A Map of Biotech Investing: Where You Earn Only If the Drug Works, and Where You Earn Whichever Drug Works

Connecting context: Two days ago, a beginner’s map of eight US biotech trends drew the big picture of money moving from semiconductors into biotech and the Korean companies connected to it. Before that, May’s sector thesis argued that this is not a market for buying good technology but one for picking stocks the market has underpriced, and June’s six-stock check argued that a catalyst being alive and a stock being cheap are not the same thing. This piece redraws the eight-trend map along the axis of how revenue is earned and goes one level deeper, asking the same questions: where does the money get made, does it recur, and how much of that is already reflected in today’s price.

TL;DR

• Sorting biotech by disease alone does not let you compare risk. You have to overlay four axes: disease, treatment modality, value chain, and economics, and the axis that divides risk is economics, meaning whether the payoff is a one-time milestone tied to a single trial or recurring revenue per patient.
• Personalized oncology is a theme where therapeutics, diagnostics, recurrence monitoring, and data all fold into one pipeline. On August 19, Moderna and Merck’s Phase 3 trial of their individualized cancer vaccine met both of its endpoints, recurrence-free survival and distant metastasis-free survival, and Moderna’s stock jumped 177% in a single day. [Fact: company statements, market prices] However, a claim that circulates often in early drafts, that “1,137 patients had a customized therapeutic manufactured and delivered to them on schedule,” is wrong. The 1,137 patients are the total randomized on a 2-to-1 basis, actual dosing covered roughly 758 patients, and manufacturing success rates, hazard ratios, and overall survival have not been disclosed. [Fact: company statements]
• The economics differ at every stage of the pipeline. Therapeutics pay off big if they succeed, but the trial outcome is binary. Recurrence monitoring tests generate recurring revenue that repeats many times per patient, and data is a high margin, long term contract business. This difference is what defines the investment structure.
• Results on the recurring revenue axis back up the structure. Natera’s second-quarter revenue grew 37.7%, oncology testing grew 57.2%, and the company guided to positive cash flow in 2026. Guardant’s revenue grew 44%, but the company plans to burn roughly $200 million in cash this year. Tempus posted its first profitable quarter, but that quarter included a $98.5 million unrealized gain. [Fact: each company’s earnings release]
• Saying the structure is good and saying it is cheap are two different claims. Natera and Guardant trade at 16x to 17x their 2026 revenue guidance, and Twist trades at around 20x, while BioNTech, Tempus, and Natera have risen 28%, 56%, and 63% respectively over the past three months. [Fact: arithmetic based on August 21 closing prices]
• For validated treatment platforms, the problem is price. Enhertu posted first-half revenue of $2.96 billion, up 29%, and in radiopharmaceuticals, Curium agreed to acquire Lantheus for up to $8 billion. This is a zone where you need to pick by asset-level metrics rather than by platform name. [Fact: company statements]
• The axis that most needs proof has, paradoxically, risen the most this year. No AI-discovered drug has been approved yet, but AbCellera is up 231% year to date, 10x Genomics is up 299%, and Twist is up 359%. [Fact: August 21 market prices] This is a place where price moved ahead of proof.
• The conclusion for portfolio structure is a barbell. Hold companies that only earn if the drug works separately from companies that earn no matter which drug works, but at current prices, the leading names on each axis have all just rallied sharply, so entry-price discipline comes first.

1. What Disease Classification Misses: Overlaying Four Axes

Classifying biotech as oncology, obesity, or rare disease only tells you what illness a drug targets. What investors need to know is what happens to the company if the drug fails, and what shape the revenue takes if it succeeds. So we add more axes.

Disease breaks down into oncology, immunology, metabolic disease, neurology, and rare disease. Treatment modality breaks down into antibodies, antibody-drug conjugates (ADCs), bispecifics, cell, gene, and RNA therapeutics, and individualized treatments made separately for each patient. The value chain runs from discovery through patient selection, treatment, manufacturing, outcome tracking, and data. Economics splits between a one-time milestone tied to a single trial outcome and recurring revenue that a single patient generates repeatedly.

Of these four axes, the one that divides risk is the last one. Two oncology companies can work on the same patients, and one can lose half its value on the day trial results come out while the other keeps earning test revenue for as long as that patient is being treated. Personalized oncology, the subject of this piece, is a theme where all four axes overlap in a single pipeline: the disease is oncology, the modality is individualized treatment, the value chain runs from selection to data, and the economics combine milestones with recurring revenue.

2. A Map of Fourteen Sub-Themes

The sub-themes are organized below by where value is created and by a current read on each. The representative names are illustrative examples for analysis.

[Fact: company statements and news reports, as of August 21, 2026. Assessments are our own]

The map’s horizontal axis is how far commercial validation has progressed, and the vertical axis is whether revenue is one-time or recurring. The upper right is the best position and the lower left is the riskiest. The coordinates are our own judgment, not a precise measurement. But one thing reads true regardless of the coordinates: the upper right is already expensive, and the lower left has risen the most this year.

3. The Internal Structure of Personalized Oncology: Economics Differ at Every Stage

Personalized oncology is a circular structure that starts with one patient’s tumor and loops back into the treatment design for the next patient.

Tumor tissue and blood are drawn from the patient and the genome is analyzed to select the neoantigens unique to that patient’s cancer. A therapeutic is designed and manufactured around those antigens and given in combination with an immuno-oncology drug. Once treatment ends, minimal residual disease (MRD) is measured repeatedly in the blood to monitor for recurrence, and those results accumulate as data that refines patient selection and trial design for the next patient.

Who makes money at each stage, and what shape that money takes, differs by stage.

The therapeutic stage rides on a single Phase 3 trial. The recurrence monitoring stage repeats every quarter for as long as the patient is alive, and the more the therapeutic succeeds, the more patients there are to monitor. That the shape of risk can differ this much within a single theme is the starting point of this piece.

4. Results on the Recurring Revenue Axis: Good Structure, Expensive Price

Here are the second-quarter results for the recurring revenue axis.

Natera’s second-quarter revenue was $752.8 million, up 37.7%, and oncology testing volume was 296,700 tests, up 57.2%. Quarterly test volume topped one million for the second straight quarter, and gross margin was 64.5%. Signatera’s average selling price per test is about $1,275, and the company has set a long-term target of around $2,000. It raised its full-year revenue guidance to a range of $2.85 billion to $2.91 billion and said it expects positive cash flow in 2026. [Fact: company statements] There is no specific dollar guidance for that figure.

Guardant’s second-quarter revenue was $335 million, up 44%. Oncology testing revenue grew 38%, and Shield, its colorectal cancer screening test, generated $52.9 million, 3.6 times what it generated a year earlier. But quarterly free cash flow was an outflow of $69.5 million, and the company guided to a 2026 free cash flow outflow of $195 million to $205 million, actually widening the range of the outflow. [Fact: company statements] In effect, the revenue guidance and the cash outflow guidance both went up together.

Tempus’s second-quarter revenue was $382.5 million, up 22%, and adjusted EBITDA was a positive $8 million. Its GAAP net income of $5.6 million marked the company’s first profitable quarter, but that figure includes a $98.5 million unrealized gain on securities, offset in part by $55.6 million in stock-based compensation. [Fact: company statements] In other words, the profit came from an unrealized gain rather than from operations. Full-year revenue guidance is $1.595 billion to $1.605 billion, with adjusted EBITDA of around $65 million. On July 20, Tempus agreed to acquire Personalis for $16.25 per share, an enterprise value of roughly $1.5 billion net of Personalis’s existing stake, with the deal expected to close late this year or early next year. [Fact: company statements] Tempus’s own MRD test, xM, runs about 9,000 tests per quarter.

Twist’s fiscal third-quarter revenue was $118.38 million, up 23.2%, with a gross margin of 52.8%. The company guided to 21% full-year growth and breakeven adjusted EBITDA in the fourth quarter. On August 4 it raised $300 million in an equity offering at $96 per share, and its August 21 closing price was $145.59. [Fact: company statements, market prices] That is 52% above the offering price in just two weeks.

Now look at the price the market has paid for these results.

Dividing market cap by the midpoint of each company’s 2026 revenue guidance gives 16.6x for Natera, 17.0x for Guardant, 8.2x for Tempus, and around 20x for Twist. This is a simple revenue multiple that does not net out cash. [Fact: our own arithmetic based on August 21 closing prices] Tempus looks cheapest because its revenue is large relative to its market cap, not because the market has bid it up less. Tempus is up 56% over the past three months.

The recurring revenue structure has weaknesses too. Test revenue is set by reimbursement, and any increase in price per test has to be preceded by broader coverage. Competition is heavy as well. Beyond Natera, Guardant, Tempus, and Personalis, Exact Sciences and Roche’s Foundation Medicine compete for the same patients. And as Guardant shows, revenue can grow 44% while cash keeps going out the door. Judging the structure to be good and judging 16x to 17x to be cheap are two separate calls.

5. The Therapeutics Axis: What the Phase 3 Trial Proved, and What It Has Not

On August 19, Merck and Moderna announced that the Phase 3 trial (INTerpath-001) of their individualized neoantigen therapeutic intismeran autogene in combination with Keytruda met both of its endpoints, recurrence-free survival and distant metastasis-free survival, in patients with resected melanoma. The trial enrolled 1,137 patients, randomized 2 to 1 between the combination arm and the control arm. Moderna’s stock rose 176.97% that day to close at $174.38, then fell 19.8% the next day, while Merck rose by around 11%. [Fact: company statements, market prices]

Here is precisely what the Phase 3 trial proved. A different therapeutic was manufactured for each patient, and a trial of Phase 3 scale was carried through to completion, with the direction of effect matching Phase 2. In the five-year follow-up of the Phase 2 trial KEYNOTE-942, the hazard ratio for recurrence was 0.51 (confidence interval 0.29 to 0.89), and the hazard ratio for distant metastasis was 0.41. [Fact: ASCO 2026, JCO] The Phase 3 hazard ratio will not be disclosed until it is presented at a medical conference.

Here is what it did not prove. With a 2-to-1 allocation, only about 758 patients actually received intismeran. The claim that all 1,137 patients received a customized therapeutic is wrong, and neither company has ever disclosed the share of patient-specific manufacturing runs completed on schedule, in other words the manufacturing success rate. [Fact: review of company statements] Absolute efficacy by disease stage, overall survival, price per patient, and reimbursement are all still undisclosed. Moderna’s personalized therapeutics plant in Marlborough, built for more than $322 million, began supplying clinical trial material in September 2025, and the company says approval could come as early as 2027. [Fact: company materials] Work in other cancer types is underway but uneven. Two Phase 3 trials are ongoing in non-small cell lung cancer, while the cutaneous squamous cell carcinoma program was discontinued after Keytruda failed on its own in that indication. [Fact: trial registries, news reports]

The same Phase 3 success means something different to each company. Moderna’s second-quarter revenue was $145 million, its net loss was $782 million, and it holds $6.9 billion in cash while cutting annual cash costs to around $4 billion. [Fact: company statements] If approval comes in 2027, the point at which the Phase 3 success changes the P&L is later still. For Merck, Keytruda’s core composition-of-matter patent expires in December 2028, with more than $25 billion in annual revenue riding on it. [Fact: news reports] A successful combination therapeutic is a way to push that cliff back. The reason Merck’s stock moved is that this Phase 3 trial became the first answer to the question of what Keytruda’s revenue looks like after 2028.

BioNTech is the name with the thickest downside cushion on this axis. At the end of June, its cash and securities stood at €16.63 billion, more than half its $29.3 billion market cap. [Fact: company statements, August 21 market cap] Its late-stage pipeline includes BNT327, the PD-L1/VEGF bispecific it shares with BMS, which is running seven pivotal trials, and gotistobart’s overall survival hazard ratio, reported in March, was 0.46. There are caveats, though. 2026 revenue guidance was cut to a range of €1.6 billion to €1.9 billion, and the US approval filing for the HER2 ADC BNT323 has not yet been submitted, though it is under review in China. Autogene cevumeran, the individualized cancer vaccine it is developing with Roche, has an ongoing Phase 2 trial in pancreatic cancer, but colorectal cancer results have slipped to 2027 and the urothelial cancer program was discontinued. [Fact: company statements, news reports] The case for good asymmetry rests on the cash cushioning the downside and the multiple late-stage trials, not on having an already-approved oncology drug.

6. The Price Problem for Validated Platforms: ADCs, Bispecifics, and Radiopharmaceuticals

On this axis, the odds of clinical success and the path to commercialization are relatively clear. Enhertu’s first-half revenue was $2.96 billion, up 29%, and Datroway and Trodelvy were approved in May and June respectively for first-line triple-negative breast cancer. Trodelvy’s first-half revenue was $859 million, up 31%. Among bispecifics, Amgen’s Imdelltra generated $288 million in the second quarter, up 115%, and in a second-line small cell lung cancer Phase 3 trial it showed median overall survival of 13.6 months versus 8.3 months, a hazard ratio of 0.60. Genmab’s Epkinly generated $312 million in the first half, up 48%. In radiopharmaceuticals, Novartis’s Pluvicto generated $1.994 billion in 2025, up 43%, and on July 31 its label was expanded to metastatic hormone-sensitive prostate cancer. [Fact: company statements]

Being validated also means being priced in. From the start of 2026 through August 18, there were 69 biopharma acquisitions with disclosed values totaling $189.1 billion, and on August 3, Curium agreed to acquire Lantheus for up to $8 billion including contingent consideration. That is a step up from earlier radiopharmaceutical deals: BMS’s $4.1 billion acquisition of RayzeBio, AstraZeneca’s $2.4 billion acquisition of Fusion, and Eli Lilly’s $1.4 billion acquisition of Point Biopharma. [Fact: deal tracking, company statements] Large companies buy the best assets first. Picking among what is left requires looking at asset-level metrics rather than platform names, and by 2026 each of those metrics had a real-world example attached to it.

[Fact: company statements, FDA, news reports]

The last row captures what defines this axis. A bispecific can be a good drug and still get stuck on a facility inspection, and a radiopharmaceutical trial stops if the isotope supply runs out. That is why Novartis is building an isotope production facility in Indianapolis and adding a third manufacturing site in California. [Fact: company statements] On this axis, the moat sits in the supply chain, not in the molecule.

7. The Axis That Still Needs Proof, Yet Has Risen the Most

AI drug discovery, synthetic biology, solid-tumor cell therapy, and parts of gene editing and RNA are technologically interesting, but the revenue moat has not been proven.

Start with AI drug discovery. As of August 22, no drug has been approved that is credited to an AI platform’s discovery. [Fact: review of regulatory records] According to one published tally, of 117 AI-derived drug candidates, 51.3% completed Phase 1 and 6.8% completed Phase 2, while another tally puts the Phase 1 success rate at 80% to 90%, falling to around 40% in Phase 2. [Fact: conference and consulting tallies, data from 2023 to 2025] If Phase 1 passes easily but candidates stall in Phase 2, the problem lies not in molecule design but in target selection and disease understanding. [Inference: reading of the tallies]

Corporate results point the same way. Recursion’s second-quarter revenue was $7.67 million, below expectations, with a net loss of $131 million and cash of $556.8 million, enough to last into early 2028. Schrödinger’s software revenue was $200 million in 2025, up 11%, and AbCellera received $84 million in upfront payments from Jazz and Vertex in the second quarter. Alphabet’s Isomorphic Labs raised $2.1 billion in May, but as of the start of the year it was still only at the stage of saying its target was to dose its first patient sometime within the year. [Fact: company statements, news reports] The conclusion that what matters is clinical entry, cash milestones, and repeat customer revenue, not technology announcements and partnership headline numbers, comes from this evidence.

Solid-tumor cell therapy has no US approval. The world’s first solid-tumor CAR-T approval went to China’s CARsgen on June 22, and autoimmune CAR-T maker Kyverna is targeting a completed approval filing in the fourth quarter of this year. The existing blood-cancer CAR-T business is stagnant. Gilead’s cell therapy revenue was $417 million in the second quarter, down 14%. Instead, acquisitions have piled into in vivo CAR-T technology since 2025: AbbVie for up to $2.1 billion, BMS for about $1.5 billion, and AstraZeneca for up to $1 billion. [Fact: company statements, news reports] This is an attempt to solve the manufacturing time and cost bottleneck inside the patient’s body rather than in a factory, and it is still at the stage of first clinical data.

In gene editing, only ex vivo editing is validated. Casgevy had 301 patients start treatment in 2025, of whom 64 were dosed, generating $116 million in revenue at a list price of $2.2 million. In vivo editing saw Intellia’s clinical hold, imposed after a patient death in November 2025, lifted in January and March of this year, and in February the FDA issued draft guidance on a new approval pathway for patient-customized editing therapies. [Fact: company statements, FDA] In RNA, siRNA has built a commercial model, yet its leading company’s stock is down 40.6% this year. Alnylam turned profitable in 2025, with revenue of $3.71 billion and net income of $313.7 million, but cut its 2026 guidance on July 30. [Fact: company statements, market prices] This is a case where having an established commercial model and having a strong stock are two different things.

Yet looking at stock prices on this axis, proof and price are in reverse order. As of August 21, year-to-date gains stand at 359% for Twist, 299% for 10x Genomics, 231% for AbCellera, and 67% for Illumina. [Fact: market prices] Tools and infrastructure companies have risen far more than biotech as a whole (the XBI is up 35.9%). Illumina’s second-quarter revenue grew 9.5% after China’s export ban was lifted in November 2025, prompting it to raise full-year guidance, and 10x Genomics saw its spatial analysis consumables grow 16%. The recovery is real. But PacBio’s second-quarter revenue was flat, leading it to cut full-year guidance and push its breakeven target out to 2028. [Fact: company statements] Within the same axis, the recovery is uneven, yet price moved first. This is the place where entry-price discipline applies most mechanically.

8. Overall Priority Ranking and the Barbell

Here is a ranking of investment structures across biotech today. The ranking reflects structural quality, not a signal to buy now.

From an equity standpoint, the conclusion is a barbell. Judge and hold separately the companies with the greatest odds of therapeutic success and the companies that make money no matter which therapeutic succeeds. Treatment optionality names like BioNTech and Moderna and diagnostics and data recurring-revenue names like Natera and Tempus sit within the same theme but are different assets. The value of the former is decided on the day trial results come out, while the value of the latter builds up through patient counts and reimbursement. Buying both under a single “personalized oncology” basket means buying without understanding the shape of the risk.

At current prices, the leading name on each axis shows something different. What BioNTech shows is asymmetry backed by cash worth more than half its market cap. Tempus shows a business position that connects diagnostics, data, and MRD within a single company, while Natera shows fundamental quality, with test volume, margin, and cash flow all improving together. And all three stocks have risen 28%, 56%, and 63% respectively over the past three months, with those gains concentrated around August’s earnings and the Phase 3 announcement. [Fact: August 21 market prices] The XBI is up 35.9% year to date and sits 2.3% below its 52-week high. At the end of July it was up 20.7% year to date, so more than half of that gain came in the past three weeks alone. [Fact: fund manager data, market prices] The ranking by structure and the ranking by entry point are different things. Right now, the latter beats the former.

Laying the same map over Korean biotech reveals the connection. Among the names this blog has covered, LigaChem Biosciences sits on the ADC axis, Alteogen sits on the recurring revenue axis through subcutaneous-conversion royalties, Lunit sits on the AI-based biomarker axis, and Samsung Biologics and ST Pharm sit on the manufacturing axis that gets volume regardless of which drug succeeds. June’s verdict was that the good companies among them were already in expensive territory. That matches the conclusion from the US map. The market recognizes first which positions earn money no matter which drug succeeds, and investors arrive only after the price in that position has already priced in most of the structural advantage.

The names mentioned in this piece are illustrative examples for analysis and do not constitute a recommendation to buy or sell any specific stock. Responsibility for investment decisions and their outcomes rests with the investor. Prices and market capitalizations are as of US market close on August 21, 2026, and revenue multiples are market cap divided by the midpoint of each company’s 2026 revenue guidance (fiscal year 2026 for Twist), without netting out cash. The coordinates on the theme map reflect our own judgment. The Phase 3 trial’s hazard ratio, overall survival, and manufacturing success rate have not been disclosed by the companies, and the contribution of the Phase 3 announcement to BioNTech’s share price gain cannot be isolated. Certain items, including Alnylam’s second-quarter revenue by product and the most recent year in the AI candidate tally, relied on secondary sources.

Related posts

• What’s Happening in US Biotech? A Beginner’s Map of 8 Big Trends and Their Korean Connections
• Korean Biotech Sector Investment Thesis: Not a ‘Buy Good Tech’ Market but a ‘Buy What the Market Has Underpriced’ Market
• Six Korean Bio Stocks, Made Simple: ‘The Catalyst Is Alive’ and ‘It’s Cheap’ Are Not the Same Thing