Moderna's Bets: Moonshots and Platforms

Plus! The China AI Stack; Closing the Loop; Ads and Addiction; The Second-Order Costs of Inflation; Fraud and Transaction Costs

Moderna's Bets: Moonshots and Platforms

In early 2020, Moderna Inc. was putting together something they called a "stopwatch drill," where they'd choose a new disease and see just how quickly they could go from identifying it to producing a viable mRNA vaccine. The company was, at this point, a decade-old biotech with no products on the market, but they were confident they could use this as a demonstration of their ability to synthesize new drugs. Their target: Nipah virus. Just before starting the drill, their CEO went on vacation. And while he was on vacation, he stumbled on a newspaper story about confused public health officials in the city of Wuhan, who were trying to track down a new respiratory disease.

The rest is history, of course, but the history of how Moderna got to this point is quite interesting, and potentially a guide to what the future might hold for drug development.

The Startup and the Story

mRNA vaccines, Moderna's specialty, are just an incredibly cool idea. Tricking the body into mass-producing spike proteins that alert the immune system that there's a virus, but that don't actually include that virus, is a wonderful sci-fi idea that happened to be close to practical deployment at just the time when there was a great new use case. mRNA vaccines are not just a new and useful way to produce drugs, but a different way to think about them: Moderna likes to call RNA "the software of life," and their medications are indeed a clever kind of hack, telling cells in the body to produce specific proteins. It's a smart way to alter the supply chain of medicine: why introduce a new substance into the body, possibly continuously, when you can instead introduce new instructions and have the required substance produced just-in-time using locally-sourced materials? For Moderna, BioNTech, and several billion vaccine recipients globally, it's quite fortunate that cellular machinery works the way it does, and that this literally code injection exploit can take advantage of it.

Moderna was founded in 2010 to commercialize early research into mRNA. The concept had been around since the late 1980s, though going from a known mechanism of action to a practical way to get cells to express desired proteins, and then going from that to an FDA-approved product, was a long journey. The company hired its current chief executive, Stéphane Bancel, in 2011, who ran the company with a culture that, depending on who you ask, either enforced strict meritocracy, high expectations, and a fast pace of development or was arbitrary and abusive, leading to slow progress and high turnover.

The company went public in 2018. The S-1 is quite a read. Biology is complicated, and biotech companies are, too. Moderna's S-1 opens with a nice tutorial on mRNA in general and on their plans in particular, but quickly dives into an extreme level of detail.1 The body has lots of defenses against arbitrary code execution—which is, after all, what viruses like SARS-CoV-2 are doing—and that makes mRNA a tricky technology, and as a corollary, mRNA vaccines particularly impressive.

More so than just about any sector, biotech investors are underwriting a technology advance, not an existing business. There are plenty of Tesla analysts who didn't study electrical engineering and plenty of software analysts who can't code, but biotech seems to have more investors with "M.D." or "PhD" after their names than "CFA." Pharmaceuticals are a bit like the entertainment business, where there's a long lag between initial concept and revenue, and where distribution and marketing are a separate discipline from creating something worth selling. So, like entertainment, it's developed a structure where the big mature companies function partly as merchant banks, offering money and infrastructure to more specialized businesses. If you look at Moderna's financials at the time of the IPO, you do see revenue, but it's mostly from deals they have with larger pharma companies where they're collaborating on research.

The financing strategy for a company like this, particularly as a platform, is tricky: the end goal is to have a portfolio of drugs that generate cash flow, which can be plowed back into new research. One option is to bet everything on a single drug, which can work as a proof of concept. But if there are multiple lines of research, then a) the overall pace of development will be faster (an insight from manufacturing one product might apply to the rest, for example), and b) it's more likely that one of the candidates will clearly get close enough to commercialization to justify more fundraising at an attractive valuation. But a large portfolio is also, in cash flow terms, a big liability—drugs require escalating cash inputs as they pass through the FDA approval process. So a big and growing portfolio of drugs under consideration is also a bet of increasing conviction that the market will still want to underwrite the biotechs in general and that company in particular.

The Vaccine

At the time of their IPO, the Moderna bet that was closest to working as a standalone product was VEGF-A, a treatment for myocardial ischemia that was in Phase 2 trials. In their "Risk factors" section, the #1 risk is the need to get more funding, #2 is how their diversification across 21 different programs won't necessarily lead to any success, and #3 is:

No mRNA drug has been approved in this new potential category of medicines, and may never be approved as a result of efforts by others or us. mRNA drug development has substantial clinical development and regulatory risks due to the novel and unprecedented nature of this new category of medicines.

Obviously, a lot has changed, and fast. On a longer scale, the compression of timelines is truly incredible: about five centuries passed between the Black Death wiping out a third of Europe's population and the proposal of the germ theory of disease (it may have been 13 centuries). Historically, plagues have been stopped through some combination of quarantines and herd immunity. When Covid hit, the virus' entire source code was available online by January 10th, 2020. Moderna designed a vaccine three days later, and it was being tested in mice by February. The vaccine got its emergency use authorization on December 18th, 2020, and 800 million doses were shipped in 2021.

This is profound not just because it's a many-orders-of-magnitude improvement in the pace of developing treatments to emerging threats, but because there's a qualitative difference in treatment protocols. A pandemic doesn't think or plan, but it's still possible for it to be inside humanity's OODA loop. Because we're naturally bad at understanding compounding, and because testing only happened in response to concerns about the disease and was held back by some fairly egregious early mistakes, pandemic policy was generally a real-time response to a blurry snapshot of the pandemic in the recent past.2 The shorter the lag between a disease outbreak and a working vaccine, the more meaningful policy options there are in the meantime: there are costs that are bearable for a year but not a decade, and if there's a likely timeline and likely efficacy for a vaccine, future pandemic policy can at least be a more meaningful debate about tradeoffs.3

Moderna and BioNTech were both able to produce mRNA-based vaccines, but the companies are pursuing different strategies. BioNTech partnered with Pfizer for manufacturing and distribution, and has been more of an R&D and product development story. Moderna is aiming to be a full-stack business; they're hiring in marketing and sales.

The Platform

Just weeks after BioNTech/Pfizer announced that their candidate vaccine was 90% effective, Moderna announced that theirs was 94.5% effective. Interestingly, given the market's difficulty in pricing exactly what Covid's impact would be on the broader economy, the market mostly got this one right: from January 1st 2020 through the day before their vaccine announcement, Moderna shares rose 365%. The day of, they were up another 10%. Or, put another way, 89% of the market's response to the vaccine was a bet, in advance, that Moderna would be able to pull it off.

But part of Moderna's bet was that their company was not an effort to spread out research dollars and eventually luck into a blockbuster product. They argue that mRNA is a broadly applicable technology. From the S-1: "We believe the manufacturing requirements of different mRNA medicines are dramatically more similar than traditional recombinant protein-based drugs across a similarly diverse pipeline." The low initial cost for mRNA drug development creates a notable competitive dynamic: since drugs are expensive to produce, sometimes when a large company goes after a particular market it can scare away smaller competitors (or their funders). Splitting the market is a challenge. We saw this in the pandemic; not everyone tried to develop a Covid vaccine. But if the cost of trying is low enough, it can make sense for Moderna to pursue many different research avenues at once—and since they have the budget to advance to more expensive stages of research, they get to do the scaring-off at scale.

One of Moderna's choices was to focus on modified RNA, rather than unmodified. Modifying RNA is, in very brief terms, a way to get around the body's natural defenses. It's less relevant for vaccines, because those defenses tend to get stronger over time, but it's more meaningful for products that need repeat dosing, like cancer treatments. One former employee says that the internal joke is that Moderna is a delivery company; making mRNA is not the same thing as making the body use mRNA, and over-optimizing for being good at the latter means having more options for future treatments.

There are still technical limitations around the mRNA approach. Viruses already have selection pressure to have very simple genomes; the less genetic code the virus needs to copy to reproduce itself, the faster it reproduces and the more virions each infected cell can make. And mechanically, what an mRNA vaccine is doing is getting a small strand of genetic material, wrapped in a tiny lipid blob, into a cell so it can get to work. Moving a larger molecule means more chances for something to go wrong. The delivery technology is basically a way to affect the size of the genetic payload, and the maximum packet size sets a rough ceiling on the maximum complexity of what an mRNA drug can accomplish. This also affects things like flu vaccines, since a standard flu shot protects against multiple strains of flu, instead of against just one. But there are other tradeoffs to consider: if there's a particular strain of flu that catches us by surprise, Moderna would be in a better position to rapidly produce boosters targeting just that variant. The vaccine can be designed quickly, and it doesn't need to be grown in chicken eggs; it's more readily mass-produced.

The fun way to look at Moderna in 2022 is that it's a wildly exaggerated version of Moderna circa 2019: lots of promising products in the pipeline, lots of cash with which to pursue them, and a credible claim that each new drug they launch makes the next ones faster and cheaper to develop. Before Covid, Moderna was well-known for how much money it had raised; $2.6bn before the IPO, and a record-for-biotech $600m at IPO. Now the company has $8bn in cash and equivalents on hand, produced $9bn in free cash flow in the last four quarters, and actually has so much cash on hand that it's buying back stock—going from no products at IPO to buying back stock four years later must be some kind of record. They can't count on future FDA approvals being remotely as swift as they were in 2020, nor can they expect the next product to be a $20bn/year blockbuster. But now they can afford to wait.

Further reading: Peter Loftus' The Messenger is a great book about Moderna, covering both their early history and the development of the vaccine itself, and was a major source for this. Moderna is an extremely chatty company. In 2020, they had some kind of investor meeting every 7.4 business days, including an annual "Vaccine Day" event with lots of detail. And readers with Tegus access can see some interviews with former employees and competitors, which was an especially good source for understanding some of the differences between mRNA approaches, and where its use cases can be stronger or weaker.

A Word From Our Sponsors

The world's best companies have a secret when it comes to onboarding their customers. Spoiler: It's Flatfile.

Flawless data onboarding acts as a catalyst to get you and your customers where they need to go, faster. Get started today!

https://flatfile.slite.com/api/files/TcKLx57Hl24efw/The%20Diff%20(1).png

Elsewhere

The China AI Stack

One of my favorite stylized facts about China is that it's an ideal country for AI companies because it has both skilled labor for algorithm development and chip design and a large number of cheap workers who share the same language and can do data-labeling tasks for very little money. This Interconnected post looks at a case study, in which a small city in Inner Mongolia is becoming a nexus for AI development: it has government incentives, a climate well-suited to power-efficient datacenters, and affordable laborers.

Closing the Loop

Local venture ecosystems usually start with entrepreneurs who have a big exit early in their career, and can start writing checks to friends and friends-of-friends a few career stages behind them. As long as those people have good exits, capital keeps getting recycled into the ecosystem, and growth ensues. A16Z may be trying to formalize this process by hiring money managers who can run money for founders. In one sense, money is fungible, but in a dealflow-driven business, it's definitely not; having money from people who can make good connections is much more valuable, and if a venture fund can raise all the money it needs for its current strategies, the best way to expand the fundable market lie in choosing who to raise it from.

Ads and Addiction

UK regulators are investigating gambling companies for using ads to target people with gambling addictions ($, FT). Gambling is certainly an industry that maximizes profits when it maximally exploits customers' compulsion to gamble. And it's an industry that tracks people very closely to learn exactly what incentives they respond to; the opportunity cost of a comped room upgrade or the ingredient cost of a free drink can be recouped in about thirty seconds of play, and casinos know it. From the article:

Much of the tracking done by gambling companies is ostensibly to monitor for dangerous play, amid concerns that gambling addiction has become increasingly prevalent, particularly during the long stints of the coronavirus lockdown.

There's a natural temptation here: at scale, you need to define problem gambling with some kind of quantitative metric, at which point regulators can tell companies to have zero tolerance for it. But once there's a strict cutoff, the incentive is to do everything possible to keep people just below that cutoff for as long as possible; even if the tail of the distribution is truncated, there's revenue to be had. So a system that's good at detecting gambling addiction is also good at exploiting it, even if the exploitation is a little less frenzied than it otherwise would be.

The Second-Order Costs of Inflation

The WSJ has a look at how a small grocery chain is dealing with inflation ($). The basic answer: it's a lot of work: "Staffers often hear price complaints from customers and swing by nearby competitors to check prices during their lunch breaks, executives said. Mr. Karns [the chain's CEO], other executives and staffers compare prices of staples, ranging from canned soup and granola bars to ground beef and chicken breast, and front-page ad items. He himself sometimes carries a three-inch price catalog to nearby stores of competitors..." Volatile prices introduce overhead at stores, at least in terms of employee time. And since stores don't update their prices at the same time, the most price-sensitive customers may also be spending more time comparison-shopping. This is less likely to happen online, where stores can scrape one another continuously and fight price wars in real time.

Fraud and Transaction Costs

Any time an economic process gets faster and more efficient, the earliest large-scale beneficiaries will include some people who abuse the system. Email led to spam and scams, cheap phone service meant cost-effective telemarketing, and new payments systems lead to new ways to steal people's payment information. This is becoming a concern for telehealth, too. One annoying feature of healthcare debates is that two big sources of costs are 1) insufficient testing, which could spot problems when they're more cheaply treatable, and 2) way too many superfluous tests for low-risk issues, suggested by people who get paid by the test. Anything that lowers the cost of a doctor's visit will increase the incidence of both useful tests and useless ones, and measuring the relative impact (and the cost/benefit thereof) is not easy. The good news is that fraud produces predictable patterns when it scales, and it's also mostly a problem when it scales. So this is fixable, but the cost of a net improvement in healthcare is probably at least a few headlines about the enormous ripoffs that happened along the way.

Diff Jobs

Diff Jobs is our service matching readers to job opportunities in the Diff network. We work with a range of companies, mostly venture-funded, with an emphasis on software and fintech but with breadth beyond that.

If you're interested in pursuing a role, please reach out—if there's a potential match, we start with an introductory call to see if we have a good fit, and then a more in-depth discussion of what you've worked on. (Depending on the role, this can focus on work or side projects.) Diff Jobs is free for job applicants. Some of our current open roles:


  1. A sample of this: " To minimize undesired immune responses to our potential mRNA medicines, our platform employs chemically-modified uridine nucleotides to minimize recognition by both immune cell sensors such as TLR3/7/8, and broadly-distributed cytosolic receptors such as RIG-I. mRNA produced using our synthesis technologies and containing unmodified uridine results in significant upregulation of secreted cytokines such as IP-10, as shown in the figure below. Administration of monocyte-derived macrophages, or MDMs, with unmodified mRNA formulated in LNPs results in an increased ratio of IP-10 transcripts relative to a housekeeping gene, HPRT. By substituting unmodified uridine with a modified uridine, we can substantially reduce immune cell activation in this assay.

  2. Perhaps the best illustration of anchoring bias in all market history is the fact that the S&P only dropped 3.6% on February 24, 2020, the day news was dominated by the pandemic's impact in Italy. It was very hard to believe by that point that if the disease could spread from China to Italy, that it wouldn't be everywhere—and that it wasn't already everywhere, and spreading fast. That was the day the pandemic as a global phenomenon was undeniable, but the market kept dropping for the next month.

  3. It's always difficult to talk about the tradeoff between quality of life on one side and life itself on the other. But tradeoffs abound! Anyone who drives a car to Chipotle is implicitly accepting that there are some burritos worth risking your life for, albeit with low levels of risk. Since we don't always maximize safety above all else, we can and should think about the tradeoffs, even when consequences are severe. At the most benign level, when you expect a future vaccine, delaying a trip you want to take just once in your lifetime is very sensible—you get a lot more utility from taking the same trip when the risk of experiencing or spreading illness is lower. More broadly, the economic output that allowed the US to have a biotech startup ecosystem, subsidize basic research in healthcare, and buy and distribute hundreds of millions of doses is a function of American economic output; at some point, losing GDP means losing lives, though it's very hard to determine the exact exchange rate.

    People don't think consistently about risk. Once they've decided they're willing to take a risk, even if they're off by an order of magnitude, the practical question of which approach is best starts to run into the political question of which approach is popular. And it's probably better, from a policy standpoint, to have a narrow and enforceable set of rules governing the highest-risk cases than to have a broad, controversial, and frequently-violated set of rules with more optimal theoretical coverage.