1. Most of the money goes to programs that fail. The report behind the chart puts the average out-of-pocket cost (the cash actually spent) at about US$229 million per drug in development. Add the programs that failed and the cost of capital over ten years of waiting, and the figure becomes about US$1.16 billion. Most programs fail along the way, and the approved drugs have to earn that money back for them.
2. Most approved drugs never become blockbusters. A blockbuster is a drug that sells more than US$1 billion a year, and drugs like Keytruda, which sell more than US$10 billion a year, are rarer still. Of the 487 drugs the FDA approved between 2014 and 2023, Evaluate forecasts that 193 (about 39%) will top US$1 billion in annual sales at their peak, and only 17 will top US$10 billion.
3. A cheaper front end does help, but only the side that pays for the clinic gets paid. The front end is the discovery stage before trials. At the same success rate, more candidate molecules mean more approvals, so a cheaper front end is worth something. But failure is concentrated in the clinic. Every extra candidate means another round of clinical spending, and the payoff goes to whoever puts up that money. A discovery engine that doubles the number of candidates earns its maker nothing if nobody pays for the trials.
4. These absolute figures run low, but the ratio holds. The figures in this model are lower than the large-pharma numbers usually quoted. DiMasi and colleagues estimate Phase 3 out-of-pocket costs at about US$364 million per drug in development in 2026 dollars (US$255 million in the 2013 dollars they used). Deloitte’s latest count, on 2025 data, puts the average cost of taking a drug from discovery to launch at the 20 biggest drugmakers at US$2.67 billion, with the cost of failures spread across the winners. The estimates disagree on the levels, but in every one of them Phase 3 is the most expensive phase, and by a wide margin.
5. A small share of the cost does not mean a small share of the value. An input that accounts for just nine percent of a program’s spending can still capture far more than nine percent of the revenue, as long as it is scarce enough. Software businesses do this every day. Discovery in drug development cannot, and one reason is that there are too few buyers. Two Phase 3 programs cost roughly US$240 million to more than US$700 million in 2026 dollars. Apart from twenty or so big drugmakers, only about a hundred biotechs can pay for two at once on their own: 72 listed biotechs in the US and Europe have more than US$500 million in annual sales, and a number of clinical-stage companies hold a lot of cash. Meanwhile, nearly 7,000 companies develop drugs. Every one of these buyers has its own chemistry team, and none of them is short of molecules.

1. The winners in this table are the industry’s biggest products. Atorvastatin was the fifth statin (a class of cholesterol-lowering drugs), and its own company’s executives nearly killed it because they thought the market was already taken. Adalimumab was the third drug against TNF (an inflammation signal), and it went on to become the best-selling drug in history. Zanubrutinib was the third BTK inhibitor (a drug that blocks the activity of the BTK enzyme) and beat the first one in a head-to-head trial. Zasocitinib, which comes up again later, was filed for approval four years after the first TYK2 inhibitor was approved, and still beat it head to head.
2. The molecule decided the winner, but the company that ran the trials got paid. Within a validated class, the molecule is the main difference, which is why chemistry is not going away. But in every row the winning molecule came from a chemistry team the winner already had, and the profit went to the company that won the trial.
3. This table only includes classes where a second drug came along. Most targets never get a second drug: the first one fails, and nobody touches the target again. And nobody makes a separate table of the classes where a fast follower had nothing to follow. The rule is narrower than it looks. Latecomers do better only when the target holds up, and someone else has usually spent ten years and a billion dollars finding that out.
4. Following is a strategy for big companies. It only works if you can afford the trials that following requires, and then have the factories and the sales force to win the market. In obesity drugs, semaglutide from Novo Nordisk (NVO) reached the market first, but it spent more than 1,000 days on the FDA’s drug shortage list, which blunted its head start. Tirzepatide, from Lilly (LLY), spent about 300 fewer days on the list. Since then Lilly has steadily taken share from Novo Nordisk, and supply may be one reason.

1. Prices are up, but the deals are still built the old way. By 8 October 2026, the year’s cumulative headline value of out-licensing deals had reached US$139.4 billion, more than in all of 2025. Upfront payments (the cash paid at signing) came to US$8 billion in the first three quarters, compared with US$7.2 billion for all of 2025. According to Evaluate, the average upfront payment in licensing deals for Chinese assets rose from US$52 million in 2022 to US$172 million in early 2026. The business works like this: a company sells a preclinical molecule (one not yet tested in people) for fifty million dollars, about the average upfront in 2022, and after that, each time the molecule changes hands, the value it gains goes to the next buyer. On 30 January 2026, AstraZeneca (AZN) agreed to pay CSPC (1093.HK) US$1.2 billion in upfront cash for the rights outside China to eight once-monthly injectable obesity programs, none of which had finished Phase 1. Eight molecules at that price would have fetched around four hundred million dollars, a third of what CSPC got. Pfizer (PFE) agreed to pay US$1.25 billion upfront for a single bispecific antibody (an antibody drug that binds two targets at once) from 3SBio (1530.HK), which was already in several clinical trials in China.
2. Analysts disagree on how much of the discount remains. By the count of Clarivate (CLVT), Chinese companies accounted for 32% of global out-licensing deals in the first half of 2025, up from 5% in 2020. Yet upfront payments for Chinese assets were still 60 to 70% lower than for comparable Western assets, and total deal values 40 to 50% smaller. In early 2026, though, an Evaluate analyst said upfronts for Chinese companies were catching up with deals elsewhere. Either way, the seller’s money comes mostly at signing, and what the drug earns once it reaches the market goes to the buyer.
3. Data from a single country is the weak point. Sintilimab, a PD-1 inhibitor (a cancer immunotherapy) from Innovent (1801.HK), was licensed by Lilly and had been tested only in China. In February 2022, the FDA’s Oncologic Drugs Advisory Committee voted 14 to 1 that the FDA should require additional trials showing that the results apply to American patients before making a decision. The committee only advises, and the vote was about whether to require more trials, not whether to approve the drug. But the FDA’s own briefing document listed three conditions under which foreign trial data alone can support an approval, and found that the trial did not meet them. The FDA then rejected the application. Surufatinib was rejected on similar grounds, even though it came with a US bridging study (a small extra trial run to show that results carry over to American patients). In May 2026, the acting head of the FDA office that reviews cell and gene therapies said that running the first human trials outside the United States might pay off in the short term, but could backfire in the long run.
4. The rule is not absolute. Toripalimab was approved on 27 October 2023 for nasopharyngeal carcinoma on the strength of a Phase 3 trial in a single region and a Phase 2 trial in a single country, after being turned down earlier over manufacturing problems. But multiregional trials are now the default expectation, and their cost either falls on the seller or shows up as a discount in the price.
5. The current test is ivonescimab. It is a bispecific antibody from Akeso (9926.HK) that targets both PD-1 and VEGF (a growth factor that tumors use to grow new blood vessels), and Akeso has licensed the rights for the United States, Europe and several other markets to Summit Therapeutics (SMMT). Akeso has reported positive results from three Phase 3 lung cancer trials run only in China, and one of them, HARMONi-A, showed a statistically significant survival benefit (one unlikely to be due to chance). But Summit’s application to the FDA rests on HARMONi, a Phase 3 trial run across several regions, and the FDA’s decision is due on 14 November 2026. The FDA had explicitly asked for a statistically significant benefit in overall survival (how long patients live), and HARMONi did not deliver one. A hazard ratio compares the risk of death between the two groups of patients: 1.0 means no difference, and the lower the number, the better the new drug. Among HARMONi’s 165 patients in North America and Europe, the figure was 0.98 at the main analysis in April 2025, against 0.79 for the trial as a whole. These patients enrolled more slowly than expected, so at that point their median follow-up was only 9.2 months. In an update released in July 2026, their hazard ratio was 0.76, close to the 0.78 and 0.79 reported for the whole trial in earlier analyses. The FDA now has to decide whether a survival result that falls short of statistical significance is good enough, and whether the results apply to American patients.

1. Schrödinger had every advantage and still ended up building a pipeline. Founded in 1990, it is the best-known computational chemistry software company, and its software was used to design zasocitinib, discussed later. For decades its answer was no. Then it switched to yes and took three wholly owned cancer programs into the clinic, one of which it dropped in August 2025 after two patient deaths. By mid-2025 it was exploring strategic options for the lead program, which in practice means looking for someone else to pay for the trials. For 2025 it reported revenue of US$256 million and an operating loss of US$167 million. Its net loss narrowed to US$103 million only because of US$65 million of other income, about three-quarters of it a paper gain on equity investments.
2. Without a pipeline, an AI company is a high-tech CRO. A CRO is a contract research organization, the industry’s outsourcing firm. Many AI companies do much the same work: they use algorithms to sift through a large number of options for a drugmaker and tell it which few look most promising, which saves the client time. The business is paid by the project, one job at a time, and cannot be built once and sold again and again like software. By the time a drug reaches the clinic, the factory and the market, the buyer holds all the cards. It has the money, the plants and the sales channels, and it can pay once to buy the patent outright or take a license, after which how much the drug earns has nothing to do with the seller. A seller that wants royalties or milestone payments needs a reason for the buyer to share, and if the buyer can build its own AI team or switch suppliers at any time, there is none.
3. Within a few budget cycles, the pipeline eats the platform’s entire budget. Taking one program from a candidate molecule to a filing for human trials (toxicology studies, process development and the first batches of drug) costs around forty million renminbi, about US$6 million. It is like a chip tape-out (sending a finished chip design to the factory for a trial run): you commit the money, you wait, and only at the end do you learn the result. A company cannot pay for that and a research platform out of the same budget. Two or three budget cycles later, the platform has become an in-house service group, and the company has become a mid-sized biotech with six programs and a few months of cash. Plenty of AI drug companies look like this today, and a better model would not have changed that.
4. The third answer turns the AI company into an asset supplier, like the Chinese drugmakers. It still makes molecules, but sells them early and uses the proceeds to fund the platform. Insilico (3696.HK) discovered the cancer molecule ISM3091 with its own generative AI platform and licensed it to Exelixis (EXEL) in 2023 for US$80 million upfront, a month after the molecule entered Phase 1. Insilico’s chief executive said one reason was an overcrowded pipeline: 31 programs aimed at 29 targets. Chinese drugmakers use a variant of this called a NewCo, in which investors set up a new company outside China that takes the rights to the asset and pays for the trials. The originating company and the investors share ownership, and the originating company still collects an upfront payment and royalties (a percentage of future sales). Hengrui (600276.SH, 1276.HK) used this route to license its obesity drug HRS9531 (in the same class as semaglutide and tirzepatide) to a NewCo called Kailera Therapeutics. Selling early keeps the platform going. The catch is that the AI company sells at early-stage prices and gives up whatever the drug earns later.
5. Recursion and Exscientia merged, then cut programs and staff. They were the two most prominent platform companies in the West, and when they merged in 2024 they promised about ten clinical readouts (the moment a trial’s results come out) within eighteen months. Layoffs began three months later. By the ninth month, two mid-stage programs had been stopped and a third paused. By the tenth, a fifth of the staff had left and research had shrunk to cancer and rare diseases.
6. Two platform deals together brought in one-fiftieth of what one molecule did. Isomorphic Labs is Alphabet’s drug discovery company. It is built on AlphaFold (a model that predicts protein structures) and led by a Nobel laureate. In January 2024 it signed its first two partnerships with drugmakers, with a headline value of about US$3 billion, of which US$82.5 million was cash. In December 2022, Takeda (TAK) bought Nimbus’s TYK2 subsidiary for US$4 billion in guaranteed cash plus up to US$2 billion in sales milestones. What it bought was a single molecule, which had already completed Phase 2b in psoriasis. Nimbus, too, is a computational drug design company. The cash is about fifty times apart, and the plainest difference is that Nimbus had a molecule to sell. Clinical data probably does not explain most of the gap. In the AstraZeneca and CSPC bar, CSPC’s lead program had not been tested in people either, yet it still brought in US$1.2 billion in cash, because eight specific molecules came with it.
7. Isomorphic has not escaped the pipeline dilemma either. A platform lives on investors’ money, because cash from buyers cannot carry it. In May 2026, Isomorphic raised US$2.1 billion in a round led by Thrive Capital, bringing its total to about US$2.7 billion, without having disclosed a single pipeline molecule or put anything into the clinic. The company says it will spend the money developing the platform and using it to fill the pipeline and move programs into the clinic faster. The best-funded discovery platform raised US$2.1 billion to turn itself into a biotech.
8. Most startups cannot name their molecule, and the listed ones are traditional drug companies underneath. Ask an AI drug discovery startup to name one molecule that its own platform produced and that a competent human team could not have. Most cannot. The business plan has a pipeline slide, and underneath the pipeline is conventional chemistry. The listed ones run as two outfits under one name: an AI front that raises the money, and a traditional drug R&D group in the back that does the work. When money gets tight, the AI group tends to go first, though the Recursion timeline in the next section cannot show which group was cut first. It only shows that under pressure, the merged company cut programs and a fifth of its people, with layoffs on both sides of the merger.
9. Two well-known AI drug companies listed in Hong Kong gave opposite answers to the pipeline question. Call them Company A and Company B. Company A essentially said no. It sells drug discovery services, lab robots and what it calls AI-for-science solutions, to customers in new materials and consumer health as well as pharma. It raised more than US$750 million in 2025 by placing new shares, and that year it turned its first full-year profit, RMB 134.6 million on revenue of RMB 802.6 million. But by its own announcement, its two core businesses, robotics and drug discovery, were still losing money, just less than the year before. The profit came mainly from two places. First, financial assets it holds rose in value by at least RMB 500 million. Second, in 2024 it had booked a large paper loss on preferred shares issued before its listing. Those shares converted to ordinary shares at the listing, so the loss did not recur. Set those two aside and Company A is a services company with a large investment portfolio, earning its revenue from clients. Company B said yes. It listed at the end of 2025, in what it called Hong Kong’s largest biotech fundraising of the year. In 2025 it had revenue of US$56.2 million, an adjusted loss of US$43.8 million, and R&D spending of about 145% of revenue. By its own account, its newest batch of eight heart and metabolic disease molecules goes after established, high-confidence targets such as GLP-1R (the target of semaglutide) and Lp(a) (a blood particle linked to heart disease), plus a few newer ones, and aims to stand apart on chemistry. That is the follower strategy, and a sensible one. Its lead program is the exception: the target is new, and its own platform found it. It expected to turn a profit in the first half of 2026, mainly from out-licensing and collaboration income, which is the third answer described above.
10. Why the AI label helps, and why so few companies will pay for the research. The AI label helps because it lowers the cost of raising money: once one drug succeeds, investors treat it as evidence that the platform behind it can produce more. Few will pay for the research itself, because anyone can free-ride. Whatever the first company learns about making the models work, everyone else can learn for free just by looking at its results, so the rational move is to sit on the cash and let someone else go first, just as with a second drug.

1. REC-4881’s molecule came from Takeda, and the platform picked the disease. The molecule was called TAK-733, and Takeda had tested it in solid tumors. The earliest version of Recursion’s platform screened human cells missing the APC gene (a gene that keeps gut cells from overgrowing) and found that blocking MEK (an enzyme that relays growth signals inside cells) made the cells look closer to normal. In 2020 Recursion licensed the molecule from Takeda for US$1.5 million upfront plus up to US$39.5 million in milestone payments. FAP is an inherited disease caused by mutations in APC. Hundreds to thousands of polyps grow in the gut, nearly every untreated patient develops colorectal cancer, and no drug is approved for it. FAP was picked by the earliest version of the platform, so the choice says little about how good today’s system is.
2. The results are encouraging, but early. As of September 2026, 13 of 15 evaluable patients had less polyp burden after 12 weeks, with a median drop of 47%. Twelve weeks after stopping treatment, 11 of 13 were still below where they started. For comparison, in a Dutch registry, 87% of similar patients who took no drug saw their polyp burden rise each year. Of the 26 treated patients, 4 had grade 3 side effects (severe but not life-threatening), and 4 stopped because of adverse events. This is a small open-label trial (doctors and patients know which drug is given) with no control group, so it cannot show much yet.
3. A registrational trial will probably have to measure something else. Another FAP drug is already in Phase 3: eRapa, a formulation of rapamycin from Biodexa (BDRX). Its trial is randomized, double-blind and placebo-controlled, with 100 of a planned 168 patients enrolled by October 2026. According to Biodexa’s chief executive, the FDA and the European Medicines Agency both said polyp burden alone could not serve as the endpoint of a registrational trial. So the eRapa trial uses a composite endpoint and has to wait until enough patients progress (needing major surgery, for example), which could be as early as mid-2028 or as late as the end of 2029. REC-4881’s results so far measure polyp burden. If the FDA takes the same view of it, the registrational trial will have to measure something else and will take longer.
4. REC-1245’s target and molecule both came from the platform. It makes cells break down a protein called RBM39, which helps with RNA splicing (the step that trims freshly copied RNA into its finished template). The company says it took 18 months to go from the biological finding to a drug candidate. In the Phase 1 trial, among patients on 45 mg a day or more, there were three partial responses (tumors shrinking by at least 30%), two confirmed in endometrial cancer and one unconfirmed in liver cancer. Disease was controlled (tumors shrank or stopped growing) in 11 of 12 patients. The first data from an endometrial cancer expansion group are due in the second half of 2027. RBM39 is not a new target, though. A paper published in Science in 2017 found that an old drug, indisulam, works by degrading RBM39, and that it was safe in early trials but fewer than one patient in ten responded. REC-1245 has to show it can do what indisulam could not.
5. The AI-designed molecule Recursion inherited now needs someone else to carry it. REC-617 is a CDK7 inhibitor (CDK7 is an enzyme that helps control cell division and gene transcription). It came from Exscientia, which promoted it as precisely designed with AI. As of September 2025, one of 29 patients with advanced solid tumors had a confirmed partial response, and five more had stable disease. In October 2026 Recursion said it would look for partners to take it and another program, REC-3565, forward. Before that, the three mid-stage programs stopped or paused in May 2025 (see the table above) were all ones Recursion had taken into Phase 2 itself. Counting REC-4881, only one of the four has reached the point of planning a registrational trial.
6. Buyers have paid, but not enough to keep the company going. Roche and its Genentech unit have paid US$216 million to date, and Sanofi US$144 million. In 2026 both picked their first project from the collaborations: Genentech chose a neuroscience target, and Sanofi chose an immunology drug candidate, which Sanofi will pay to develop. That is buyers grading the work with real money. Yet together these payments roughly match one year of Recursion’s cash spending. The company has capped its 2026 cash operating costs at US$375 million, while its third-quarter revenue was only US$18.6 million against a net loss of US$119 million. The stock market fills the gap. In October 2026 it sold another US$200 million of shares, which the company says will keep it funded into 2029.
7. The shares trade at less than a quarter of the IPO price. Recursion went public in 2021 at US$18 a share, raising US$501.8 million including the underwriters’ extra shares. On 3 November 2026, after the latest REC-4881 data, the shares barely moved, trading around US$3.84.
8. What to watch. Recursion says it will give more details on the registrational trial later in 2026. The questions are which endpoint the FDA will accept and whether the trial needs a control group. Three things are due in 2027: the Sanofi candidate entering Phase 1, the REC-1245 endometrial cancer data, and whether the REC-4881 registrational trial starts on schedule.

1. Human Genome Sciences was not hype, and the science held up. The genome, the patents and the drug were all real: belimumab was the first new lupus drug in fifty years. It just took twenty years to produce a single drug, because in 1990s drug discovery, sequencing was the cheap half, as molecule design is now. The company that patented genes was eventually bought by a company that could run trials, and its management was gone within fifteen days.
2. BenevolentAI went down the same road, only faster. It called itself a machine brain for bioscience, raised money at a valuation of two billion dollars and went public through Europe’s largest SPAC deal (a SPAC is a listed shell company that merges with a private one to take it public). When investors asked where the drugs were, it built a pipeline. When the pipeline burned through the cash, it cut the pipeline. It laid off staff round after round, announced it would go back to selling the platform, and finally delisted. The whole cycle took twelve years.
3. The technologies all survived. Genomics is now the basis for target validation (confirming that a target really is linked to the disease in people) across the industry, and nobody would start a program without it. Natural-language systems that read the literature are in use at every major drugmaker. The molecule in the next section was designed with free-energy calculations.
4. They misjudged who gets paid for a tool. The toolmaker does not own what the tool finds: Illumina (ILMN) sells sequencers, but it does not own the drugs discovered with them. If buyers can build the tool themselves, as with chemistry software, the toolmaker gets even less, because a tool cannot sell for more than it would cost the buyer to make one. Schrödinger sells FEP software (free-energy perturbation, the free-energy calculation mentioned in the previous item), yet the money it made from zasocitinib (a US$111 million distribution) came from its stake in Nimbus, not from the software. Until its first clinical readout, a discovery platform is just a tool valued as if it were a drug company.
5. So far, the only way out anyone has found is to sell the asset. Human Genome Sciences did it by getting acquired, and Nimbus by selling a subsidiary. BenevolentAI did not manage it, because by the time it wanted to sell, it had nothing anyone wanted to buy.
6. Biotech stocks have badly lagged tech. Using adjusted monthly closes, from the end of 2011 to the end of October 2026, XBI, an index fund of small and mid-sized biotech stocks, rose about 632%, while QQQ, which tracks the Nasdaq-100 and its large tech companies, rose about 1,304%, a gap of nearly 700 percentage points. Before the pandemic the two ran close: from the end of 2011 to the end of 2019, XBI rose about 343%, slightly ahead of QQQ’s 315%. Almost all of the gap opened after 2020, the years when AI advanced fastest and AI drug discovery raised the most money. Index returns reflect many things, including interest rates, acquisitions and drug-pricing policy, so not all of this can be pinned on AI, but there is no sign that AI has lifted biotech as a whole.

1. The aggregate. In 2024, a team at the consultancy BCG went through every molecule from an AI-native biotech that had completed a clinical phase. Of the 24 that finished Phase 1, 21 passed, about 88% (the report gives 80 to 90%), against a historical 40 to 65%. Of the 10 that finished Phase 2, 4 passed, about 40%, against a historical 30 to 40%. The authors’ own reading is that AI is good at producing molecules with drug-like properties (absorption, metabolism, safety and the other properties a drug needs). Comparing the midpoints of the ranges, AI-designed molecules beat the historical pass rate on the engineering question by roughly twenty to forty percentage points. On the scientific question, ten molecules are far too few to prove anything, and their pass rate looks like the historical one. The authors also list well-validated targets as one possible reason for the high Phase 1 pass rate, though they argue it is not the whole explanation.
2. What a doubled hit rate is worth to a buyer. The arithmetic is simple. If the buyer’s own screen succeeds ten percent of the time and a vendor’s succeeds twenty percent of the time, the buyer can screen twice. It only needs one hit (one compound that binds the target), and it can afford to screen, so doubling the success rate only halves its work. It is like a coupon from a food-delivery app: people on a tight budget use it, and people with money do not even look at it. That value is real, just far smaller than the business plan says. There are two exceptions. One is time: if a patent filing is waiting on the screening results, speed is worth money. The other is a target that cannot be screened, and only there does a model open up something new. A model earns its keep by getting a hit where there was none, not by raising ten percent to twenty, and nobody has shown that yet.
3. The best case: zasocitinib. In 2016, a joint team from Nimbus and Schrödinger set out to make a selective TYK2 inhibitor. TYK2 belongs to the JAK family of enzymes, and the binding sites of TYK2, JAK1, JAK2 and JAK3 are almost identical, so selectivity (hitting TYK2 without touching the others) was the whole challenge. They started with JH1, the catalytic domain (the part of the enzyme that does the actual work), using free-energy calculations to score about four thousand designs, pick 46 to synthesize, and find nine that met the project’s criteria. They then turned to JH2, a regulatory domain away from the catalytic one, and screened 2.4 million commercially available compounds computationally against a set of protein structure models. That turned up a new kind of starting molecule, which they optimized into zasocitinib. Binding the regulatory domain instead of the catalytic one is what made selectivity possible: according to Takeda, the molecule’s selectivity over the other JAK enzymes is more than a million-fold. It was a hard, expensive chemistry problem, and the computation solved it.
4. Up close, the story is less impressive. It was Bristol Myers Squibb (BMY) that validated the target and found the allosteric pocket (a binding site away from the enzyme’s active site). Nimbus came later, working on a target whose risk someone else had already removed, and making a second drug. Its method was physics-based computation, the FEP described earlier, not a generative model trained on chemistry text. Nimbus’s own president of R&D has said plainly that he would not call a compound identified back in 2020 the first AI-approved drug. And the company that supplied the software was Schrödinger, which now runs its own pipeline alongside its software business. What works here needs three things: a target someone else has already validated, a question narrow enough to be solved as a physics problem, and a wet lab (a lab that really synthesizes and tests compounds) that can afford to make and test dozens of molecules in every round.
5. What FEP calculates, and what it does not. FEP estimates the difference in binding free energy between two similar molecules binding to the same protein structure. It simulates one molecule gradually turning into the other, accounting along the way for the energy of the water, the protein and the ligand (the candidate molecule). Done well (with enough sampling and a reliable structure), it is accurate to about one kilocalorie per mole, which corresponds to roughly a fivefold difference in affinity (how tightly a molecule binds). That is enough to rank similar molecules and decide what to make next. It does not tell you whether you picked the right target. Nor does it predict whether the compound will be metabolized within an hour, block the hERG channel (a heart ion channel whose blockade carries a risk of arrhythmia), build up in the liver, or work in mice. It assumes you already know how the molecule sits in the pocket, and it quickly becomes unreliable once the chemical change is large, the protein moves, or the structure itself is wrong. It answers one narrow question well. Every other question in the cycle of designing, making and testing compounds still needs an experiment.

1. What an old hand actually decides. He has to judge which differences in a structure–activity table (a table of how changes to a molecule’s structure changed its activity) are signal and which are assay noise. Test the same compound twice and the results can differ twofold, so a twofold improvement proves nothing. Knowing when a change is real is a knack nobody has written down. He remembers that a scaffold (the core structure of a molecule) was dropped at his previous company because of a toxicity that was never published. Working within a synthesis budget, he picks thirty analogs out of a thousand plausible ones, knowing he will see results only once a month. And he has to decide when a series (a family of molecules built on the same scaffold) is dead. Of everything he does, this decision costs the most and rests on the least evidence.
2. Failures are not published, and the published record is unreliable. Failures stay in internal company databases, the most tightly guarded assets in the industry, so a model trained on the public record learns mostly from attempts that worked. Amgen’s hematology and oncology team spent years trying to reproduce 53 landmark cancer papers before betting programs on them, and confirmed six. A Bayer team reviewed 67 of its own in-house projects and found that the published data fully matched its own results in only about 20 to 25% of them. A third replication effort, the Reproducibility Project: Cancer Biology, repeated 50 experiments from 23 high-impact papers. For results that were positive in the original papers, the replicated effect sizes (how big a change the experiment measured) were 85% smaller at the median. Of the effects it could evaluate, fewer than half (46%) met most of its criteria for success. Part of the reason is the incentive: every paper needs a gene nobody has claimed yet, so search almost any gene together with the word tumor and you will find a paper linking the two. So a large share of a big drugmaker’s internal research is filtering: before a program is bet on, its scientists repeat the published findings and throw out the ones that do not hold.
3. Who sets the test. Drug discovery cannot pick its own judge. First the human body judges, then the FDA. Nobody can define a benchmark, win on it and declare the problem solved, the way computer vision and language modeling do every day. Outside grading is slow and costly, and it also makes it hard for the field to fool itself. To train the kind of judgment described above, you would need boring, specific records from inside real programs: the situation at the time, the decision made and what happened next. For each series, such a record would show what was known, which thirty compounds were picked, what data came back and where the series died. It would include the failures (which are most of it) and carry timestamps, so that a model is scored only on what could have been known at the time. Data like this exists only in fragments inside a few dozen companies, and it will not be pooled, because the record of failures is itself a competitive asset. This is an institutional problem, and more computing power does nothing about it.
4. The science benchmarks in model launches are Kaggle competitions. A Kaggle competition gives everyone the same fixed dataset and a question, and the organizer holds the answer key. You submit, you get a score right away, and you can submit again. The science benchmarks reported in model launches come in roughly three kinds (see the table above), and each has the same parts: a fixed dataset, a question, and an answer written in advance. One widely used agent benchmark consists of 75 Kaggle competitions, scored against Kaggle’s own leaderboards. Some newer tests use real data and real research workflows, which is progress, but every question still comes from a result someone already has. A high score shows a model can find answers that already exist, and a drug program pays only for the ones that do not.
5. Critics inside the field agree. The leaderboards most used within drug discovery are built the same way. A researcher who has spent many years in industry cheminformatics (the use of computers to work with chemical data) argues that the most widely used leaderboards are too flawed to draw conclusions from. In December 2025, a paper with 56 authors argued that mainstream science benchmarks test knowledge out of context and overlook what drives discovery: step-by-step reasoning, proposing hypotheses and interpreting observations.
6. What a real AI-for-science result looks like. A real result is a prediction about something nobody knows yet, checked afterwards by someone else. In drug discovery, that means predicting that a molecule will bind before it has been made, or predicting that a target matters before the clinical trial, and then letting a lab or a trial grade the prediction. Mathematics already has results like this: a model raised the known lower bound on the kissing number in 11 dimensions (the largest number of equal-sized spheres that can touch a central sphere of the same size at once) from 592 to 593. Nobody had found that arrangement before, and anyone can check it. A result in mathematics can be checked by other mathematicians. A result in drug discovery can only be checked in a lab or in patients, which takes months or even years and real money. So model companies report tests that can be scored the same day.

1. Parts of these decisions have already been modeled and audited. The dose is set by exposure–response analysis (how the drug’s effect changes with the dose and its level in the blood). AstraZeneca has published a scorecard comparing its results before and after it overhauled how it decides which programs to advance and which to stop: the share of candidates that went from selection to the end of Phase 3 rose from 4% to 19%. But the reasoning behind individual decisions is not recorded anywhere. That is no knock on the people making them, whose experience is real and usually beats the alternatives.
2. Why the record does not exist. Hospital medicine explains this best, and portfolio meetings (the meetings where a company decides which programs to fund) work the same way. In the clinic, the real basis of a decision is liability: whoever pays if something happens to the patient gets to decide. And the decision rests on a judgment nobody has quantified, such as how bad this kidney is, how high this blood pressure is, and how much this patient’s age should adjust the formula. So the decision comes out as “the other center doses at 200, we dose at 150”, and the reasoning is never written down. This is street smart, not book smart, and it works. Portfolio meetings run on the same logic: whoever takes the blame if a program fails makes the call, and a written prediction could later be held against whoever wrote it, so it never gets written. Yet the moment someone writes a prediction down before the outcome, this kind of knowledge becomes testable.
3. These decisions suit a model well. They are rare and expensive, so being a little less wrong each time is worth a lot of money, and nobody needs an answer in milliseconds. The decisions themselves are already recorded in the minutes, because they are made at a meeting with a fixed agenda. What is not written down is what was known when each decision was made, and recording that only takes a change to the format of the minutes.
4. The hard part is scoring a decision to stop. You only ever see the outcome of the decision you actually made, and a series you stopped will never produce data. So a look back after the fact cannot score a decision to stop. Two things can be scored. One is predictions on the series that were continued, registered before the results came in. The other is the judgment of whichever side was followed when the model and the humans disagreed. Settling it takes forecasts written down and dated in advance, not a database.
5. Nobody has published this test. Some companies sell pieces of it: trial-outcome prediction, digital control arms (using historical data in place of real control patients), and portfolio analytics. But none has published a time-stamped record that puts the model’s calls next to the humans’ calls on the same set of decisions, registered before the data read out. Meanwhile, the money and the talent have poured into molecule generators.
6. The two test cases will not settle it. Zasocitinib’s regulatory outcome arrives in the first quarter of 2027, but the company that designed it does not call it an AI drug. Rentosertib’s trial runs at 47 centers in China, and to win US approval it first has to get past the single-country data problem described earlier. That leaves rentosertib as the one real test of AI molecule design, and the result will not arrive until 2029. Even if it succeeds, it answers a question about the nine percent of the budget.
7. A 2026 Perspective reached the same conclusion. A Perspective published in Nature Reviews Drug Discovery in 2026, whose authors include the researcher who coined Eroom’s law (the observation that the cost of each new drug has kept rising for decades instead of falling), calls AI’s clinical impact “disappointingly limited”. AI-designed drugs still hit the same Phase 2 wall as everything else. An improvement of a given size is worth most when it raises the clinical success rate, above all in Phase 2. On this view, a model is worth roughly how much it raises the clinical success rate, multiplied by how many programs it can be used on. A molecule can bind its receptor tightly, kill cells in a dish and shrink tumors in mice, and still do nothing in people, or help only a small fraction of them. Only a molecule that shows a statistically significant effect in a Phase 2 trial has earned the right to be called a drug. But academia chases impact factors and grants, so important negative results rarely get published, and companies built around AI prefer a big story. The statistician John Tukey put it this way: an approximate answer to the right question is worth far more than an exact answer to the wrong one.
8. What would overturn this conclusion. First, a platform company reaches a Phase 3 readout on an asset it still owns, without turning itself into a conventional biotech along the way to raise the money. That would show the pipeline dilemma can be avoided and the structure described here is wrong. Second, a discovery platform with no clinical assets gets an upfront payment of US$500 million or more. An equity round does not count. It would show that buyers have started to price the method itself, and the fifty-to-one ratio above would be out of date. Third, someone publishes a time-stamped scorecard from inside real programs showing that, on the same budget and with failures counted, a model beats experienced people at the decisions listed above. It is the cheapest of the three: a company that already holds these meetings could register every prediction before the results come back and publish the scorecard when the program ends, at no extra experimental cost. Still, nobody has published one.

1. Fraud has become an industry. More than 10,000 research papers were retracted in 2023, a record, and more than 8,000 of them came from journals of a single publisher, Hindawi. The retracted Hindawi papers had already been cited more than 35,000 times. A 2025 study in PNAS found that the output of suspected paper mills (operations that mass-produce fake papers and sell authorship) is doubling every year and a half, while the scientific literature as a whole doubles only every fifteen years.
2. Retractions catch only a small part of the problem. A screen of 20,621 biomedical papers found problematic duplicated images in 3.8% of them, at least half suggesting deliberate manipulation, and that study checked only one kind of error. An economic analysis estimated that more than half of preclinical research is irreproducible, costing about US$28 billion a year in the United States alone. The Amgen and Bayer replication efforts described earlier point the same way: most published findings do not fully hold up when someone tries to build a program on them.
3. One paper can lead a field astray for nearly two decades. A 2006 Nature paper reported that a specific amyloid assembly, Aβ*56, impairs memory, and it became one of the most cited papers in Alzheimer’s research, steering a great deal of later work and funding. In June 2024 it was retracted after investigators found manipulated images, making it the most cited paper ever retracted. Any model that learned the field from its literature learned that result as fact.
4. A model that reads the literature learns its errors too. Text cannot tell a model which results were fabricated, which were never replicated, and which experiments failed and were never written up. A knowledge graph built from papers turns a claim into an edge, and a claim cited a hundred times looks a hundred times more certain, even when every citation traces back to the same flawed experiment. More reading only adds more copies of the mistake.
5. Why wet-lab data can be trusted. Wet-lab data is not free of noise either: as noted earlier, the same compound tested twice can differ twofold. The difference is that the noise can be measured. The protocol is fixed, the controls run alongside, the failures are recorded, and any result can be run again. That is also why Lilly’s TuneLab and J&J’s antibody design challenge, both covered in the next section, insist on data measured in their own labs.

1. AI can already help run experiments, but it cannot yet replace judgment. Some companies already use AI to design experiments and decide the order to run them in, and the boost is considerable. When it comes to decisions and judgment, though, AI still falls short. One company running a randomized Phase 2 trial worked closely with Anthropic in the months before database lock (the point after which trial data can no longer be changed).
2. The hardest part of biotech is translating academic findings, and translation is at heart a reproducibility problem. Most academic results cannot be carried straight into a drug program, which is the same problem as the failure to reproduce published results described earlier. One remedy is to work closely with academic founders who are willing to keep working on a finding. Their labs keep refining it long after the paper, which is closer to engineering, so what reaches the company is far more mature than what a typical lab hands over.
3. A company founded this way starts much further along. By the time the technology moves into a company founded by the professor who runs the lab or by one of the lab’s students, it is still pre-IND (the IND is the application to start human trials), but it already has months of research and several rounds of lead optimization behind it. It is no longer just the version in the paper, and building the company becomes much easier.
4. Lilly’s TuneLab is a drugmaker getting into the data business itself. In September 2025 Lilly launched TuneLab, with models built on its own drug disposition (how a drug is absorbed, distributed, metabolized and excreted), safety and preclinical data covering hundreds of thousands of molecules, backed by more than US$1 billion in research. Biotechs can use the models, and in return they contribute their own data to improve them through federated learning, which trains a model on each party’s data without moving or exposing it. Distribution has been quick: TuneLab is already plugged into Benchling, whose customers include more than 1,300 biotechs, and into CDD Vault, while Charles River (CRL) provides testing for participating companies. TuneLab itself also sits under Lilly’s Catalyze360 program, right next to Lilly’s venture arm. So Lilly keeps its own data, benefits from everyone else’s without ever seeing it, and gets an early view of which biotechs are worth investing in or buying. That last part may be worth more than the models themselves.
5. TuneLab still works on the cheap half. By its own account, the models help participating companies weed out candidate molecules faster. That is useful, and it is graded by real experiments, but it falls within the nine percent before the first human dose and does not touch the question of whether a target matters in people. There are two further risks. A biotech feeding its data to a company that may well end up acquiring it will sooner or later hold some of that data back. And there is no public, pre-registered evidence yet on how much federated learning gains across labs whose experimental methods differ. There is also a side effect: a drugmaker that opens models trained on a billion dollars of its own data to biotechs, asking for data instead of money, puts direct pricing pressure on companies that sell property-prediction models. Frontier model companies such as Anthropic complement TuneLab: the language model runs the workflow and the property model scores the molecules, but neither makes the decisions that cost the most money.
6. Johnson & Johnson shows what it looks like when the buyer sets the test. Investors now say they want companies that can run their own design, make and test loop: companies that build AI into their own systems and, above all, generate their own data. Companies that build foundation models on public datasets are everywhere and cannot set themselves apart. In October 2025, J&J (JNJ) launched a challenge in computational antibody design. According to people in the industry, each of the dozen or so antibody design companies taking part designed about 100 sequences for each of three targets, J&J is testing them in its own labs, and the winners will get the chance to work with J&J. The result is that J&J now holds head-to-head data on a dozen or so of the world’s leading AI protein design companies, on the same targets. The companies have never had a benchmark like this, and all of them want the data, which J&J keeps.
7. Data has to be distinctive, not just big. At a recent panel of AI biotech founders, a few practical points came up. First, a model that understands DNA does not understand disease. The chain from DNA to RNA, proteins, cells and finally patient outcomes is long, and the data that exists is not enough to predict it end to end, so it is more practical to model each layer and then connect them. In the same way, understanding the variants that cause disease is not the same as understanding the disease. Second, an unusual biological phenotype can teach more than a pile of samples. One company, for example, studies how the hearts of hibernating animals repair themselves after injury, looking for new leads in heart failure. Third, the goal for a model is to predict how people respond to treatment, and predicting sequences better is only a step toward that. Fourth, drugmakers may come to want drug assets more than pure technology platforms. AI should also explore chemistry that traditional methods cannot reach, against targets that already have clinical validation, instead of squeezing more efficiency out of known chemistry. Finally, some think the wave of protein language models has matured and the next opportunity may lie in small molecules, but in the end the value of any model or dataset has to be proven by commercial results.

1. The samples are small. The Short Volatility index launched in 2015 with 15 funds; by early 2022 it had two. Relative Value had nine funds in 2023, and Tail Risk seven in 2020. An average of two funds says little about a whole category.
2. Funds that failed are not counted. XIV lost more than 90% in a single day in February 2018 and was shut down; LJM Preservation and Growth lost roughly 80% in two days; Malachite and Parplus were wound down in March 2020; Allianz Structured Alpha lost about $7 billion. The index’s −31.75% drawdown is the drawdown of the funds that survived.
3. A few months can change the long-term numbers. The Long Volatility index’s annualized return since inception was 4.17% at the end of February 2020 and 5.38% a month later. One month changed a fifteen-year record.

Month by month, the Short Volatility index lost 12.53% in February and 15.85% in March, then made 9.10% in April, 12.36% in July, 6.84% in August and 5.12% in September, ending the year down 0.80%. The reason is that after a spike, implied volatility (the volatility priced into options) falls more slowly than realized volatility (how much the market actually moves). So the extra return from selling options is largest in the weeks after a spike. A fund that survives the spike earns its loss back within months. Malachite was wound down in March and never got that chance.

Research shows when this premium is small and when it is large. Before a spike, protection gets cheaper, not more expensive. Cheng (2019) finds that the premium built into VIX futures falls or stays flat while risk rises, because demand for hedging falls too, and VIX futures were cheap just before both the 2008 crisis and the 2011 euro crisis, because buyers who had cashed out did not buy again. During a spike, selling volatility may not be rewarded at all. Johnson (2017) shows that the slope of the VIX term structure (how implied volatility differs across maturities), not the level of the VIX, predicts returns on variance swaps, VIX futures and straddles, and on the days with the lowest slope, when short-term volatility was priced above long-term volatility, owning variance paid off the next day in 17 of 18 asset–maturity combinations. The large rewards come after the shock: in stocks, the price of variance risk rises sharply after shocks such as 1987 and 2008. In short, the premium is small before a spike, may be missing during it, and is largest after it.

The Tail Risk index showed the same pattern in reverse: up 37.44% in March, then down in seven of the next eight months, after losing 10.40% in 2019. The Long Volatility index did the same, going from +39.70% in the first quarter to +27.37% by the end of August. A hedge that is never sold just goes up and comes back down. The most important skill of a tail fund is knowing when to sell its gains, not when to buy protection.

The Tail Risk index has a negative Sharpe ratio (return per unit of risk) and is still worth owning, because a good Sharpe ratio is not its purpose. Its median rolling 12-month return is −7.41% and its correlation with MSCI World is −0.50, and it has cost about 2.65% a year since 2008. The buyer pays roughly 3% a year for something that made 37% in the worst month of the decade.

Individual funds vary far more than the index:

• Universa reported 3,612% for March 2020, and LongTail Alpha’s tail fund rose roughly tenfold, or 157% counting client cash held in reserve. Those are returns on the hedge itself, not on the whole portfolio. Universa recommends putting 3.33% of a portfolio in its fund; by its own calculation, a portfolio of 96.67% S&P 500 and 3.33% Universa returned +0.4% in March 2020 against −12.4% for the index, and 11.5% a year since inception against 7.9%.
• Cambria’s TAIL ETF, which holds Treasuries plus a ladder of out-of-the-money S&P 500 puts (puts that pay off only after a large fall), has lost 44% since its 2017 launch. The idea is the same as a tail fund’s, but the method is simple and mechanical.
• 2024 was a calm year with one spike, and volatility-trading hedge funds averaged +4.03%. On August 5 the VIX went from 16.36 at the end of July to 65.73 before the open, closed at 38.57 and was back at 14.80 by August 16. The ten largest short-volatility ETFs lost $4.1 billion, and among the 25 volatility funds PivotalPath tracks, those that were short volatility lost about 10% that day while the group as a whole gained 5.5–6.5%.

The gap between the VIX and the volatility realized over the following 30 days has averaged 3.33 points since 2013. Even in 2025, April included, realized volatility came in about 3 points below implied, in line with the average since 2000. About three points a year is all that short-volatility strategies have to work with.

The PUT index sells one-month S&P 500 puts at the current index level every month, and holds enough cash to cover them. From July 1986 to August 2023 it returned 9.40% a year against 9.91% for the S&P 500, with 10.26% volatility against 15.38% and a −32.66% maximum drawdown against −50.96%, and it beat the index 95% of the time when the index’s rolling one-year return was negative. Selling options systematically earns a return similar to stocks, but along a different path: steadier, more prone to sudden losses, and better in bear markets. These are returns before costs, and options are expensive to trade.

That makes PUT a free benchmark, and professional investors use it that way: a short-volatility fund earns its fees only if it beats both PUT and the Short Volatility index after adjusting for drawdowns. Many do not, which is part of why the category’s average looks poor. Three points a year leave little room for the usual 2% management fee and 20% performance fee, plus trading costs.

1. Night. Muravyev and Ni split the daily return on delta-hedged S&P 500 options (options hedged so that small moves in the index do not change their value) into two parts. The average loss of 0.7% a day breaks down into −1% from close to open and +0.3% from open to close, and the pattern holds across maturities, strike prices and single-stock options. This fits the idea that option prices do not account for stock volatility being much higher during the day than at night. In the working-paper version, the two parts are about equally volatile and have similar extreme moves, so risk cannot explain the gap. Options are expensive at the close and cheap at the open. Selling them every day earned 0.7% a day before costs but lost 80% of capital in the financial crisis; selling only overnight earned 1.0% a day before costs and was positive in every three-month period, the crisis included. The crisis losses happened during trading hours. Weekends show the same pattern: option returns are significantly lower over non-trading periods, mostly weekends, because of mispricing rather than risk, although an earlier version found little of the effect in S&P 500 index options. All of these figures are before costs, use mid prices and mostly come from before SPX options traded overnight, and the effect may have weakened since publication. The cost of trading twice a day would eat it.
2. Sharp moves that actually happen. Dew-Becker, Giglio, Le and Rodriguez use variance swaps (contracts that pay according to realized volatility) of different maturities to separate realized variance from news about future variance. From 1996 to 2014, hedging news about variance at horizons from one quarter to 14 years cost nothing on average; only unexpected, short-lived realized variance carried a premium. Their example is the 2011 debt-ceiling fight. The one-month variance swap rose from 16.26 at the end of June to 25.96 at the end of July while realized volatility stayed near 15, and then August’s realized volatility was 47.18. Across 19 option markets, the premium for exposure to actual price swings (gamma) is negative and the premium for exposure to changes in implied volatility (vega) is positive, and both become more extreme for options further out of the money. Investors pay to avoid the sharp move itself, not the fear of one.
3. Long shots. Options give leverage that some investors cannot get elsewhere, and the more leverage an option offers, the lower its risk-adjusted return: a portfolio that buys low-leverage and sells high-leverage index options earns abnormal returns with a t-statistic of 6.3. Prediction markets offer a test with no economic risk at all. On Kalshi, across more than 300,000 contracts, a 10-cent contract wins less than 10% of the time, takers (who accept posted prices) lose almost 32% on average and makers (who post them) about 10%, and the bias holds in politics, entertainment and economic-data contracts alike. Same-day options, known as 0DTE, show the same behavior: about 75% of retail S&P 500 option trades are 0DTE, and since daily expiries began, retail has lost about $358,000 a day, roughly 60% of it to transaction costs.

So a short-volatility fund sells two different things. The first is insurance: index puts near the current price, like the ones in the PUT index, which earn a premium for bearing the risk of sudden drops. The second is leverage and long shots: far out-of-the-money and very short-dated contracts, which earn a premium that has nothing to do with crash risk, as Kalshi’s entertainment contracts show. And the gap between makers and takers shows that how well a fund trades is part of its return, not just a cost subtracted from it.

For the investor:

1. Short volatility offers a risk premium other than the stock-market return: the profit from selling insurance, which a standard 60/40 stock-bond portfolio does not earn.
2. Relative value offers returns with genuinely low correlation. A Sharpe ratio of 1.18 at 3.9% volatility with a 6.5% drawdown is rare among hedge fund strategies, and Dominicé found that adding the Long Volatility and Relative Value Volatility indices to the S&P 500 gave about 7% a year at more than double the index’s Sharpe ratio. That low correlation cannot come from trading skew (the extra price of downside protection compared with upside bets). Kozhan, Neuberger and Schneider find that almost half of one-month S&P 500 skew is a risk premium, but skew risk and variance risk are nearly the same risk: if you collect one while hedging out the other, the premium almost disappears. A relative-value fund that is net short skew is really short volatility, so its low correlation has to come from other trades, such as dispersion (index volatility against single-stock volatility), spreads between maturities or trades across asset classes.
3. Long volatility and tail risk are the only exposures that pay out by contract, and immediately, when stocks fall sharply. Bonds failed at that job in 2022, and stocks and bonds have become more correlated. The math behind this is sometimes called the volatility tax: long-run compound growth is roughly the average return minus half the variance, so deep losses do outsized damage (a 50% loss needs a 100% gain to recover). A hedge that costs about $5 million a year per $1 billion protected raises the long-run compound growth rate if it turns a −34% drawdown into −6%. An independent backtest of a Universa-style strategy on Cboe data from 2006 to 2020 produced 11.0% a year, close to Universa’s claimed 11.5% against 7.9% for the S&P 500. The hedge also lets an investor take more risk elsewhere, for example holding 100% in risky assets instead of 60/40, and it pays out exactly when assets are cheapest. CalPERS is the cautionary example: it hired Universa and LongTail Alpha in 2017, then pulled its money out, and the Universa position it gave up returned about 3,600% in March 2020, more than $1 billion it could have used to buy assets at their lows.

For the market, the reason is structural. Demand for options is one-sided: institutions regularly buy index puts and sell calls, and structured products pass long-dated volatility and correlation risk to dealers. Since 2008, dealers have not been able to keep that much risk on their balance sheets, so someone else has to hold it, for a fee. Short-volatility funds lower the price of protection for everyone. Long-volatility funds supply liquidity in a crash, because they sell puts and VIX calls when they cash in. Relative-value funds keep option prices consistent across strikes, maturities, asset classes, and index and single-stock options. Even VIX exchange-traded products, which lose money steadily when they roll their futures in normal markets, need someone on the other side.

This role shows up in the data. Chen, Joslin and Ni measure dealer constraints from trading in far out-of-the-money index puts. When puts get more expensive while the public buys fewer of them, it is the dealers’ supply that has tightened, and such periods bring more expensive options, higher risk premiums across assets, harder funding conditions and dealers cutting back their risk. Barras and Malkhozov find that the variance risk premium measured in options and the one measured in stocks move together but are not the same; the gap tracks the financial health of banks and dealers and at times exceeds 3% a year, twice the average premium itself. Muravyev shows that market makers’ inventory risk moves option prices more than informed trading does. So part of what a short-volatility fund earns is payment for using scarce dealer capacity. That payment is highest when dealers are constrained, which shows up as more expensive puts with fewer public buyers, and it should fall as income funds, quantitative investment strategies (QIS) and other short-volatility money pile in. That limits the size of the whole category, not just of any one fund.

Whether to use a fund or do it yourself is a practical question: margin, the risk measures known as Greeks, trading in thin far out-of-the-money options, and the discipline to take profits. The difference between good and bad execution of the same idea is the whole business. TAIL’s −44% and Ambrus’s +51.48% since 2022 are both called “buying tail protection.” Demand is also rising: HFR counted 562 hedge fund launches in 2025, a four-year high, as investors added strategies positioned for higher volatility, and Aberdeen’s outlook for the second half of 2025 turned positive on both relative-value and directional volatility strategies.

AQR’s 2011 paper “Chasing Your Own Tail (Risk)” argued that tail insurance costs more than it pays out and offered five alternatives: diversify by risk rather than by asset class, manage volatility actively, add uncorrelated alternative investments, use low-beta (less market-sensitive) stocks, and write a crisis plan in advance. Ilmanen, Thapar, Tummala and Villalon (2021) compared out-of-the-money index puts with trend following across asset classes and found a generally stronger case for trend following, except for investors who mainly fear fast drawdowns.

AQR’s 2022 follow-up showed the pattern. In the short February–March 2020 drawdown, tail-hedging funds and put buying did best; in the long drawdown that began in January 2022, trend following did best, and options-based strategies paid for their crash protection with negative long-run average returns. Harvey, Hoyle, Rattray, Sargaison, Taylor and Van Hemert (2019) found that continuously holding short-dated S&P 500 puts was the most reliable defense and also the most expensive, returning −7.4% over all periods. FactorResearch compared tail-risk funds with cash-like assets from 2007 to 2020 and found similar risk-adjusted returns and drawdowns, with tail funds most in demand when they were least attractive.

Dean Curnutt of Macro Risk Advisors summed it up: tail hedging works, but the price matters a great deal.

Ambrus describes itself as a Chicago-style proprietary trading firm combined with a tail strategy. The flagship Ambrus Volatility Fund launched in May 2022; the firm was incorporated in September 2020 and traded separately managed accounts from January 2021. Co-chief investment officers Kris Sidial (Xanthus Capital, Chimera Securities, nearly four years on BMO’s exotic derivatives desk) and William Wise run the portfolio, and Sal Abbasi (Citadel, 2006–2012) runs quantitative risk. The fund uses no leverage and holds a large cash balance, and Sidial keeps 60% of his liquid net worth in the funds.

The return pattern matches what Sidial told Forbes in 2022 the fund would do: flat to slightly down in normal markets, sharply up in tail events. Split the past 24 months, and the three event months compound to roughly +47% while the other 21 months add about +4% in total. In the data, the fund really does not lose money while it waits for a crash.

The Hedge Fund Journal’s 2024 profile sums up the fund’s mandate in three points: keep the cost of holding protection as low as possible; take profits during volatility spikes, generally when the VIX is above 40; and always keep enough protection to pay off if volatility keeps rising. The portfolio has two parts. The first, which the firm calls loss mitigation, buys protection across VIX-related products (VIX options, index derivatives and exchange-traded products), choosing maturities and strike prices and watching market flows. It is built to pay 100–200 times its cost at the peak, which is what a VIX call bought with the VIX in the mid-teens pays if the index reaches 50–70. The second, called bleed mitigation because it offsets the steady cost of protection, is a set of mostly intraday, risk-limited strategies in standard listed options. They trade on hedging behavior, the term structure, intraday volume and overreactions in single stocks, with win rates and win/loss ratios only slightly better than a coin flip; the firm compares its edge to a casino’s 51%. Most of that profit is spent on more protection. A minimum level of protection stays in place even when trading loses money, and more can be bought when a spike looks close.

The protection mostly sits in listed options with two weeks to two months to expiry, based on the view that short-dated options carry extra volatility from dealer hedging and that large moves since 2017 have played out within six weeks. The fund buys skew and far out-of-the-money options on both sides when they are cheap, prefers single stocks and sectors where positioning is lopsided to plain S&P 500 puts, and accepts being wrong 70–80% of the time. In April 2025 its best trades were out-of-the-money S&P 500 options with one to two months to expiry and out-of-the-money VIX options with one to three months. Its short-volatility positions are always capped with option spreads; its long positions never are. The risk the firm fears most is that volatility rises while far out-of-the-money options fail to follow, so that both sides of the portfolio lose. Decisions on when to sell are judgment calls, weighing hedge fund and dealer positioning, the volatility of the VIX itself and the economic backdrop, but they follow one rule: first sell enough to recover the cost of the trade, then sell the rest in stages. Giving back profit is allowed; losing money on the trade is not.

Size is the variable to watch. In September 2024 the firm had raised about $50 million without a seed investor and put its capacity near $500 million. By spring 2025 firm assets were about $115 million, and its Form ADV reports $485 million as of March 1, 2026, with the main fund at $439.5 million from 157 investors and a $250,000 minimum investment. The April 2026 investor letter said assets were approaching $1 billion, that the fund would stop taking most new money from July 1, and that large multi-strategy hedge funds were among its investors. Wise has said that the bleed-mitigation profits are much harder to scale than the loss-mitigation side.

The firm’s May 2022 paper, Volatility and the Changing Market Structure Driving US Equities, describes four self-reinforcing loops. First, when dealers are “short gamma,” their hedging forces them to buy as prices rise and sell as they fall, which makes moves bigger; when they are “long gamma,” their hedging does the opposite and dampens moves, and that long-gamma position comes from the many funds that sell options near the current price for income. Both exist at once, so markets stay calm until prices move far enough that dealers switch from calming the market to speeding it up; sudden gaps cannot be hedged at all. Second, structured products leave banks with positions that supply volatility in calm markets and need to buy it back when prices approach certain trigger levels. Third, passive, market-cap-weighted flows behave like momentum flows concentrated in a few large stocks. Fourth, rules introduced after 2008 removed the shock absorbers: dealers cannot hold much risk, liquidity comes from trading algorithms that pull back under stress, and the notional value of options traded now exceeds the volume of stock trading. The conclusion is not a crash prediction but a prediction about frequency: events that used to happen once a market cycle will happen about twice as often.

Three other claims are easier to test.

A steady supply of option selling. Ambrus estimates that the amount of short-volatility exposure has doubled since 2018 and assets in derivative-income funds have grown tenfold, with multi-strategy hedge funds and quantitative dispersion trades adding to the pressure. Flows like these move prices: an earlier version of Muravyev’s paper found that rolling positions at expiration creates enough selling pressure to push the prices of other options down 5.7%. Lower implied volatility makes protection cheap in calm periods. When the sellers are forced to cut their positions, they have to buy back, which causes a spike (even hedged short-volatility funds buy some exposure back), and then the same sellers return and push volatility down again. That is why spikes are V-shaped, and why the firm sells into them. But “cheap” needs a qualification. Park finds that a rise in VVIX (the volatility of the VIX) makes out-of-the-money S&P 500 puts and VIX calls more expensive and lowers their returns over the next three to four weeks (a one-standard-deviation rise predicts next-day returns 1.32–2.19 percentage points lower on the puts and 0.68–1.01 lower on the calls), so far out-of-the-money options are cheap when VVIX is low. But a steep term structure, which is normal in calm markets, predicts low returns on variance swaps, VIX futures and straddles. Calm markets can make far out-of-the-money options cheap; they do not make options near the current price cheap.

A VIX more prone to extreme jumps. The firm’s December 2022 paper argued that in the next extreme crash the VIX would reach levels never seen before. The index is calculated from the midpoints of bid and ask quotes across all out-of-the-money strikes, until it reaches two strikes in a row with no bids, and the growth of short-dated trading, plus fewer and larger market makers, had made the 23-to-37-day options it uses less liquid. On August 5, 2024 the VIX jumped from 23.9 to 42 within 15 seconds of the 3:15 a.m. pre-market open and passed 65 by 8:30, while front-month futures stayed below 35. An SEC working paper found that options with lower strikes entered the calculation (the lowest strike included fell from about 2,700 to as low as 1,400) but added at most 6 points. The main cause was inflated midpoints: the gap between the bid and ask prices of the 3,500-strike put widened from about $0.20 to more than $15, and for at-the-money puts from under $1 to $50 or more. The BIS concluded that VIX ETFs and dispersion trades were unlikely to be the main cause. The firm was right about the index level but said nothing about the futures, and that matters for anyone holding VIX calls: the options are priced off the futures, so a spike that appears only in the index pays off only if it lasts past the open.

0DTE options. Ambrus’s April 2023 paper, Dispelling False Narratives About 0DTE Options, was written in response to JPMorgan’s “Volmageddon 2.0” warning about same-day options. It surveyed ten market makers and concluded that institutions (advisers running overlay and income programs, market makers, investors hedging specific events, and volatility funds) dominate the trading. By its account, end users are net sellers on normal days, which leaves dealers long gamma and calms the market; the exception is a panic in which end users rush to buy puts. The paper estimated that on October 13, 2022, dealers would have had to sell more than $10 billion of S&P 500 exposure had the index fallen back to its intraday low. Its conclusion, that 0DTE options speed volatility up but do not create it, has held up. The August 2024 spike happened before the market opened, before that day’s 0DTE options had traded, and an academic study finds that large 0DTE gamma positions do not amplify price moves that have already happened. For a long-volatility firm, the same tool is useful: it can buy protection for the one day that matters instead of paying for a whole month. By February 2026, 0DTE contracts were 63% of SPX volume.

What holds up is the mechanism: dealer hedging, the steady supply of option selling and the VIX calculation, all confirmed in August 2024. What does not hold up yet is the frequency prediction, which four events cannot settle; a theory that explains both long calm periods and sudden bursts says nothing about timing. The trade that remains has two parts. Spikes will be larger relative to the fall in stocks, so own far out-of-the-money options and VIX calls. They will also be shorter, so sell above a VIX of 40 or, based on the VVIX evidence, sell when the volatility of the VIX is high rather than at a fixed VIX level.

Sidial’s own framework: divide the expected yearly running cost by the expected return in a crash; define the crash in advance (a 20% monthly fall in the S&P 500, or a VIX near 70); judge the fund like a goalkeeper, on saves rather than goals; hold it all the time; keep the position small; and measure returns on the investor’s total allocation, not on the capital inside one trade.

Universa, at 3.33% of a portfolio, added roughly 13 percentage points in March 2020 (+0.4% for the hedged portfolio against −12.4% unhedged), at a running cost of about 0.5% a year of the amount protected. Ambrus made 25.21% on its own capital in April 2025 at a running cost near zero, so an investor holding it at 10% of a portfolio received about 2.5 points. Per dollar of running cost, Ambrus wins. Per dollar invested, Universa’s crash payout is roughly five times larger, and the gap would grow in a 2008-style decline that lasts for months, because the rule of selling above a VIX of 40 and keeping only a minimum level of protection limits the payout, while pure protection keeps paying as the fall deepens. These are rough calculations on reported figures, not a like-for-like backtest. What a zero running cost buys is the ability to hold the strategy on its own, rather than as a cost to the rest of the portfolio.

This trade-off is not specific to Ambrus; it is how the market prices risk. If only sharp moves that actually happen carry a premium, any position that pays out by contract in a sudden crash has to cost something. So a long-volatility fund with no running cost has to use one or more of three things: protection that mostly profits from rises in implied volatility (vega), which is cheap or even earns a premium but does not pay out by contract in a sudden crash; a separate trading profit that pays for the protection; or a second premium earned on the side, such as selling volatility with capped risk. Ask the fund to show the three separately. The same evidence tests the view about short-dated options. On average, short-dated out-of-the-money options are the most expensive point on the curve, so if short-dated options really carry extra volatility, the contracts Ambrus picks have to beat that average. The way to check is to look at the long positions’ realized volatility minus implied volatility on their own, before the profits from bleed mitigation.

Two questions remain open. The first is robustness. The bleed-mitigation strategies have been measured as uncorrelated in calm markets, but whether intraday profits survive several days of price gaps with very wide bid-ask spreads is untested, and if they failed, the fund would stop covering its costs exactly when its protection is paying out. If part of those profits comes from the day–night pattern (options cheap at the open, expensive at the close), they rest on a documented pricing anomaly rather than a casino’s 51%, and it is fair to ask how much of the profit that pattern explains. The second is capacity: a firm that estimated its capacity at $500 million in 2024 is approaching $1 billion in 2026. If bleed mitigation cannot grow with it, either the short-volatility side grows and the fund drifts toward being a short-volatility fund, or it does not and the fund goes back to paying a running cost like an ordinary tail fund. The test is whether a year like 2023 still comes in at +1.7% with $1 billion under management.