Sep 20 – Sep 26, 2026
English translation of the Korean original, prepared with AI assistance. Korean original
Not investment advice — these are resources for learning and forming your own view.
Deep dive — one theory Organization
Why a Founder's Stock-Backed Loan Is 'Governance News,' Not 'Personal Finance'
- Origin
- Michael Jensen and William Meckling, 1976 paper.
- In short
- The moment a principal delegates work to an agent, costs arise, because the agent's interests never fully match the principal's, and the principal cannot fully observe what the agent actually does. Monitoring costs, bonding costs and a residual loss inevitably follow -- and Jensen and Meckling's central point was not that these costs can be eliminated, but that they must be managed.
- Limits
- It has been criticized for assuming that managers will shirk whenever they get the chance; stewardship theory sits at the opposite end of that spectrum. Tying compensation to performance, meanwhile, breeds short-termism focused on hitting the numbers, so the fix creates a new problem of its own.
Agency theory is useful right now because, at a moment when higher rates have made borrowing expensive, an executive's personal financial position quietly seeps into corporate decision-making. Shareholders want executives to grow the company's value, but they cannot fully observe what executives are actually weighing when they decide, and that information gap is precisely what generates monitoring costs and residual losses. When a founder is also the controlling shareholder, it's tempting to assume ownership and management interests automatically align -- but that assumption wobbles the moment personal debt enters the picture. The higher rates go, the heavier the interest burden and margin-call pressure on collateralized loans become, which makes this exactly the moment to reach for this lens.
News this week that Larry Ellison pledged another 67 million Oracle shares as collateral, taking his total pledge to $9.2 billion, looks on the surface like a story about a wealthy individual's personal financing. But from an agency perspective, it ties the top decision-maker's personal liquidity directly to the company's own stock price, so that if the share price falls below a certain level, he faces pressure to post more collateral or sell shares. That in turn makes choices that prop up the short-term stock price -- the timing of capital spending, the tone of disclosures, the intensity of the AI growth narrative -- personally advantageous for management as well, and it is very hard for outsiders to verify whether that incentive points the same way as shareholders' long-term interests or in a different direction. Worse, if a forced sale actually happens, the sale itself pushes the stock price down further, passing volatility unrelated to the company's actual performance on to shareholders.
The takeaway for classmates is this: raising stock compensation to align managers' and shareholders' interests is the standard playbook, but the moment that same stock is pledged as collateral for a personal loan, the alignment of 'rising and falling together' flips into the asymmetry of 'if it falls, I'm the one who gets squeezed first.' So investors should track disclosed pledge ratios and their trends as closely as earnings, and board members are better off setting collateral limits or pre-approval requirements in advance. Jensen and Meckling's point that agency costs cannot be eliminated, only managed, sounded unusually practical this week, in a high-rate environment.
Source story: Ellison Pledges $9.2 Billion More in Oracle Shares for Loans
Cases by layer 6
US bond sell-off pushes long-term yields to highest since 2004
A continued sell-off in US Treasuries pushed long-term rates to their highest level since 2004. In the same week, the 30-year fixed mortgage rate surged to 7.45%, and Wall Street expects the government to issue $1 trillion in short-term debt to ease its funding burden.
Discounted cash flow is a framework for converting future earnings into today's value by discounting them for time and risk, and this week's events fit that textbook case almost perfectly: the discount rate itself was reset in the market to its highest level in 20 years. Long-term Treasury yields form the floor on which every asset's required return is built, so when that floor moves, everything from mortgage rates to corporate borrowing costs gets pushed up with it -- exactly what we saw in this week's 7.45% mortgage rate and SoftBank's 9.75% bond issue.
This news is often read as a bond-market-only story: 'bonds sold off, so rates rose.' But through the discounting lens, it's not just a bond-market event -- it's the price tag on every asset class, including stocks, real estate and private equity, being rewritten at once, with the essential feature being an asymmetry: assets that pay off further in the future get discounted more heavily. That's why, given the same rate rise, AI growth stocks hurt more than steady dividend payers.
Through this lens, this week's market wasn't shaken by deteriorating earnings outlooks; the numerator (future cash) held steady while only the denominator (the discount rate) grew, cutting valuations. The same shock shows up wearing different faces -- shrinking demand for homebuilders, valuation compression for growth stocks, a spike in funding costs for highly leveraged companies -- but the root is the same. KB Home lowering its margin outlook and bond investors moving to offset losses through tax strategies both read as branches of that same trunk.
Takeaway — When reviewing your portfolio, rank holdings first by when each asset's cash actually arrives, rather than by whether the stock looks good or bad. Just grasping the fact that assets whose cash is further out swing more sharply with a single rate move explains the pattern of gains and losses in this phase.
Financial Times
Refineries Are Now the Main Chokepoint for Global Energy Supplies
Overlapping geopolitical conflicts, including Iran and the war in Ukraine, exposed how heavily the world's diesel supply depends on refining capacity in the Middle East and Russia. Refining capacity has become a tighter bottleneck than crude oil itself, and some in the US have even argued for considering a ban on diesel exports.
For the aggregate demand-and-supply framework to apply, you need to identify which side the shock came from, and this is a textbook supply-side shock: it's not that buying power suddenly surged, but that refining capacity -- the producing side -- was choked off by war and attacks. And because diesel is an input woven into agriculture and transport costs across the board, its effects don't stop at one industry; they hit prices and output economy-wide, which meets the condition for treating the whole economy, not a single market, as the unit of analysis.
This news is often read as a sector story: 'oil prices are up, so energy stocks should do well.' But through the aggregate demand-supply lens, the real story is an uncomfortable combination -- falling output alongside rising prices, a force pushing the economy toward stagflation -- which narrows the central bank's room to respond with rate policy. That's the thread connecting why long-term rates spiked and Asian bonds sold off on inflation worries in the same week.
Through this lens, this week's energy news isn't just one more inflation indicator -- it's an event that narrows monetary policy's own set of options. Rates can only suppress buying power; they can't restart a refinery that's shut down, so the tighter the central bank squeezes to control prices, the more likely already-weakened housing and consumption crack first. Oil prices easing late in the week on US-Iran diplomatic progress relieved some of that pressure, but as long as the bottleneck sits in the equipment itself, that relief can easily reverse.
Takeaway — When reading inflation news, ask first which side the shock came from, rather than focusing on the size of the number. If the inflation is supply-driven, factor in that cyclical assets face a double squeeze as expectations of rate hikes grow.
WSJ Markets
What It’s Like to Work in One of America’s Data Centers
Data centers are being built at scale across the United States for AI infrastructure, driving massive power consumption and industrial expansion. The surge in power demand is lifting long-term profit outlooks for regional utilities and pushing up demand for specialized cooling and power equipment as well.
The theory of constraints is most powerful in chains where one stage's output becomes the next stage's input and total throughput is capped by the narrowest point. The AI industry's chain runs from chip design to manufacturing, data center construction, and finally power and cooling, and what this news shows is that the bottleneck has narrowed not around money or chips but around things that can't be scaled up quickly, like power grids, transformers and cooling equipment -- exactly matching the condition that the constrained resource can't be easily expanded.
This news is often read as a growth story: 'AI demand is growing, so data centers are multiplying.' But through the bottleneck lens, the question flips, arriving at the conclusion that AI's real output is set not by model performance but by how much power a company has secured. The point that pouring capital into anywhere other than the bottleneck won't raise total throughput lands especially sharply right now, when everyone is talking only about chips.
Through this lens, the next phase of the AI investment cycle turns not on model competition but on the race to secure power. A real gap in operating capacity will open up between operators who locked in power contracts early and those who didn't, and in the middle, utilities and transmission, distribution and cooling equipment makers gain bargaining power -- in Porter's terms, this is also a phase where supplier power rises. Capital flowing toward physical equipment, like Caterpillar's autonomous industrial machinery or Bezos's space investments, can be read in the same context.
Takeaway — When evaluating AI-related companies, check how much power and land they've secured rather than what they've announced. Money flowing to whoever holds the bottleneck is an old rule of chain-structured industries.
WSJ Markets
Meta's standoff with Amazon over Muse could be a sign of things to come
Meta's AI agent Muse racked up 2.5 million downloads within 13 days of launch, rapidly gaining users, and a platform clash broke out when Amazon blocked the app's distribution on its own site. In the same week, Shopify announced it would integrate Muse into its e-commerce checkout.
Network effects hold when value to remaining users grows as more people join and complementary products follow, and with Muse this week we actually saw that loop in action: more users give commerce partners like Shopify a reason to plug in, and having a partner attracts still more users. Amazon's decision to block distribution is itself an attempt to cut off a rival's user growth before it crosses critical mass, satisfying the condition that active competition is under way.
This news is often read as another conflict story: 'big tech is fighting again.' But through the network lens, what's really at stake isn't one app but the new gatekeeper position an AI agent occupies between consumers and commerce, and whoever fills that position first ends up brokering all the transactions that follow. Since the contest turns not on which feature is better but on whether people are already there, Amazon's block reads not as an irrational tantrum but as a calculated defensive move.
Through this lens, this clash signals that the AI ecosystem is splitting into rival camps rather than converging on a single open standard. If it's easy for users to run several agents at once, the pull toward any one camp weakens and no one becomes a clear winner; but once hard-to-move data like payment and purchase history starts accumulating inside an agent, switching costs emerge and that pull hardens. So the thing to watch next isn't download counts but the share of downloads that actually convert to checkout.
Takeaway — When assessing platform competition, look first at how hard it is to leave, rather than user counts. User numbers with no switching costs attached are just a provisional metric that can drain away at any time.
CNBC Markets
SoftBank pays steep price for record bond sale to fund OpenAI bet
SoftBank issued a record-setting more than $11 billion in high-yield corporate bonds at rates of up to 9.75% to fund its investment in OpenAI. With long-term rates at a 20-year high, the cost of raising high-risk credit has climbed sharply.
For signaling theory to apply, there must be an information gap between sender and receiver, and the signal itself must actually be costly -- and the conditions clearly hold here, since SoftBank chose to signal that it knows more about OpenAI's prospects than the market does by taking on a real interest burden of 9.75%. Saying 'we're confident' out loud and raising money while accepting nearly $1 billion a year in interest are two very different costs to mimic.
This news is often read as a risk warning: 'betting on AI with excessive borrowing.' But through the signaling lens, it explains exactly why the bond was priced this way at this moment -- the more expensive the funding, the larger the conviction the act conveys. At the same time, the theory delivers a cold warning of its own: a signal being costly doesn't guarantee the underlying judgment is correct, and once everyone is pouring money into signaling contests, society as a whole is left with nothing but waste.
Through this lens, SoftBank's bond issue is both fundraising and a message aimed at the entire AI camp. It effectively sets an entry barrier -- telling rivals this is the cost of playing this game -- and puts SoftBank in a stronger position in OpenAI valuation talks; read alongside Meta launching hardware first with Google and OpenAI following, it amounts to the start of a contest to buy conviction at a premium. But if rates keep rising or AI monetization is delayed, this signal simply becomes a repayment burden.
Takeaway — When reading a company's fundraising disclosure, look beyond the amount raised to how much it cost to raise it. Expensive funding is both evidence of conviction and, at the same time, an irreversible commitment.
Financial Times
Paramount settles lawsuits, allowing for Warner Bros. deal
Paramount Skydance cleared the way to proceed with its $110 billion acquisition of Warner Bros. Discovery by settling lawsuits brought by attorneys general from 12 states and the writers' union. In Los Angeles, however, politicians and workers remain focused on ways to soften the merger's negative impact.
For this framework to apply, there needs to be a value path that only exists through the combination, and the premium paid must be measurable against the expected synergies -- and combining the two studios has a clear path, pooling content libraries and streaming subscribers to spread fixed costs, with a concrete $110 billion price tag attached. The cost of settling the lawsuits and the process of merging staff and organizations after the deal also directly illustrate the difficulty of integration execution, satisfying the condition that execution capability is a real variable.
This news is often read as procedural progress: 'the legal hurdles are gone, so the deal moves forward.' But through the lens of synergy and the winner's curse, the real question isn't whether the deal closes but who captures the larger pie -- the buyer or the seller -- and the old pattern that the highest bidder in an auction is usually the one who misjudged the most could well repeat itself here. The objections from workers and politicians against the merger read not as emotional backlash but as an early signal of integration costs to come.
Through this lens, this deal's success will hinge not on the announced synergy figures but on cutting redundant departments and merging the two organizations' production cultures. Synergy estimates always skew optimistic, and conditions attached during negotiation -- like the lawsuit settlements and employment-retention promises -- eat into that optimism; set alongside Barry Diller's withdrawal of his MGM Resorts takeover bid in the same week, it shows that not overbidding is itself a strategy.
Takeaway — When you come across major M&A news, write the announced synergy figure down side by side with the premium paid. The simple math that a narrower gap between the two leaves less for the acquirer's shareholders gets forgotten surprisingly often.
Bloomberg Markets
Other signals this week 18
- Wall Street Week | China’s Trade Reckoning, Syracuse’s Second Chance, Humanoids in Healthcare Bloomberg Markets
- Xi Urges AI Cooperation WIth US Bloomberg Markets
- Trump presses Zelenskyy to stop hitting Russian refineries Financial Times
- Lula Plans $28 Billion Debt Buyout to Aid Consumers Before Vote Bloomberg Markets
- The timely tax strategy that can help bond investors make the best of runaway yields and big losses CNBC Markets
- Here's who we know is going to the Trump-Xi dinner so far CNBC Markets
- Insurance Australia Says Greensill Settlement Isn’t Finalized Bloomberg Markets
- 30-year fixed mortgage rate jumps sharply Thursday to 7.45% CNBC Markets
- OpenAI ‘agent’ hacked an Australian health service website Financial Times
- Reform Should Cut Back Britain’s State Pension, Think Tank Says Bloomberg Markets
- Vietnam’s Leader Declares US Trade Deal Close Bloomberg Markets
- L.A. Comes to Grips With a Studio Merger Many Didn’t Want WSJ Markets
- Here are the 3 big things we're watching in the stock market this week CNBC Markets
- Mark Zuckerberg debuts $1,299 Meta VR Glasses and Muse Charm pendant as part of AI agent push CNBC Markets
- Meta's Muse AI agent downloads are surging. Here's how it compares to ChatGPT, Grok and Claude CNBC Markets
Based on 179 items over 7 days