling systems, waste processing, agriculture and defense. The bet is on the physical layer beneath nearly every other bet.
Watch on YouTube →The number under the floorboards
Kwok’s premise is deliberately broad. Infrastructure, he says, includes “all the buildings, all the assets that support every industry in the global economy and broader society.” Add them up and the estimate reaches $106 trillion—about twice the investment of the previous 15 years. The number is closely tied to global GDP because roads, power, water and communications are not decorations on growth. They are its operating system.
The geography matters as much as the total. Roughly $70 trillion is expected in Asia, where urbanization and economic expansion create demand for new systems. In the United States and Europe, the work tilts toward replacing and modernizing what is already there. Kwok cites studies suggesting that failing to modernize aging US infrastructure could cost the economy $10 trillion by 2039. Elsewhere, particularly across parts of Asia and Africa, the job is more elemental: build enough transport and energy capacity for growing cities and economies to function.
That creates a funding question no finance minister can dodge. Governments have historically carried much of the load, but many are fiscally constrained. Pension funds and insurers, meanwhile, want long-duration investments with infrastructure-like risk and return. Green hears a renewed appetite in Europe and the Americas for private participation, even if policy has not yet caught up with the speeches. “We certainly haven’t seen the volume of real P3s kind of matching the rhetoric yet,” he says. The door is open. The traffic through it remains light.
We simply cannot get the things we need to build built without the help of our private investor friends.Alastair Green
AI brings its own fuse box
Artificial intelligence appears in the conversation as both appetite and tool. The appetite is obvious: training and running models requires data centers; data centers require racks, cooling, fiber, electricity and sometimes large amounts of water. Kwok calls these the “picks and shovels” of the AI boom. Whatever happens among competing models at the surface, the buildings and equipment underneath still have to exist.
That demand radiates outward. Power investment in the United States, relatively stagnant for decades in Kwok’s account, is accelerating. Massive training campuses can sit far from metropolitan areas; smaller edge facilities must stay near users when latency matters. Governments pursuing sovereign AI add another local requirement. Digital infrastructure is becoming global and local at the same time, determined by a mix of electricity, temperature, latency and political control.
Then AI turns around and helps build the boom it caused. Green points to generative scheduling and design tools used on complex projects costing more than $200 million. Even among sophisticated clients building renewables and data centers they have built before, he and Kwok regularly see cost and schedule compression of 10%, 15%, 20% and sometimes 25%. Adoption remains below half the market, Green cautions, but the effect is large enough to change project economics rather than merely polish them.
Kwok sees the same shift in the information itself. Infrastructure data used to be scattered and unstructured. New tools can consolidate it, find patterns and influence investment committee memos and diligence. “You can’t really look in any direction and not see what AI is doing to infrastructure investment,” he says. That is the loop: software demands more physical capacity, then helps squeeze more capacity from the people and assets available.
The bottleneck wears boots
The $106 trillion vision sounds abstract until Green starts listing trades. Mechanical. Electrical. Concrete. Flatbed transport. Over-the-road trucking. Anyone who has recently tried to hire a plumber already knows the pattern in miniature. Data centers and power systems are being ordered in extraordinary quantities, but they still depend on finite crews working in sequence. Capital can be raised faster than an electrician can be trained.
That is why scheduling is more than a software demonstration. When labor is scarce, rearranging a project to use each subcontractor more efficiently changes both cost and completion time. The pressure is forcing innovation among leading investors and their portfolio companies. It also punctures a comforting idea about infrastructure: a giant total addressable market does not guarantee that projects can be delivered on demand.
When a resource cannot scale quickly, coordination becomes a source of capacity. Better sequencing does not create more electricians, but it can waste fewer electrician-hours.
Need is not the same as a deal
The most revealing mismatch in the episode is between what the world needs and what private funds can buy. Kwok expects transport to represent about 40% of the $106 trillion requirement. Yet in 2025, only 12% of private infrastructure deal activity went to transport. Digital and telecom took 27%; energy—including utilities, renewables and storage—took 49%.
Kwok’s hypothesis is not that investors dislike transport. The investable deal side has not caught up. Much transport infrastructure remains government-funded, and the projects available may not match the size, structure or return profile that large funds require. One major investor told him that even within waste and water, the firm simply looks at everything capable of absorbing a $500 million equity check because so few opportunities clear that bar.
Scale splits the market. The top 20 asset managers have raised roughly half of infrastructure capital over the past decade. Mega-funds need mega-projects; a smaller fund with $150 million to deploy can pursue niches. Meanwhile, managers from private equity and real estate are asking whether adjacent holdings—from senior living and warehouses to data centers and infrastructure services—can be regrouped under an infrastructure banner. The label is expanding because the assets are expanding, and because limited partners are asking for more.
From toll roads to 100-plus niches
Green and Kwok’s map contains more than 100 subverticals across seven broad territories. The familiar pillars remain, but the interesting action increasingly happens where they overlap: power for data centers, transport electrification, waste converted into energy, smart parking, microgrids and district energy. A fund can call itself a generalist. It cannot be equally good at everything.
Kwok’s recommendation is to choose themes rather than chase every category: digital and power, electrified transport, waste-to-X. Green takes the argument one step further. Assume someone else always has cheaper capital. To win, an investor needs another edge—pattern recognition, an appetite for a particular risk, operating knowledge or an ability to use AI differently. The question before setting the final purchase price becomes brutally practical: what can we do with this asset that somebody else cannot?
The money arrives; expectations follow
Fundraising finally caught up with the enthusiasm in 2025. Nearly $200 billion flowed into designated closed-end infrastructure funds, while private infrastructure assets under management approached $2 trillion—three to four times the level of a decade earlier. More than half of respondents to McKinsey’s 2026 survey intended to increase their infrastructure allocation, the strongest appetite among six private asset classes covered.
The reasons are evolving. Diversification remains first, followed by expected increases in returns and performance. When limited partners judge managers, 70% place performance among their top five criteria, 55% cite the quality of the investment team and 53% cite value-creation strategy. That third measure is rising quickly. Investors are no longer satisfied with a manager buying a stable asset and waiting. They want a plan for capex, procurement, predictive maintenance and commercial improvement.
Deployment looks healthy at first glance. Undeployed capital has fallen from almost 40% of assets under management about 15 years ago to 23%, even as the asset class has grown. Returns are murkier. Five-year distributions as a share of paid-in capital have slipped to about 13%, the lowest since roughly 2012 or 2013. Some fiber, data-center and renewable assets have proved harder than buyers expected, particularly as labor scarcity pushes on operating and capital costs. More money is finding assets. The assets still have to find their way back to investors.
Infrastructure deserves your attention. There’s a lot happening in it that is fundamentally going to change the way society operates.Adrian Kwok
Old concrete, new eyes
Not every opportunity requires pouring something new. Green describes sprawling portfolios whose owners do not fully know which heating and cooling equipment sits inside each building. Satellite infrared images can reveal which roofs are venting heat in summer. Engineers can then visit the worst ten rather than inspect hundreds or thousands, capturing perhaps 80% of the benefit quickly. A problem that looked like an inventory exercise becomes a targeting exercise.
The same logic applies below ground. A municipal employee may confront aging water pipes with only local experience. A company specializing in trenchless repair can carry lessons from hundreds or thousands of places and transfer a better method from one town to another. Private scale does not always mean a larger institution. Sometimes it means a narrower company with a much wider memory.
Sustainability is woven through these choices: renewable power, batteries, cleaner building materials, waste-to-energy and the decarbonization of old properties. New construction in emerging markets can leapfrog obsolete systems. Retrofitting is harder because the assets are inconsistent and poorly documented. AI cannot make those tradeoffs disappear, but it can help decide where the first engineer should go.
The edge after the check clears
The episode ends on an unusually humble note for a conversation full of enormous numbers. Green urges investors and CEOs to try the new tools, knowing the first deployment may not go perfectly. The second or third might spread benefits across a portfolio. Three years ago, he says, some of the results he now hears about would have sounded like science fiction.
That is a fitting rule for the whole $106 trillion build. No single government, fund or technology can carry it. Public authorities need structures that turn necessary projects into workable partnerships. Investors need to bring more than capital. Operators need to find productivity inside scarce labor and imperfect data. Communities need a genuine win when power-hungry facilities arrive nearby. The total is colossal, but the work remains stubbornly specific: a schedule improved, a pipe repaired, a grid connected, a building made visible. The future of infrastructure will be decided one unglamorous bottleneck at a time.