In this issue:
- The Elasticity Business—Over long periods, it's astonishing how much the elasticity of supply and demand can change. Indispensable goods become free, impossible-to-create ones get manufactured, and elasticity is only limited by effort, ingenuity, and capital.
- Jensen!?—The case for either keeping him out or giving him super-voting stock; if he can't be fired, make it official.
- Elasticity, Illustrated—China can reduce global emissions, but not because the CCP cares about emissions.
- Hiring—AI makes it easy to mass-apply for jobs. But we've already dealt with that; email did it, too.
- Strategy—A company that briefly simplified retail investors' access to Bitcoin is now over-complicating it.
- Opsec and Romance—Wealthy single people are prudently paranoid about their personal lives.
Talk to this post on Read.Haus.
The Elasticity Business
In economics we talk about the concept of "elasticity," for both supply and demand: when prices go up, do people stop buying that product, or just pay more for it and either reduce spending elsewhere or draw down savings? And similarly, how do suppliers react to pricing changes? Do they build more of whatever the product in question is—more fabs, more power plants, more ML researchers, more natural gas rigs? Or do they just relax and enjoy supernormal profits that they expect to come down on their own?
When you talk about elasticity, you shouldn't treat it as a constant with respect to either space or time. On a sufficiently short timescale, elasticity can be nearly zero because there isn't time to react. Consider a case like being in line at Starbucks. Even if you're willing to pay 10% or 20% or 50% more than the customer in front of you, you're still going to stay in line. You face extremely inelastic supply in that you could get yourself to the front of the line, maybe by paying the person in front of you to switch places. But you need to offer enough to overcome the intrinsic weirdness of the situation, but not so much that you make it really weird. But wait a minute and you're fine.
Similarly, on long timescales, few things are truly inelastic unless they're n-of-1 by definition, like a work of art. Land sometimes gets used as an example of an inelastic resource. As the old joke that, like everything attributed to "Mark Twain" and not to a specific work by Mark Twain, was not created by Mark Twain goes, "Buy land. They're not making it anymore." But the Dutch beg to differ; about a fifth of the land area of the Netherlands was marshland or literally underwater and subsequently reclaimed. Dutch land was economically productive; ports meant that they could specialize in value-added processing of imports (initially, importing wheat to feed cows so they could export dairy products, and later, much more). If land is worth enough, a sufficiently commercial country can expand by conquering nature rather than its neighbors.[1]
And that works in the abstract. If you'd been writing about elasticity in the mid-19th century, and wanted to appeal to a nerdy audience, you could talk about nitrogen. We need it to fertilize plants, and we need more of it than we can get naturally. There are some sources of bulk nitrogen; the late 19th century's answer to the first Gulf War was when Chile fought a war with Bolivia and Peru over control of nitrate deposits, including guano. But it turns out—very fortunately for us—that fixed nitrogen can be extracted from the air, though it's energy-intensive. Depending on how you quantify things, this was the most significant invention in human history; half of the nitrogen in your body is the result of the Haber-Bosch process, and, as a rough shorthand, that means half the world's calories and thus around half the world's population. It's very fortunate for us that the available supply of a useful element turns out, at some cost, to be elastic.
Some companies, and some industry trends, work because they shift elasticity. Fracking is one of the most powerful examples. Oil prices have historically been volatile because the fixed cost of finding oil is both high and unpredictable, while the marginal cost of pumping another barrel can be quite low. Finding another Ghawar is effectively impossible. (Venezuela officially has more reserves, but that doesn't come from particularly trustworthy officials). Because the marginal cost of getting another barrel out is pretty low, it's hard to find a circumstance where Saudi Arabia will stop pumping.[2] So, at $11/barrel, they pump. At $147, they pump a little more. But not enough to get it back to $11.
Fracking has a faster decline curve for individual wells, which means that for a fracker to keep producing, they have to keep investing. And that means that they're adding elasticity to the market. When prices decline, they don't spend as much to replace production (or they do, and eventually can't raise enough money to keep doing so before too long). When prices rise, they don't have to underwrite hydrocarbon demand two decades from now, just a few years out. They don't eliminate the cyclicality of the industry by any means, but they add new dimensions to it through capital markets: if investors demand that cash be returned to shareholders rather than reinvested, they'll restrain supply, at least until the point that shareholders start to believe that they ought to grow again. Since the volatility of oil and gas prices erodes energy companies' returns—when times are too good, they have to overpay to return capital, and when times are bad, they have to overpay for capital—they're basically making the industry more stable in the aggregate in exchange for being the first to take advantage of higher prices or adjust to lower ones.
Cloud computing businesses are, in a way, in the business of creating ways for their customers to express different levels of demand elasticity over different time scales. You can buy spot instances if you're insensitive to price spikes or willing to go without; if you absolutely need some baseline level of service, you can commit to that and get it. Having that customer mix means that they can hit many different points in the multidimensional space of future demand for various kinds of compute—you can buy low latency with a long lead time or vice-versa. And this means that customers are, collectively, sketching out the world's demand curve for different kinds of hardware at different points in time. Since hardware businesses are also notoriously cyclical—much money has been made over the years shorting memory businesses at single-digit P/Es and buying them back when they're reporting losses—this is transmitting some stability to the business.
Economic abstractions are valuable analytical tools for understanding broad phenomena, but they don't give you much practical guidance day to day. In fact, the business of selling elasticity is valuable in part because making it happen is so concretely specific and annoying. If you'd told George Mitchell all about elasticity and the potential rewards for tweaking it, he would have gotten bored, rolled his eyes at you, and gone back to trying to figure out the optimal viscosity of fracking fluid, and he would have been right. These kinds of high-level models tell you a lot about the general shape of the economy and why it's hard to predict, but they also tell you that the specifics are much more specific than that.
Disclosure: long AMZN.
Excessively firm estimates of supply elasticity get smuggled into discussions about whether or not price gouging is good. Some people oppose price gouging on the grounds that the supply of important goods during a natural disaster is perfectly inelastic. Very well. They should, for consistencies' sake, also oppose Vote Gouging and Gratitude Gouging, where some completely pedestrian activity like giving someone a bottle of water somehow makes you a hero just because they were dying of thirst after an earthquake or hurricane. To the extent that price gouging is bad because no amount of money can make products materialize where they're desperately needed and infrastructure is lacking, we should privatize the loss from people trying to do it anyway. But what matters is not what the average person thinks they could do during a disaster, but what the most determined and clever person would do if offered a substantial reward in exchange for moving a pallet of X to GPS coordinates Y, Z as soon as possible. So opposition to price gouging gets stronger the more ineffectual the person making the argument is, and there are some people who are very, very strongly opposed to gouging. ↩︎
Not that this is a complete escape from the phenomenon of variable elasticity. Low oil prices don't make Saudi Aramco a money-losing proposition; they do that to Saudi Arabia, which has a generous official and de facto welfare state and various military interests that allow it to, expensively, maintain the government's legitimacy. There is, at least theoretically, an oil price low enough that Saudi Aramco could profitably pump oil, but the state can't protect either the oil or the proceeds. That's a Black Swan scenario. So was $150/barrel from the perspective of 2002 or -$40 from that of 2014. ↩︎
You're on the free list for The Diff. Last week, paying subscribers read about the gap between guidance companies give the market and projections they give acquirers ($), when and why it might make sense to make ads for AIs (or, equivalently, when it might make sense for AI agents to pay attention to ads) ($), and how thoroughly vindicated Richard Posner has been ($). Upgrade today for full access.
Diff Jobs
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- Ex-Palantir team re-routing $500bn of home & auto insurance premium through AI assistants is hiring two roles: an orchestration engineer to coordinate agents across thousands of discrete problems, and a growth engineer to make them the default tool for insurance. If you’ve been building tools and infra for other companies but now want to be in the driving seat, this is for you. (NYC)
- a16z, Sequoia, Thrive-backed next generation pharma company that instruments the best drug hunters in the world with deterministic software, data and AI to accelerate the process of acquiring, developing, and monetizing promising therapeutics (three active immunology programs) is looking for senior software and data engineers. If you are excited about sublinear headcount scaling and building technology that drives therapeutic abundance, please reach out. (NYC, Boston, SF)
- Ex-Anduril, Ex-Abnormal Security, Ex-Bridgewater, fast growing startup building frontier agentic cybersecurity to make the long tail of businesses defense dominant via MSPs is looking for a senior machine learning / quantitative researcher. Working in low data environments or working in low signal-to-noise environments e.g, mid-frequency quant finance is a big plus. Evidence of data rigor required. (SF)
- High-growth startup building dev tools for wrangling and debugging complex codebases is looking for someone who can personally execute the SaaS bear case: review the third-party software they use and figure out what to keep, what to drop, and what to implement in-house. (SF, DC)
- Series A startup that powers 2 of the 3 frontier labs’ coding agents with the highest quality SFT and RLVR data pipelines is looking for growth/ops folks to help customers improve the underlying intelligence and usefulness of their models by scaling data quality and quantity. If you read axRiv, but also love playing strategy games, this one is for you. (SF)
Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.
If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.
And: we're now actively deploying capital into early-stage companies through Anomaly. Our focus is on defense, logistics, robotics, and energy. If you'd like to chat, please reach out.
Elsewhere
Jensen!?
In between the decades when nothing happens and the weeks when decades happen, there are the weekends when at least a couple quarters'—maybe even a years'—worth of narrative shift happens. There's a reason "Merger Monday" exists, and why bailouts tend to get announced at the start of the week: it's a stretch during which markets aren't open so there's less incentive to leak and less distracting feedback about what does leak, and also when important people are less distracted at work because everyone else has gone home. So, there's a petition to reinstate Jensen Huang as CEO of Nvidia, claiming 168k TK signatures, including around 20% of Nvidia's employees (here's more on how that came about ($, WSJ): behind every surprisingly viral new site is one or more engineers desperately discovering annoying new bottlenecks), and he's talking about coming back ($, FT). As The Diff noted last week ($), Nvidia doesn't have a formal system to give him majority control, but he's done a great job of being informally in command by having so many direct reports, handling so many supplier negotiations himself, and, perhaps, having so many handshake deals and tacit understandings with Nvidia's partners. Strictly speaking, it's hard to definitively say he's "irreplaceable" because we don't even know what he would be replacing!
Bringing him back would solve some problems: neoclouds would worry less about their financing, and the rest of the chip supply chain would be more willing to invest. On the other hand, if he has sixty direct reports and is dealing with a government investigation of what he knew about where his chips went, Nvidia will have a CEO who's busier than ever and who has been granted the equivalent of temporary turbo-super-voting stock. But all that is contingent on the stock going up! If he comes back and the system is too complicated to manage, or it turns out that too much of Nvidia's revenue comes from customers it pretends not to know about because it's not allowed to sell to them, then he's out again.
So, the end debate is about transparency in corporate governance. It's more honest to just give a CEO voting control; if the board is going to reinstate him, they should find some way to give him extra votes that grant him disproportionate but not total control simply because that reflects reality and he has more of an incentive to delegate. But the board should also consider a broader kind of responsibility: if he does come back, the message to every CEO is that the more imperial their management style, the more de facto control they have. And if the way to keep control is to keep delivering numbers, that gives them an incentive to deliver the numbers at all costs: if he comes back, and thinks he can pull it off, his incentive to get Chinese labs access to chips, whether by lobbying or looking the other way, is stronger than ever.
(As a style note, the dramatis personae includes two major Huangs who aren't related, not to mention two kids and a son-in-law who work there. But Mr. Huang is too formal for The Diff's uncompromising, but also unwritten, house style. So we must call him "Jenson," as if we're buddies or something. I regret that I, too, join the ranks of people who feel betrayed by this whole sordid situation. And it is worth noting that a CEO's extramarital affairs, while not strictly something that will affect EPS next quarter, do reflect poorly on one's character and in particular one's ability to handle power without being tempted to misuse it. There are, of course, plenty of people who manage to be terrible in many respects while remaining faithful to their spouses, but it should still update you in the direction of wanting checks and scrutiny that you might not otherwise ask for.)
Elasticity, Illustrated
People who are serious about climate change have one of two options: learn about geoengineering or learn Mandarin. The US and Western Europe have both worried about emissions for years, and done something about it; US CO2 emissions per capita, for example, are down to pre-[First World]war levels. Growth in emissions comes from catch-up growth in populous countries that are not experiencing any sort of catch-up policy prioritization. A good illustration of this is that China's aggressive reduction in oil use after the Iran war will actually lead to a decline in global emissions this year ($, Economist). But that cause illustrates how far the policy gap is; as it turns out, China had the option to reduce year-over-year emissions, but when they did it, it was a side effect of a completely unrelated decision involving their actual priority of economic stability.
Hiring
John Hempton has a good piece on how AI has made it harder for him to hire: it's easier to spam people with semi-customized cold emails (though if you get enough of these, you'll recognize the tells), so companies face a higher cost of filtering. This will be more of a temporary adjustment than a long-term trend, for the simple reason that we've been through exactly this before. Pre-email, it was much more costly and inconvenient to print a cover letter, address it to the correct company, and mail it to them along with a résumé, and doing so didn't automatically create a paper trail making it trivial to keep track of where one had or hadn't applied. Email lowered the cost of spamming job applications—so it raised the return on writing a customized cover letter. To worry about AI having an indefinite negative effect on hiring (separately from other labor displacement concerns) is to claim that no one will come up with a verifiable proof-of-work. But that implies that it's not a hiring problem at all—it means that the AI can do all the work! In the meantime, changes in the transaction cost of hiring are an opportunity for clever junior people to find a way to distinguish themselves before they're at a point where they can get warm introductions.
Strategy
The company once known as MicroStrategy pivoted a few years ago to a simple strategy: issue stock, use it to buy Bitcoin, and benefit from public investors' willingness to pay more for Bitcoin if it had a stock ticker and they could own it in their Fidelity account. Since then, that simple business proposition has been replicated endlessly, to the point that there's now a glut of Bitcoin plus a ticker minus some SG&A, and they have complicated their strategy in response (FT). When crypto was rising, investors liked levered crypto exposure even more than the vanilla kind, so the company, now Strategy, started issuing more complex financial instruments. But lenders don't like it when a company has minimal income and significant debt service, so they also started raising cash. So Microstrategy has gone full circle: it used to own a portfolio of cryptocurrency, which was worth a premium to fiat investors. Now they have some fiat liquidity, some crypto exposure, and a lot of extra paperwork, which is a bundle that sounds more like it deserves a discount.
Disclosure: short a little MSTR.
Opsec and Romance
In light of Jensen's personal and business struggles, the WSJ has a look at high-end matchmakers for wealthy executives ($). There's a mutually-illegible power asymmetry here, in that they are richer and more influential than their dates, which means that their potential liability from a bad first date is a lot higher, even if they aren't up to anything nefarious. "I had a relationship with someone, and here is the story of why I didn't like it; I'll give him a pseudonym and limit my disclosure of personal details to his job, appearance, hometown, hobbies, birthdate, mother's maiden name, and last four digits of his social security number" is an established genre of confessional writing. And some people just use Instagram. The modern information environment makes it hard for people to sequence things: it's possible to get rich and established professional without getting established personally, and it's hard for people to navigate their personal lives after that happens. The more you expect information availability to rise, the bigger a priority you should put on finishing up your dating life and getting settled earlier in life.
Comments load as you approach this section.