Sep 13 – Sep 19, 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 Operations
AI's Bottleneck Isn't GPUs -- It's Power and Permits: What Oracle's $18 Billion in Debt Reveals
- Origin
- Eli Goldratt, The Goal (1984 novel).
- In short
- Every system has at least one point that limits its performance, and total output can never exceed what that point can handle. The sharp implication behind this simple proposition, which Goldratt worked out through fiction, comes next: fixing anything that isn't the bottleneck, no matter how hard you work at it, doesn't raise total output at all -- which breaks our intuition that effort translates proportionally into improvement.
- Limits
- This lens doesn't work well in systems where the bottleneck keeps shifting around or where many workarounds exist, and the simple prescription doesn't hold when there are multiple bottlenecks at once either.
The theory of constraints is useful right now because the AI investment race has stopped being a question of who spends the most money and become a question of where the money spent gets stuck on its way to becoming actual compute. A data center is a classic sequential process -- from securing land to permitting, substation equipment, power supply and server delivery -- where each stage's output feeds the next, so if one link jams, no amount of optimizing the rest raises total output. And because power grids and local permitting aren't resources you can expand within months just by throwing more money at them, the condition that they're hard to substitute in the short term holds exactly. Where chip supply once bottlenecked the industry, power and land have now taken its place as the new constraint.
Layer this week's news over that lens and the picture sharpens. News that Oracle's $18 billion data center debt plan is under strain from permitting delays and community opposition in New Mexico looks like a financing problem, but what's actually shaking is the timing of when that debt turns into a revenue-generating asset -- and the interest keeps accruing in the meantime. Amazon locking up backup power supply by securing warrants in Generac in the same week, sending that stock up more than 40%, is evidence that the market is already pricing not 'compute capacity' but 'when and where you can secure electricity.' When OpenAI says it will burn $280 billion by 2030, how achievable that figure is depends not just on capital markets' generosity but on the pace of substation buildouts and local community hearings. That the bottleneck has moved from semiconductors to power and land is the most practical piece of information this week offered.
There are two general lessons for classmates. One is that in any industry, once money starts pouring in, the constraint is guaranteed to move one step further down the chain, so picking this year's investments based on last year's bottleneck already puts you a beat behind. The other is that the asset holding the bottleneck is usually the most boring, most regulated one -- unglamorous resources like power infrastructure or permitting capability end up holding the bargaining power instead. That said, this prescription doesn't stay simple in systems where the constraint sits in several places at once or where there are many workarounds, so the first step is to check whether the bottleneck actually converges on one point.
Source story: Oracle’s $18bn data centre debt under strain amid local pushback
Cases by layer 6
Oil Executives Say the Great Fuel Crisis Is Here
Oil industry executives, including Chevron's Mike Wirth, warned that the global fuel supply shortage isn't a temporary phenomenon. In the same week, a key Saudi pipeline was shut down by an attack, closing off the Hormuz workaround, and US diesel prices climbed to record levels, hitting the transport sector directly.
This shock isn't about any single refiner's earnings -- it spreads by truck and rail into input costs across every industry, so the condition that the unit of analysis is the whole economy, not a single market, is naturally met. Its origin is clear too: fuel prices didn't rise because consumption suddenly jumped, but because a pipeline was cut and strait negotiations were delayed, choking off the producing and shipping side. And with companies struggling to set prices because they can't immediately pass higher costs through to sale prices, the premise that short-run prices are sticky also holds.
This news is often read as sector good news -- rising profit estimates for energy. But through the aggregate demand-and-supply lens, a far more uncomfortable fact emerges: the same rise in prices looks completely different depending on its source -- if it comes from stronger buying power, output and employment improve together, but if it comes from a block on the producing side, you get rising prices alongside falling output, and this week's oil price is clearly the latter. This lens makes visible that good news for energy stocks and bad news for the broader economy are two faces of the same event.
Through this lens, diesel and LNG prices handing developing Asia a $7 billion extra bill, and Thailand straining to hold its growth target, aren't separate news items -- they're the same supply shock rippling out in the same shape across multiple countries. Once the combination of rising prices and falling growth outlooks sets in, corporate profit estimates get cut on the margin side before the revenue side, and the more energy-intensive the industry, the faster that happens.
Takeaway — Don't read a higher energy weighting in your portfolio purely as improved performance -- check whether the same cause is eating into the margins of your other holdings too. Supply-driven inflation is the kind of pressure policy can't easily reverse until time resolves it.
WSJ Markets
Fed defies Trump with first rate rise since 2023
Citing the need to curb inflation, the Fed raised rates for the first time since 2023, putting it on a direct collision course with former President Trump, who has demanded rates below 1%. The dollar surged immediately afterward, Asian government bond prices fell, and the 10-year US Treasury yield briefly topped 5% before pulling back.
This is a phase where both a change in the policy rate and expectations around it exist together, and this week's data captured every trace of that change actually moving through the channels of market rates and exchange rates. The dollar index jumping, Asian short-term bonds falling and long bonds wobbling around the 5% line show that financial intermediation is working without a hitch; the effect just hasn't shown up in prices yet, because reaching consumption and investment still takes time.
This news is often consumed as an independence drama -- 'the Fed stood up to political pressure.' But the transmission-mechanism lens changes the question. What matters isn't the size of the hike but which channels are open and what sits at the end of them, and if this round of inflation originates in pipelines and straits, rates can only suppress demand -- they can't reopen a shut-down oilfield. Behind the standoff between politics and the central bank lies the more troubling fact that the tool and the problem live in different places.
Through this lens, the fastest effect of this hike shows up not in prices but in exchange rates and asset prices, following a sequence where dollar strength first cuts translated returns on emerging-market-currency assets, then pushes up the discount rate on growth stocks. Real demand cooling enough to tame prices is a matter of several quarters out, so if the supply shock persists in the meantime, the Fed carries the burden of tightening further on top of an already-tight stance.
Takeaway — Cultivate the habit of watching which channel a rate decision flows through and how fast, rather than just the decision itself. In particular, during dollar-strength periods, factor in at the allocation stage that returns on foreign assets can struggle even to break even.
Financial Times
American Businesses Have No Idea How to Set Prices Right Now
Reports emerged that US companies are struggling to set prices because they can't tell how long elevated energy costs will last. Costs have already risen with no clear sense of when they'll come back down, leaving companies caught between worrying about demand if they raise prices and eating into margins if they hold the line. In the same week, Macerich's CEO pointed to a pattern of selective consumer spending -- people still spending, but concentrated in specific categories.
Whether higher costs actually pass through to consumer prices is the biggest worry for companies right now, putting the question of price pass-through squarely at the center of this story. The observation that consumers are still spending, but selectively, is direct evidence of how sensitive demand has become, and because the sales-volume gap between companies that raised prices and those that held off shows up immediately, there's room to isolate that sensitivity.
This news is often summed up as a macro story about wavering inflation expectations, but the price-elasticity lens shows the same cost increase ending very differently for different companies. Sellers of goods with plenty of substitutes and no real need see sales drop the moment they raise prices, forcing them to absorb the cost, while those holding hard-to-replace goods can quietly pass it on. This is where winners and losers split apart behind a single line of inflation statistics.
Through this lens, this quarter's key earnings variable isn't revenue growth but pass-through rate. The same oil-price environment produces opposite margins depending on whether a logistics company has shipper contracts that let it pass diesel cost increases through in freight rates, and whether a consumer-goods maker holds brand-loyal customers versus selling into a category where people shop purely on the price tag. The observation that spending is turning selective is a signal that this gap has already started to widen.
Takeaway — When evaluating a company, check first how much of a price increase customers will actually accept, rather than just how much costs rose. Previously measured sensitivity doesn't hold up well when prices jump across the board, though, so re-measure it using the most recent quarter's actual sales volumes.
WSJ Markets
Musk urges top AI labs, Chinese companies to test each other's models amid calls for slowdown
Elon Musk urged major AI labs and Chinese companies to cross-verify each other's models before release, and Amodei and Altman also joined in warning about the pace of development. Congress, however, went into recess without concrete legislation, and Nvidia's Jensen Huang distanced himself from the debate, calling the framing of innovation versus safety a false choice.
The players are narrowed down to a handful of frontier labs and chip suppliers, and it's a structure where if one slows down, the benefit passes straight to whoever didn't -- so each side's payoff is directly tied to the other's move. On top of that, Congress's failure to legislate leaves no referee to enforce any commitment, and with Chinese firms now in the game too, there's no real way to sanction a defection. Huang immediately taking a different stance is itself evidence that each player is reading the other's move and playing its best response.
This news is often read as tech leaders' confessions of conscience or as safety discourse, but through the game-theory lens, what's exposed is a structural problem, not a moral one. If everyone slowed down, regulatory risk and accidents would both fall and everyone would be better off -- but because whoever slows down alone falls behind, everyone is stuck at the point where no one can move first. This lens shows the call for mutual verification not as an expression of goodwill but as an attempt to find a device that binds everyone together.
Through this lens, a voluntary self-regulation agreement is a promise that breaks easily until verification and enforcement are attached to it, so what's more likely to actually increase is the cost of safety reporting and incident disclosure, not a real slowdown in development. OpenAI's string of disclosures about concerning model behavior and Anthropic's rollout of a development-pace measurement metric both read as moves to stake out position in this game, while the companies selling infrastructure are betting the pace won't slow.
Takeaway — When reading AI regulation news, check who loses what if they break the agreement, not just what was proposed. An agreement with no enforcement mechanism barely needs to be reflected in valuations.
CNBC Markets
OpenAI expects to burn $280bn by 2030
OpenAI said it expects $280 billion in negative cash flow by 2030, citing infrastructure buildout and pricing pressure. Over the same period, the company is weighing an additional funding round at a $1.2 trillion valuation ahead of an IPO, and has decided not to go public this year. Meanwhile, the 10-year Treasury yield briefly topped 5%.
This company's value comes almost entirely from cash-generating power beyond 2030, not from what it earns today, and the company itself laid out the path of cash burn to get there in hard numbers, giving us something concrete to weigh future flows against. The same holds on the discount-rate side: an environment where long-term Treasury yields touch 5% means the yardstick used to convert distant future money into today's value has moved up across the board.
This news is often consumed through the binary of 'is it an AI bubble or not,' but the discounted-cash-flow lens separates the numerator from the denominator. Burning $280 billion pushes the numerator further out, and a 5% rate inflates the denominator, and when both happen at once, today's value of the same business plan shrinks noticeably. In place of the word 'bubble,' what's left are two questions: how far out is the money, and how heavily is it being discounted?
Through this lens, Anthropic easing cash-burn worries by signaling two straight quarters of profitability, and OpenAI declaring losses through 2030, aren't just company-specific news -- they're the story of how, in a 5% rate environment, companies whose cash arrives sooner and those whose cash arrives later get treated completely differently. Even within AI, chips and power equipment that are generating revenue today hold up, while platforms whose cash payoff is far off see their valuations swing sharply with even a small further rise in rates.
Takeaway — When looking at growth stocks, check when cash actually starts coming in before you look at the growth rate, and the further out that point is, the more you should recalculate value under two or three rate scenarios. Also keep in mind that when applying this tool to assets that aren't yet profitable, the conclusion tends to be driven by whatever assumptions you plug in.
Financial Times
Samsung Faces Local Hedge Fund’s Push to Cancel Preferred Shares
A South Korean hedge fund directly demanded that Samsung Electronics buy back and cancel its preferred shares, pressing for improvements in shareholder value. The reasoning is that reducing the share count lifts earnings per share and book value per share; in the same week in the US, Jana Partners demanded a CEO change at Cooper.
Ownership and management are separated, and it's management that decides where to put excess cash, with outsiders unable to easily verify whether that choice matches what shareholders want -- and this situation carries that condition intact. The very fact that a shareholder singled out a specific capital-allocation method, preferred-share cancellation, and demanded it publicly is a scene where a mismatch in objectives that had gone unresolved internally gets pulled out into the open.
This news is often read as a hedge fund chasing a quick gain or shaking up management control, but the agency-theory lens reclassifies this kind of intervention as a costly monitoring activity. Monitoring has a price, and there's always a residual loss left over besides -- what matters isn't whether this cost can be eliminated, but who pays it and what changes as a result. The arrival of an activist is the moment governance costs get priced by the market.
Through this lens, whether this demand succeeds hinges less on whether the preferred shares actually get cancelled and more on whether management changes how it explains its use of excess cash going forward. The gap between companies where hedge-fund pressure works and companies where it doesn't is essentially the size of the governance discount, and a CEO-change demand in the US and a capital-policy demand in Korea landing in the same week reads as the market becoming willing to pay this monitoring cost.
Takeaway — When picking undervalued stocks, check not just business quality but whether shareholders have a channel to change management's capital allocation. Keep in mind, though, that tying compensation too tightly to performance invites number-chasing in turn, so remember that the fix can create a new problem of its own.
Bloomberg Markets
Other signals this week 18
- Trump says he's banning MS NOW, CNN and Politico from White House CNBC Markets
- Supreme Court rejects Trump bid to lift block on mail-in ballots rule CNBC Markets
- Toxic Fume Complaints Ground Planes From Airbus Venture WSJ Markets
- Gulf states postpone talks with Iran over Hormuz impasse Financial Times
- Trump says he still has confidence in Fed Chair Warsh, demands 1% or lower interest rates CNBC Markets
- Oaktree Sees Credit Investors ‘Finally Being Paid to Take Risk’ Bloomberg Markets
- 10-year Treasury yield hits 5% before reversing as traders await Fed meeting CNBC Markets
- Bloomberg This Weekend 09/12/2026 Bloomberg Markets
- How record diesel prices will rip through the U.S. economy. Trucks, rails are only the start CNBC Markets
- A $7 Billion Gas Bill Sees Developing Asian Nations Sour on LNG Bloomberg Markets
- Lakers Buyers Lay Out Plans to Reach $30 Billion Valuation in Pitch to Investors WSJ Markets
- The 40-Minute Phone Call That Left Ed Sheeran’s Tour Hanging by a Thread WSJ Markets
- EPA to Roll Back Power-Plant Emission Rules WSJ Markets
- Why Weil Gotshal’s Highest-Paid Partner Jumped to Cravath, Shocking Big Law WSJ Markets
- Holtec pulls IPO over ‘perfect storm’ in AI sector, founder says Financial Times
- OpenAI rules out IPO this year as Altman, Musk & Amodei warn AI is moving too fast CNBC Markets
- Jana Partners Pushes Contact-Lens Maker Cooper to Replace Its CEO WSJ Markets
- Anthropic tells investors it will be profitable for second straight quarter Financial Times
Based on 174 items over 7 days