Goldman Sachs says the most capital-hungry investment cycle in history has arrived. I read the note twice. Not because the conclusion is controversial. Because the conclusion is hiding the real problem. Capital is not the bottleneck. Verification is.
Every cycle has a signature phrase. In 2008 it was “innovative mortgage products.” In 2021 it was “algorithmic stability.” In this cycle, the phrase is “capital intensity.” The bank’s analysts point to AI data centers, energy grids, semiconductor fabrication, defense supply chains, and a deliberately revived industrial base. The list is broad. The implications are enormous. The report says this capital-intensive cycle could reshape global economic structures and drive significant growth in infrastructure and finance. I do not dispute the size. I dispute the framing.
Goldman Sachs is a good bank. It has better data than almost anyone. But “capital-hungry” is an interpretation, not a measurement. It turns an aggregate spending forecast into a story about demand. It assumes the world urgently needs all this capital, and that the only remaining question is who will supply it. That assumption is worth tearing apart. Because if you tear it apart, you find a different headline: we are entering the first capital cycle where the scarce resource is not money, not even engineering labor, but the ability to prove that the money went where the model said it would.
Let me reconstruct the cycle in data terms. AI data centers are now multi-gigawatt projects. Grid interconnection queues are measured in hundreds of gigawatts. Chip fabrication plans run on multi-year construction schedules while the underlying technology depreciates in months. Defense procurement, reshoring programs, and energy transition mandates are simultaneously competing for the same steel, concrete, transformers, and skilled labor. This is not a normal capex boom. It is a collision of several capex booms, each with its own lead time and each with its own failure mode. The word Goldman chose — “hungry” — is accurate. The question is what that hunger is for.
The technical leg of the argument goes like this. Capital intensity is not a vague word. It is a ratio: capital expenditure divided by output, or capital stock divided by GDP. During the industrialization of the United States, the electrification of Europe, the postwar reconstruction, and the Chinese urbanization boom, capital spending as a share of GDP behaved like a KPI for the age. It rose, peaked, and fell. Every peak paid for infrastructure that later became the platform for a productivity boom. The current cycle has the same shape, but a different substrate.
Previous infrastructure cycles built physical assets that depreciated slowly. A bridge lasts fifty years. A port lasts longer. The balance sheet math was straightforward. You borrowed for thirty years, built something that would still be productive in thirty years, and repaid with a margin embedded in the economy. This cycle builds assets that depreciate violently. An AI training cluster does not last fifty years. It lasts three. A chip fabrication plant is dated the moment the next node is announced. Grid interconnects last, but the compute they feed is obsolete every eighteen months. We are financing compute with steel-age financial instruments. The real news is not that the cycle is capital-hungry. It is that the cycle is time-hungry in a way the capital market has never priced.
This mismatch is where the “finance” part of the forecast enters. Balance sheet expansion is the name of the game. Project finance, infrastructure debt, tax-equity deals, refinancings, asset-backed structures, and the entire machinery of institutional allocation will grow because the number of moving pieces is growing. But watch the plumbing, not the headline. The financial system that will be asked to carry this cycle is still running on legacy assumptions about settlement, reconciliation, and custody. In a traditional repo or a construction loan, the counterparties reconcile at the end of the day. There is a ledger, but it is not a real-time ledger. There is a custody chain, but it is often a custody handshake. The costs of that handshake were the price of trust. In a capital-hungry cycle, those costs compound.
Based on my audit experience, I can tell you what that compounding looks like. In 2024, when the spot Bitcoin ETFs were approved, I audited custodial wallet solutions used by major asset managers. The marketing materials said “bank-grade security.” The actual threshold signature logic had three potential attack vectors in the key-shares distribution protocol. I reported them privately. The teams fixed them. But the experience stayed with me. The money was there. The plumbing was not. If top-tier custodians run on assumptions that fail under stress, what will happen when the next-generation institutional giant tries to move ten billion dollars through a data center special purpose vehicle? The problem is not necessarily a hack. The problem is a reconciliation error so expensive that nobody notices until the cycle turns. Capital is not the bottleneck. Verification is.
This is a technical observation, not a philosophical one. Capital allocation is a data pipeline. Every dollar that moves from a pension fund to a power plant passes through a series of states: commitment, settlement, drawdown, capitalization, progress report, cash flow. At each state, somebody has to verify that the physical world matches the financial world. In the old economy, verification was done by lawyers and engineers with binders. In the new cycle, verification is done by software, and software is never neutral. “Code is law, but bugs are reality” is not a slogan. It is the closest thing we have to an audit framework.
The crypto industry should be the natural auditor of this cycle. That is not the same as saying crypto assets will somehow capture the infrastructure boom. But the technical primitive that crypto invented — the immutable, machine-verifiable record of state transitions — is precisely what a capital-hungry century needs. The problem is that too much of crypto is still trying to be the borrower instead of the verifier. I have spent the last two years working on zero-knowledge proofs. I know how this sounds to people who think ZK is just a way to obscure token holdings. It is not. The ability to prove that a statement is true without revealing the statement is the most important accounting tool since double-entry bookkeeping.
Here is the translation. A sovereign wealth fund wants to fund a renewable energy project in a foreign jurisdiction. It wants to know that the project meets emissions standards, that the labor is legal, that the equipment was actually installed. The counterparty wants to keep its commercial terms private. In the old world, this is a conflict. In the world of composable privacy, it is a technical exercise. You build a proof that the emissions certificate is valid, that the title transfer is clean, that the employee count is real. You do not reveal the underlying data. Privacy is a feature, not a bug, because privacy is what allows the verification to be honest. If the counterparty has to hand over the entire commercial ledger to prove one fact, the incentive to fake the fact is too high. If they can prove the fact without exposing the ledger, verification becomes cheaper and more likely to actually happen.
I built enough ZK systems to know they are not magic. In the 2022 bear market, I spent six months building a minimal zkSNARK proving system from scratch in Rust, implementing Groth16 from first principles. I debugged more than two hundred lines of assembly. The algebra is unforgiving. A single constraint error breaks the whole proof. But that is exactly the point. Math doesn’t negotiate. The proof either verifies or it doesn’t. That non-negotiability is what makes ZK valuable for infrastructure finance. It replaces a relationship with a proof.
In 2025, I worked with a legal-tech startup to integrate zero-knowledge compliance proofs into a DeFi lending protocol. We designed a circuit that verified creditworthiness without exposing personal data. We reduced proof generation time from 500 milliseconds to 150 milliseconds. That is not a meaningless benchmark. It is the difference between a system that can be embedded in an application and a system that lives in a research paper. The same optimization can be applied to project finance. Imagine an infrastructure loan where the borrower proves that the equipment has been delivered, the site is active, and the headcount is real — without exposing payroll, pricing, or supplier identities. That is not a futuristic vision. It is a circuit design problem.
This is what Goldman’s “finance sector growth” should mean. Not more fee layers. Not more exotic vehicles. More verification infrastructure. But I have learned to be suspicious of the word “growth” when it comes from a sell-side institution. Growth in finance is often growth in complexity. Complexity is revenue. The financial sector grows when the number of handoffs grows, and the number of handoffs is not correlated with economic value. It is correlated with the ability of intermediaries to insert themselves into the flow. The real world is about to do a massive amount of capital construction. The financial sector will grow around it regardless. The question is whether the growth is a scaffold or a tumor.
Watch what happens to the terms. It will start with co-investment. It will move to special purpose vehicles. Then it will move to securitization. Each step is a legitimate response to a real friction. Each step also creates a new layer of counterparties. Somewhere in that stack, the ratio between value creation and fee creation inverts. That is not a conspiracy. It is an equilibrium. The more complex the allocation problem, the more expensive it is to verify, and the more intermediaries charge for verification without actually doing it. A document signed by a consultant is not verification. It is a claim that verification happened.
The same problem exists across digital infrastructure. The blockchain industry has spent years building bridges between chains, but the security model of most bridges is still a story about oracles and relayers. Far from being decentralized, the destination chain has to trust that a handful of off-chain actors are telling the truth. That is not a bridge. That is a handshake with extra steps. If the handshake is too expensive to verify, you do not fix it with more tokens. You fix it with a proof. The same logic applies to sovereign funds, project lenders, and grid operators.
Here is the blind spot in the Goldman framing. The forecast assumes that “capital-hungry” is the same as “opportunity-rich.” It isn’t. Capital hunger can be the product of cheap capital. When interest rates are low and fiscal deficits are wide, every project looks like an infrastructure project. Every corporate treasury looks like a buffer against inflation. Every data center looks like the new rail. That is not necessarily because the world has never had so many good investments. It is because the price of wrong bets is being subsidized by the people who supply the capital. The cycle is capital-hungry in the same way a recovering addict is drug-hungry. The desire is real. The object of the desire is not always good.
I have seen this pattern before. In the crypto industry, the popular phrase is “liquidity fragmentation.” It is supposed to describe a real problem: dozens of Layer2s, dispersed users, thin books. But after watching the 2021 collapse and the 2022 bear market, I have a different read. Liquidity fragmentation isn’t a real problem. It is a manufactured narrative used to justify launching more infrastructure products. The same underlying capital is sliced into smaller and smaller pieces, each with a new token, a new bridge, a new TVL dashboard. That isn’t scaling. That’s slicing already-scarce liquidity into fragments. The gold of the system is not fragmenting; the attention is. A fragmented market is not an opportunity for more building. It is an opportunity for more intermediaries.
The capital-hungry cycle will produce the same story at the macroeconomic scale. The response to fragmentation is always more financial engineering. Need to connect fragmented suppliers to fragmented demand? Build a new layer. Need to allocate capital across new infrastructure assets? Create a new fund vehicle. Need to manage risk across jurisdictions? Introduce a new derivative. Each new layer is defensible on its own. But the aggregate effect is not necessarily a deeper, more resilient capital market. It is a longer chain of handoffs, and every handoff is a place where verification fails.
Notice what the market is not building. There is no global standard for verified infrastructure reporting. There is no protocol that lets a bank prove, in real time, that a construction milestone has been met. There is no clean way to verify that the chips bought for a subsidy program are actually installed in the servers funded by the subsidy. There are thousands of analysts modeling these flows on spreadsheets. There are settlement and clearing systems that still operate on a T+2 basis in a world where AI models generate new business plans every second. That mismatch is the real fragility of the capital supercycle.
When I say “verify,” I mean it at the protocol level. Not a PDF signed by an executive. Not a sustainability report reviewed by a consultant. A proof with a machine-checkable output. The math behind the proof doesn’t care about the investment committee. Math doesn’t negotiate. That is why cryptographic verification is the only tool that can scale beyond the speed of human trust. If the next infrastructure boom is built on thirty-year assumptions, it will need to be verified faster than human beings can read. ZK proofs, secure enclaves, verifiable inference, and on-chain settlement are not token-holder tools. They are macroeconomic instruments.
Let me be explicit about the counterintuition. Most people think of verification as a cost. The point is the opposite. In a capital-hungry cycle, verification is a pricing advantage. A project that can prove its own progress will attract capital at a lower cost than a project that can only present a PowerPoint. A lender that can monitor a construction site through verified data flows will charge a lower risk premium. A sovereign fund that can check the provenance of every dollar before it leaves the balance sheet will not waste a crisis on bad counterparties. The cost of trust is the hidden line item in every capital allocation. This cycle will make that line item large enough to see.
I built a prototype in 2026 that used a zero-knowledge circuit to prove that an AI model’s output was generated without tampering. The proof covered both the model weights and the input dataset. This was for a leading AI lab, and the result was published as a whitepaper on “Verifiable Inference.” The reason a macro analyst should care is not because of the AI trend. It is because AI models are now making loan decisions, supply chain forecasts, and power grid optimizations. If the inputs to those models cannot be verified, then every downstream financing decision is built on an unverified foundation. The next capital cycle is not just an infrastructure story. It is an inference story. We are not just building physical machines. We are building machines that decide where the capital goes. If the decision engine is a black box, the capital market is a confidence game.
This is where the Goldman report gets dangerous. The report is not wrong about scale. It is wrong about control. By framing the cycle as capital-hungry, it centers the problem on capital supply. That framing has a political consequence. It invites a policy response that maximizes capital deployment at the expense of verification. It tells institutions to move faster, to write bigger checks, to enter new exotic asset classes, to trust the emerging special purpose vehicles that promise enormous returns. The pressure to deploy is enormous. Fund managers are measured on allocation. The market rewards the ones who commit, not the ones who verify. That is not a critique of Goldman. It is a critique of the incentive structure in finance.
But incentives are not immutable. If a critical mass of capital allocators starts treating verification as a screening criterion, the structure flips. The borrowers who can produce verified, machine-checkable claims will get the cheapest capital. The borrowers who cannot will get nothing. That is how the cycle eventually cleans itself. It does not happen through regulation alone. It happens through the plumbing. It happens because the cost of being wrong in a capital-hungry cycle is too high to be paid in trust. It happens because code is law, but bugs are reality, and the next crisis will not be caused by a bug in a smart contract. It will be caused by a bug in the underlying assumption that a lender’s model represents a real asset.
So what is the takeaway for a reader trying to survive the buildout? Stop counting how many billions are being announced. Start counting how many of those billions can be audited. When you hear “capital intensity,” ask not who is supplying the capital. Ask who is verifying the claim. When you hear “infrastructure renaissance,” ask whether the financial plumbing is being rebuilt at the same speed as the physical grid. When you hear “finance sector growth,” ask whether that growth is in intermediary fees or in machine-checkable proof.
The most capital-hungry cycle in history will not end with a shortage of capital. Capital is abundant. It will end with a shortage of receivers. I don’t mean receivers in the bankruptcy sense, though there will be many of those. I mean receivers in the signal-processing sense. The infrastructure that receives a claim and determines whether the claim is true. The one who builds that infrastructure owns the cycle. The one who only brings a bigger checkbook will be the exit liquidity.