Oct 18 – Oct 24, 2026
Not investment advice — these are resources for learning and forming your own view.
Deep dive — one theory Strategy
If AI compute is 'the new oil', where does the money end up?
- Origin
- Michael Porter, 1979 HBR article and Competitive Strategy (1980).
- In short
- Whether an industry makes good money is decided by its structure, not by how capable its companies are, and Porter summarised that structure as five forces: how fiercely competitors fight, how easily new companies can enter, whether substitutes exist, and whether suppliers and customers have the power to raise or cut prices. The stronger these forces, the thinner the profits for everyone in the industry.
- Limits
- The snapshot works well only when an industry's boundaries are clear, so the five forces cannot capture places such as platforms that join several markets at once, or markets where complements shake up the game. Because it is a picture taken at one moment, its shelf life is short in industries whose boundaries are melting fast.
Porter argued that whether an industry makes money is decided by its structure, not by how clever the companies in it are. Few places fit that claim as well as AI infrastructure does now. Until recently AI's boundaries were too blurred to capture in a snapshot of the five forces. But talk of buying and selling compute by the hour, and even hedging it with derivatives, means compute is shifting from equipment each company procures for itself to a standardised traded good. At that moment the industry's boundaries and the power at each stage finally become visible. As with oil, standardisation and futures markets make prices transparent, but transparent prices act first to squeeze the margins of the middle stages. That is why now is a good moment to take a snapshot of the structure.
Viewed through this lens, this week's news forms a fairly clear picture. Upstream suppliers of high-performance chips, power and data centre equipment still have strong pricing power, with demand queuing up: Western Digital posted quarterly net profit of $1.78 billion, and Anthropic is reportedly discussing a second multibillion-dollar deal with Nscale. By contrast, the so-called neoclouds caught in the middle face thin barriers to entry, since anyone with capital can come in. And once compute trades as a standard product, they lose any basis for saying 'it has to be our GPUs', so substitutability rises too. CoreWeave's funding troubles freezing the whole AI junk bond market and Solidigm shelving its IPO are the bill for those thin margins. Customers, too, are concentrated in a few large model companies, giving them strong power to push prices down. If four of the five forces press down on the middle stage, there is no reason the money should stay on neocloud income statements, however large AI demand becomes.
The generalisation for alumni is this. In a new industry, identifying where the bottleneck lies along the route that sells 'the new oil' is far closer to investment returns than identifying what the new oil is. The more explosive an industry's demand, the more that demand is sucked into the one or two stages with the greatest bargaining power. But Porter's snapshot has a shelf life: if power constraints ease or chip supply diversifies, the bottleneck quickly moves to another box, so you should not hold the same picture for years. In practice, it is safer to drop the habit of treating AI as one theme, split it into suppliers, intermediaries and end customers, and re-score each box's bargaining power every quarter.
Source story: AI compute is the new oil—and needs a new market
Cases by layer 6
Oil hits $117 for first time since April on fears of flare-up in Iran hostilities
Rising fears of hostile conflict with Iran pushed oil to $117 a barrel for the first time since April. In the same week Brent swung around $109 on news of Ukrainian strikes on energy facilities and attacks on tankers in the Middle East, and Houthi attacks even led the US to withdraw some staff from Saudi Arabia. Earnings estimates for the energy sector rose, while the bond market came under upward pressure on yields.
The premise of this lens holds exactly, because the source of this oil price spike is clear. Prices jumped not because stronger demand is consuming more crude, but because the 'producing and shipping side', the Strait of Hormuz and Saudi oilfields, came under threat. On top of that came measures that narrow supply, such as Niger banning diesel and petrol exports outright. Shocks like this are not the kind where prices and wages all adjust within a few months; they become visible only when the unit of analysis is the whole economy's output and prices, not individual oilfields.
This news is usually read as a sector profit-and-loss calculation: good for energy stocks, bad for transport and consumer goods. But a framework that separates demand-side booms from supply-side shocks changes the question. If rising prices come from a strong economy, the central bank can push them down. If they come from blocked oilfields, raising rates however far only drags the economy into the worst combination: falling output with inflation that stays. That is the essence.
Through this lens, this week's oil price rise does not end with one line about 'upgraded energy sector earnings'. It reads as an event that adds another layer of transport and raw material costs on top of an already accumulated burden of essential spending, as in the news that US household health insurance premiums have passed $28,000. With output squeezed and prices rising, the central bank has to give up one of the two, and that dilemma reaches prices through rising long-term yields and valuation pressure on richly valued growth stocks.
Takeaway — When you read an oil headline, classify first whether demand pulled prices up or supply was blocked, rather than how much they rose. Then you can gauge in advance which way rates and equities will move on the same number. In a supply-shock phase, it is safer to look at inflation-hedge assets together with companies that have pricing power.
Financial Times
Japan core inflation in September surges to its highest this year at 2.1%
Japan's core inflation rate for September reached 2.1%, the year's high. The market is again raising expectations of a shift in the Bank of Japan's policy stance, while over the same period the Japanese government is steering pension funds and households to absorb $120 billion of new government bonds. The direction of the yen and Japanese government bond yields hangs on this.
The situation sits squarely in the phase this theory addresses: the policy rate itself has not moved, yet bond yields and the yen are reacting to the mere expectation that it will. At the same time, the government's effort to place $120 billion of bonds with pension funds and households means a large volume sits right in the middle of the channel from the policy rate to market rates. If domestic institutions hold up as buyers even as rates rise, long-term yields will not climb as much as expected, and the time it takes to reach the real economy changes too.
This news is usually consumed as a one-line forecast: 'inflation is above 2%, so the Bank of Japan will hike soon'. But this lens, which traces the intermediate channels, brings a more important question to light. In an economy like Japan's, where households hold more deposits than loans and companies sit on cash, a rate rise works first through the exchange-rate channel rather than the lending channel. In the end, the real effect of policy depends not on the size of the hike but on which channel is open.
Through this lens, the Bank of Japan's next move reads less as a step to cool domestic consumption than as a choice to support the yen and reduce pressure coming in through import prices. Oil at $117 is directly pushing up prices in a country that imports all of its energy. If so, what moves on announcement day is not consumption or capital investment but the yen-dollar exchange rate and long-term bond yields, and it will take several quarters for traces to appear in real-economy indicators.
Takeaway — If you hold Japanese assets, checking government bond auction results and the yen's reaction comes closer to your real profit and loss than the rate number in the policy statement. Remember that monetary policy is transmitted through channels, not declarations.
CNBC Markets
CoreWeave Woes Risk Shutting Down AI Junk Market: Credit Weekly
As CoreWeave's finances wobbled, doubts spread about the funding environment for AI start-ups as a whole, and junk bond investors are cutting the money they supply to new tech companies. In the same week Solidigm withdrew its IPO and credit risk indicators for neocloud providers rose. The risk premium on bonds issued by weaker-credit companies is rising across the board.
The actual utilisation of AI infrastructure companies and the quality of their long-term contracts are a classic asymmetry: management knows, bond investors find it hard to know. The size of that asymmetry shows in how a single piece of news about one company's funding trouble overturned the market's view of the whole sector, while key variables such as the collateral value of GPUs and customers' ability to honour contracts remain unverified from outside. And because the funding channels are junk bonds and IPOs, those who bear the losses are separate from those who decide the risk. The conditions the theory requires are all in place.
This news is usually read as a story about sentiment: 'the first crack in the AI bubble'. But through a lens on how information asymmetry breaks markets, the problem is not that investors got scared. It is that there is no way to tell good companies from bad, so everyone is charged the same high rate. The essence, then, is a vicious cycle in which financially sound companies leave the market first and the average quality of the remaining issuers gets worse.
Through this lens, the dividing line in the AI infrastructure funding market will become not growth forecasts but 'can you prove your condition at a high cost to yourself'. Funding channels are likely to stay open only for companies with mechanisms that weak companies find hard to imitate, such as disclosed long-term forward contracts, payment guarantees from large customers or asset-backed structures. Conversely, smaller issuers without such signals will be pushed out not through higher rates but by losing access altogether.
Takeaway — When approaching AI credit, look first at whether the contract structure can be verified, rather than at the coupon. In a phase of widening spreads, weeding out lemons matters more for returns than picking cheap bonds.
Bloomberg Markets
Eight Months of War Turns China Petrochemical Glut Into a Crunch
With crude supply constrained by the US-Iran conflict, China is rapidly running down the surplus petrochemical inventories it had built up, and glut is turning into shortage. Industrial materials that were in oversupply only recently are now objects of a scramble. Expectations of rising prices across industrial materials are forming.
This chain is ideally set up for amplification. In a multi-stage chain running from crude to naphtha, petrochemical intermediates and finished products, each stage sees only orders from the stage just ahead of it, not end consumers, and it takes months from order to delivery. The evidence: there is no sign that end demand changed sharply, yet glut flipped to shortage within just a few months as each stage made the rational decision to build inventory ahead of a long war.
This news is usually interpreted as 'Chinese demand has revived' or 'the war has created a commodity supercycle'. But to those who understand how the whip cracks, much of the current shortage may be a shift in inventory positions rather than demand. If so, the key point is that the inventory cycle, more than war news, will decide how long this lasts.
Through this lens, the current strength in industrial material prices splits into two layers. The cost component that oil has actually raised will last, but the component created by the stockpiling race will reverse just as sharply the moment fears of supply disruption ease. If suppliers mistake current order volumes for real demand when deciding to expand output, the ones left holding excess inventory a few quarters later will be the capital investors at the very top of the chain.
Takeaway — When you see an earnings surprise from a commodity-related stock, check once more whether rising shipments move in the same direction as end-consumption indicators. The sound at the end of the whip is always louder than the movement of the wrist.
Bloomberg Markets
Tesla and SpaceX Formed Committees to Discuss Potential Merger, Elon Musk Says
Elon Musk said committees have been set up at Tesla and SpaceX to discuss a potential merger of the two companies. In the same week SpaceX presented a long-term revenue outlook of up to $500 billion a year in its results, while Tesla, with car revenue shrinking, lowered its sales targets and is restructuring around AI. The market is focused on the possibility of vertical integration spanning AI, space and autonomous driving.
The value path that exists only if the two companies combine is clear on paper: putting autonomous-driving compute, satellite communications and rocket manufacturing under one roof offers quite a lot to share. At the same time, the premium Tesla shareholders pay depends on how SpaceX, an unlisted company, is valued, and that premium can be compared with the expected synergies. And because effectively the same person leads both sides of the deal, integration and price-setting are not separated. All of this framework's premises are in play.
This news is usually read as the narrative of 'a genius executive building a giant tech empire', and consumed as fuel for both share prices. But through a lens that asks who gets the pie rather than how big the combined pie is, the key is not whether the two companies grow by merging. It is who values the unlisted asset, on what basis, and from which shareholders to which shareholders value is transferred. The very fact that synergy numbers are traded without that question being answered is a warning sign.
Through this lens, the merger talks risk becoming a deal that covers the slowing growth and valuation burden of Tesla's car business with SpaceX's $500 billion long-term story. The independence of price-setting is further clouded by the fact that the two companies' shares are already entangled in collateral structures, as in the news that Larry Ellison borrowed against his Tesla and SpaceX stakes. Until the committees' independence and valuation basis are disclosed, it is more rational to question the distribution structure before the size of the synergies.
Takeaway — In big merger news, the first number to look for is not the synergy estimate but the premium paid and who did the valuation. The rule of thumb that the highest bidder is usually the one who misjudged most is especially dangerous in deals where the founder controls both sides.
WSJ Markets
FICO Holds $10 Pricing for Mortgage Lenders Amid New Competition
FICO decided to keep the price of the credit scores it supplies to mortgage lenders at $10 per score. That is more than double the price of the rival VantageScore 4.0. For lenders, higher operating costs could put downward pressure on the sector's margins.
That FICO holds its price even though a substitute exists at less than half the cost means the company sees demand for its product as highly insensitive to price. The $10 is a tiny share of the total cost of a mortgage, and securitisation processes and guarantor standards are built around FICO scores, so lenders have neither the time nor the freedom to switch right away. The conditions under which raising the price (or not cutting it) does not reduce sales are in place almost exactly as in a textbook.
This news is usually read as a story of intensifying competition: 'a cheap rival has arrived, so FICO will soon cut its price'. But a lens that asks how much less you sell when you raise the price shows that FICO's ability to keep a price twice as high makes whether switching actually happens far more important than the mere appearance of a rival. At the same time, with the Trump administration moving to lower refinancing costs, the real issue becomes where this fixed fee will be passed on.
Through this lens, FICO holding its price reads not as arrogance but as the result of an accurate calculation of its product's low price sensitivity. The $10 burden therefore flows not into lower revenue for FICO but into higher costs for lenders and, ultimately, fees for borrowers. However, sensitivity grows the longer customers have to switch, so if a regulatory change lets guarantors officially accept multiple scores, past inelasticity will no longer be a reliable number.
Takeaway — When choosing companies with pricing power, look at 'how long it takes customers to switch' rather than market share. A price that holds only while regulation or standards block switching will collapse all at once the day that regulation changes.
WSJ Markets
Other signals this week 18
- Trump moves to block certification of Alaska Senate election if Republican loses Financial Times
- Protests across France spiral into riots and looting Financial Times
- Moody's holds steady on France's credit rating as country grapples with turmoil and higher yields CNBC Markets
- Pulp Maker Mercer Is Weighing Possible Chapter 11 Bankruptcy Bloomberg Markets
- Capital One's strong quarter and progressing Discover integration are rewarded on Wall Street CNBC Markets
- Euro Falls as Russia Risk, France Protests Dent Trader Sentiment Bloomberg Markets
- US Treasury proposes tighter rules on investing in sensitive Chinese tech Financial Times
- Amazon warns new cyber flaws detected using AI are being ‘broadly exploited’ Financial Times
- Tradeweb to Buy Trumid For About $1.97 Billion WSJ Markets
- Anthropic commits 5% of shares to new public benefit fund that could address AI job disruption CNBC Markets
- Australia’s Victoria State Approves $700 Million Gas Project Bloomberg Markets
- Latest Oil Market News and Analysis for Oct. 20 Bloomberg Markets
- Wealth Tax Debate, Midterm Investing, AI Anxiety Bloomberg Markets
- Latest Oil Market News and Analysis for Oct. 21 Bloomberg Markets
- Tesla beats on earnings and sales even as auto revenue declines CNBC Markets
- Five Takeaways From SpaceX’s Earnings Call WSJ Markets
- Mynt’s Record IPO Brings a New Heavyweight to Philippine Stocks Bloomberg Markets
- SoftBank in Early Talks to Buy Allied Universal WSJ Markets
Based on 178 items over 7 days