본문 바로가기
Economics × Management Theory Weekly

Oct 25 – Oct 31, 2026

Korean original

Not investment advice — these are resources for learning and forming your own view.

One tension ran through this week: the price of borrowing shows no sign of falling, while the places that need money keep multiplying. On the macro side, the Fed held rates even as inflation stirred again, and it gave no hint about December. Gold drifted around $4,000, and long-term yields firmed upward alongside fiscal worries. The difficulty is that much of this inflation comes from the production side, through oil, freight and labor costs, so higher rates do little to suppress it. On the growth side, AI remains the only growth story, but that story now runs on debt rather than equity. Google’s $51 billion in spending sent its shares down 10%, Meta’s outlook was cut, and talk that Anthropic might fail to win an investment-grade rating after listing shook tech stocks across the board. A third force, organization and governance, sits on top of both. The 95-year-old Weil Gotshal is moving toward dissolution as its people leave, and Trade Desk used super-voting shares to reprice executive stock options. Together they show how fast a company can melt when its assets are people and trust rather than buildings. In short, the market this week faced two constraints, cost and financing, tightening at once. Next week, watch whether Big Tech capex guidance and the jobs data tighten those constraints further or ease them.

Deep dive — one theory Operations

It is not a shortage of money but a shortage of places to put it: rereading the AI investment race as a bottleneck problem

Theory
Theory of Constraints (TOC)
Origin
Eliyahu Goldratt, in his 1984 novel The Goal.
In short
Every system has at least one point that constrains its performance, and total output cannot exceed what that point can handle. The sharp implication of this simple proposition, which Goldratt set out in a novel, comes next. However hard one works on improving anything other than the bottleneck, total output does not rise at all. This is where our intuition that effort brings improvement breaks down.
Limits
The lens works poorly in systems where the bottleneck keeps shifting or where many substitute routes exist. When several bottlenecks bind at the same time, the simple prescription also fails to hold.

The AI industry is a chain in which each stage's output is the next stage's input. Chips are made, data centers are built to house them, power is brought in, and cloud revenue follows. It is hard to find an industry with stronger dependencies between stages. The narrowest points in this chain are sites, power connections, transformers and skilled construction labor. Spending more money cannot widen them within a few quarters. That fits the condition the Theory of Constraints assumes: a constrained resource that cannot be expanded in the short term. So the question to ask of the AI sector now is not who has invested more, but whether the money has gone into widening the bottleneck. This is exactly where the intuition that more effort yields more output breaks down.

Through this lens, Microsoft's quarter reads very differently. Revenue rose 19% to $92.3 billion, yet capital expenditure came in below expectations, and the company itself pointed to delays in data center construction as the reason. This is closer to an admission that the places to put the money are not yet ready than a sign that there is no money to spend. In the same week SK Hynix posted results well above market expectations on AI memory, and Google committed $51 billion. But if the bottleneck sits in server halls and power, the chips and cash poured in at the front of the chain cannot yet be converted into revenue. They pile up at the front as inventory. Google's 10% share price fall and the downgrade of Meta's credit outlook on cash-flow concerns can be read as the market belatedly pricing in a time lag: spending goes out now, but returns arrive only after the bottleneck is cleared. The proposition that no amount of effort outside the bottleneck raises total output by even a step showed up this week as earnings reports and share prices.

The generalization alumni can take away is simple. When evaluating any growth story, ask first where the narrowest point of the system is today, not how much has been invested. The speed at which that point widens is the speed at which the industry can actually grow. Bottlenecks do not stay put. For a time it was GPUs, now it is power and construction, and next it may be skilled operating staff. So finding the bottleneck once is not enough; the work must be repeated every quarter. The lens also has clear limits. Where several bottlenecks bind at once, or where many workarounds exist, a prescription that assumes a single constraint works poorly. When reading Big Tech's capex guidance and any remarks on power and construction lead times next week, underline where they say things are blocked rather than how large the numbers are.

Source story: Microsoft blames data centre delays for lower capital expenditure

Cases by layer 6

Macro environment · Monetary Policy Transmission Mechanism

Fed holds fire on interest rates despite intensifying inflation

The Fed held rates even as inflationary pressure strengthened again, and only reaffirmed its resolve to contain prices. It gave no clear signal on whether it would raise rates in December. It also said it would cut the number of monetary policy meetings over the coming year from eight to seven. The market read this combination as a source of greater volatility in long-term yields and turned toward real assets such as gold and commodities.

The policy rate itself did not move, but expectations for December were left open, so a change in expectations clearly exists. This week those expectations spread straight into Treasury yields, the gold price and the mortgage rate debate. That shows the premise still holds that policy reaches the real economy through several channels. The decision to cut the number of meetings in particular narrows the very window through which the central bank can send signals. It bears directly on this lens's question: the same policy can produce different results depending on whether a channel is open or shut.

This news usually gets read as a one-line story: 'the Fed held with a hawkish tone.' But once we ask not what level rates were set at, but through which path and with what delay the decision reaches households and firms, one picture explains why long-term yields moved despite a hold, why the home-buying debate flared again, and why gold is holding at $4,000.

Through this lens, the hold was not a decision to do nothing but a decision that loaded the entire burden onto the expectations channel. If short-term rates are pinned and signals are withheld, the market fills in the blank itself, so volatility rises in long-term yields. That volatility spreads to corporate bond spreads and mortgage rates, and reaches firms' funding costs and households' housing decisions with a lag. What makes it harder is that much of this inflation came from the supply side, through oil and freight costs. Interest rates can suppress the willingness to buy, but they cannot restart a stalled supply.

Takeaway — Companies need the habit of working backward from the rate figure in the policy statement to the channel through which that decision reaches their own funding costs. Firms that hold both floating-rate borrowing and debt maturing next year, in particular, can use the hold period to buy time to rework their funding structure.

Financial Times

Macro environment · Aggregate Demand - Aggregate Supply (AD-AS)

Chocolate Price Spook Families for Halloween

News arrived that Halloween candy and chocolate prices were up about 50 cents from a year earlier, driven by higher labor, energy and transport costs. In the same week Brent crude stayed near its elevated wartime levels despite higher Gulf exports. Reports tracking gasoline prices also traced how consumer-level energy costs spread into logistics and freight costs.

The object of analysis is not one item, chocolate, but the economy-wide price trend in which common inputs, namely labor, energy and transport, are layered onto many consumer goods at once. The source of the shock is also not consumers suddenly buying more candy but rising costs on the side that makes and moves it, so the conditions of a supply-side shock are plainly met. Firms also cannot pass higher costs on all at once and are raising prices by 50 cents in steps. That fact itself supports the premise that prices move stiffly in the short run.

This news usually gets consumed as a household story: 'holiday prices are up again.' But if we first distinguish whether a shock came from the buying side or the making side, we can explain at once why this price rise reads not as a sign of a boom but as a precursor of weaker consumption, and why a prescription of rate hikes feels awkward.

Through this lens, the price rise was not caused by hot demand but pushed up by costs. The result is an uncomfortable combination in which prices rise while volumes and margins are squeezed together. When consumers' disposable income is pressed, spending outside necessities falls first. Amazon's lowered year-end guidance and the shift of funds from tech stocks into consumer staples can be read as phenomena with the same root. Airlines, by contrast, could raise fares and pass fuel costs on, so they protected profitability even as passenger numbers fell. In the end, who held the power to pass costs on decided who won.

Takeaway — In a cost-push inflation phase, check the cost pass-through rate before revenue growth. A business without that power will not see margins return even if interest rates come down. The choice that matters most in this phase is to direct investment and resources toward those with both a justification for raising prices and the brand to support it.

Bloomberg Markets

Industry · Porter's Five Forces

SK Hynix third-quarter profit sharply beats estimates on AI memory boost

SK Hynix reported third-quarter results well above market expectations, on the back of AI memory demand. Demand for high-performance AI memory lifted earnings expectations across the industry, and sentiment toward semiconductor equipment and back-end processing improved with it. In the same week, Big Tech shares wobbled on delays in data center investment and worries about cash burn.

The unit of analysis is not SK Hynix's quarterly results alone but the position a memory supplier holds within the AI infrastructure chain. Only a handful of firms can make high-bandwidth memory, and expanding capacity takes years, so new entrants find it hard to get in. On both counts the conditions fit this framework, which takes a snapshot of industry structure. Although Big Tech buyers hold the demand, the supplier side's profits jumped more. That fact itself is evidence that, among the five forces, supplier bargaining power is working unusually strongly.

This news is usually read as good news for one company: 'Hynix benefited greatly from AI.' But through a lens that asks not which firm is capable but which link in the chain captures the profit, one picture explains why, in the same AI boom, memory makers collect profit while the Big Tech buyers of that memory see their shares fall on cash-burn worries.

Through this lens, the profits of the AI boom are not spread evenly along the chain. They pool in the links that have no substitutes and are closed to new entrants. Those who run models face fierce competition and relatively easy switching, so their margins thin. The high-performance memory used on top of them has nothing that can be swapped in right away, so it holds the upper hand on both price and volume. This picture has a short shelf life, however. Once capacity expansion is complete or a substitute technology arrives, the bargaining power of the same link could cool quickly.

Takeaway — When investing in or attaching a business to the AI theme, look not for 'where demand is growing' but for 'where there is no substitute when demand grows.' Buyers, conversely, protect their costs by securing long-term contracts and multiple suppliers before the current supplier advantage hardens.

CNBC Markets

Industry · Moral Hazard / Adverse Selection

In an AI blow-up, would insurance pay out?

As fears of AI-related overheating and a systemic debt crisis grew, trading in the credit default swap market picked up again, and the question 'If AI blows up, will insurance pay?' was raised in public. In the same week, sharp outflows were observed in the high-risk debt market, and the large debt exposure of private-equity-backed companies was also discussed.

Between those who fund AI infrastructure and those who use the money to build data centers and run models, there is a large information gap about how much a project will actually earn. A large part of the losses can be passed to third parties through credit default swaps, insurance and private lending structures, so the condition of loss shifting is clearly met. A buyer of protection may borrow more aggressively, and the seller cannot see the counterparty's true risk. This is the terrain the theory addresses.

This news is usually read as a sentiment gauge: 'the market has started to worry about an AI bubble.' But through a lens that examines a structure in which information sits on one side only and losses pass to others, the problem turns out to lie not in the size of the bubble but in a design that separates those who bear risk from those who create it.

Through this lens, the revival of CDS trading is both a thermometer of anxiety and, paradoxically, a device that makes still larger borrowing possible. Because protection can be bought, lenders also keep their standards looser. The result is an accumulation of low-quality debt, such as the outlook that Anthropic might not win an investment-grade rating even after listing, or the debt warnings on private-equity portfolio companies. When it becomes hard to tell good borrowers from bad ones, the market sets an average price. Healthy borrowers then leave the market first, and this week's scene repeats: the riskiest debt melts away first.

Takeaway — When reviewing AI exposure, check not only who earns the money but who absorbs the loss if things go wrong, down to the level of the contract. The mere fact that there are many protective devices is not grounds for comfort. It may be a signal that risk is being moved elsewhere and piling up.

The Economist (Finance)

Company · Resource-Based View (RBV)

What killed Weil Gotshal?

Weil Gotshal, a flagship Wall Street law firm with a 95-year history and 1,200 lawyers, has reached the point of discussing dissolving its partnership after a chain of departures by key people. The incoming leader, due to take office in January, gave up the post citing a talent crisis. Management took restructuring advice throughout the weekend in an attempt to stabilize the firm. In the end, the leadership put a proposal to the partners to dissolve the firm.

This firm's competitiveness lay not in factories or patents but in the client relationships that particular partners built over many years, their experience in difficult advisory work, and the organizational culture that held those people together. Resources of this kind can be neither bought in the market nor copied, so the premise that resource heterogeneity does not easily disappear fits exactly. Yet those very resources took the form of people who could walk out on their own two feet. That is why this view, which takes the individual firm as the unit of analysis and looks inside it, explains the most.

This news is usually consumed as an incident story: 'a law firm collapsed through management failure.' But through a lens that asks where the competitive advantage resided, we can explain why 95 years of reputation and a headcount of 1,200 did not serve as a shield within five days, and how quickly assets that do not appear on the books can evaporate in a knowledge-based industry.

Through this lens, Weil's collapse happened not because profitability worsened but because the resources that generated profit did not belong to the firm. When a key partner leaves, the clients and deal pipeline he brought leave with him. Other partners who see the departure move in turn. That chain makes the firm's cohesion collapse in a self-reinforcing way. The warning of this view that yesterday's strength becomes tomorrow's shackle can also be read here: a structure that grew by relying on individual partners' reputations became hostage to those same individuals.

Takeaway — An organization whose people are its assets needs not only evaluation and pay design but also a design that attaches client relationships to the organization rather than to individuals. When considering an acquisition or investment, it is also practical to add the question 'can this company's value resign?' to the due diligence list.

Financial Times

Company · Agency Theory

How Trade Desk Rescued Executives’ Stock Options From a Dive in the Share Price

Trade Desk CEO Jeff Green used super-voting shares to approve a repricing of executive stock options over the objections of outside investors. Because the exercise price was lowered after the share price fell sharply, a governance dispute arose over how gains and losses are divided between shareholders and management. The market took it as a source of added short-term share price volatility.

Ownership and management are separate, yet voting power is concentrated in the founder, so the decision could not be blocked even though shareholders objected. The beneficiaries of the decision were the executives, including the decision-maker himself. Both points expose the classic setup in which the two sides' aims diverge. Outside investors have no way to verify from the outside whether the option repricing is really meant to retain talent or to avoid accountability, so the condition of imperfect observation and verification is also met.

This news usually drifts into a moral criticism: 'the founder abused governance.' But through a lens that treats agency costs as something to be managed, not eliminated, the key point is structural, not about who is a bad person. The contract design was built from the start so that when the share price falls, only shareholders bear the loss and management does not.

Through this lens, the option repricing deepens, after the fact, the asymmetry in pay for performance. If executives share in the gains when the share price rises, and have their exercise price lowered to compensate when it falls, the pay loses its original function of putting them in the same boat as shareholders. What is left is only an incentive to take on risk. Super-voting shares act here as an amplifier that makes the outcome unchangeable even when monitoring costs are paid. That is why the only response open to investors is to sell the shares, and that is how volatility grows.

Takeaway — When assessing a growth stock, the voting structure and pay provisions are as important a due diligence item as the financial statements. In particular, read in advance how those provisions work in a down market. Executives, for their part, should choose designs that carry a cost, such as performance-conditioned grants, over repricing, so that the message sent to the market is not distorted.

WSJ Markets

Other signals this week 18

Based on 179 items over 7 days

Past issues