Anthropic Is Selling Compute - These Biotechs Are Selling What Compute Cannot Buy
Wall Street knows how to price a bigger computer or a chip. It even knows how to price a Huang / Huang affair. It is less comfortable pricing a freezer full of tumors until someone figures out what to do with them.
We get these disconnects in Wall Street all the time, and those that can spot them early can reap the benefit of extraordinary returns. As Anthropic approaches its public-market debut some investors are going to ask what the next act for Anthropic (NASDAQ: ANT) or AI for that matter? Is the talking point going to be more compute, or will other ideas enter the narrative? Investors need to be prepared because on November 19th that the Claude developer could price its IPO above the indicated $70 - $78 range.
The company’s growth is almost viral. Anthropic reported $33.24 billion in revenue for the first nine months of 2026, according to the prospectus. This is not a pre-revenue science project wrapped in an AI presentation anymore, it's a trillion dollar IPO.
But the IPO raises a question that extends beyond chatbots and coding assistants, what’s next? Is AI going to cure cancer? According to the CEO of Anthropic it will, but investors want more than big talk. They want to know how all this computing capacity is going to be monetized. Are future valuations derived from those who own the most computing capacity—or whoever controls information that the computers cannot simply invent?
Cures for cancer are coming, but not from Anthropic. When it comes to biotech discovery there is still no computer substitute for research, and those that control it, possess only what computers can hallucinate about. On a November 10th video message Dario Amodei, CEO of Anthropic predicted that
artificial intelligence could make cancer, Alzheimer’s disease, heart disease, and mental illness “relics of the past” in less than a decade.
Find out what compute cannot buy.
MEDS: The Asset You Cannot Download
DataMEDS AI (NASDAQ: MEDS) announced today, through its Helomics subsidiary, that it is partnering with Canadian drug-discovery company Diagen AI. This partnership is expected to combine specialized algorithms with a biobank containing more than 70,000 primary tumor biopsies, including more than 50,000 gynecologic cancer samples. These sample counts sound big, but it’s not about the sample count, it's about the access to test new drugs on this massive library of tissue.
Helomics brings patient-derived tumor models, laboratory capabilities, and side effect observations. Diagen brings a computational discovery platform. The combination allows researchers to move between predictions and experiment. Under their new partnership, they can identify a drug candidate, expose tumor material to it, measure the response, and use the findings to inform the next decision. This breaks the paradigm of accepting a drug that works okay versus one that is fine tuned.
One of the shortcomings of the Large Language Models (LLMs) is that they can read what researchers have published, but they cannot conjure an experiment that has never been performed. No amount of processing capacity grants access to privately controlled tissue, associated annotations, research permissions, or laboratory workflows. While the biobank is a very unique asset and one of the largest repositories of tissue, there are risks and limitations of its use. There are ethics regarding use of the tissue, sample quality, and a limited supply. The biobank highlights the shift from renting intelligence to owning the inputs that make intelligence useful.
Small Models Can Be More Effective
When we think of traditional “AI drug discovery" we are imaging the models trying every single input imaginable and basically leaving no stone unturned. It’s an inclusive model that wants to look at everything. This thinking is woefully inefficient, because there is a better way.
Research co-authored by Diagen’s Mohit Pandey and published in Nature Communications describes a reinforcement-learning approach to screening large peptide libraries. The method reduced the search space by more than 90%. Then the researchers experimentally evaluated 105 selected candidates, identifying 15 with activity against a triple-negative breast cancer cell line.
The researchers prioritized task-specific models and molecular descriptors over computationally intensive large-language-model architectures they considered impractical for that particular task. This is a different way of looking at things versus how Claude may have been designed. While Claude may be the largest general-purpose model, it might not be the best tool for biotech discovery. It’s like searching for gold, a larger excavator can move more dirt, but a skilled geologist can tell you where to dig. The advantage comes not from processing everything, but from knowing what deserves a closer look.
Collaboration Cash Neutral
The Helomics collaboration is designed to investigate mechanisms, test the peptides against primary tumors, and pursue small-molecule candidates and subsequent animal validation. The deal itself is cash neutral and is a win-win for both companies. Each company owns a piece of the other that amounts to roughly 10%. For MEDS investors it means the issuance of 630,000 MEDS shares to Diagen, and approximately $1.5 million of expected transaction/program value through specified milestones. This is cash neutral because Helomics was going to spend money on their biobank independently from Diagen.
This is a transformative deal for MEDS shareholders, who now have the upside of validated drug discovery in cancer without bearing any of the development costs. While the opportunity is early-stage they are developing intellectual property at an accelerated pace, which makes this an intriguing investment with significant upside and few of the drug developers risks or dilution.
SELLAS Just Reminded the Market What Matters
While investors wait for the Anthropic IPO to price, the regulators are hard at work analyzing the survival data of oncology drugs. Regulators want to know if drugs extend human life and as good as the LLMs are at predicting what we will say their models have yet to account for the human effect tied to the FDA decision making process.
On November 17, SELLAS Life Sciences (NASDAQ: SLS) announced that its randomized Phase 3 REGAL trial met its primary overall-survival endpoint. Its investigational cancer vaccine, galinpepimut-S (GPS), produced median overall survival of 24 months versus 14.7 months with best available treatment, with a reported hazard ratio of 0.59 and p-value of 0.035.The study involved AML patients in second or later remission who could not undergo allogeneic stem-cell transplantation.
SELLAS plans FDA discussions and preparations for a biologics license application in the first half of 2027. Its separate CDK9 inhibitor, tambiciclib/SLS009, is another investigational program that is in the works.
This is a classic example of how drug discovery is supposed to work and how investors are ultimately rewarded. In software, a useful answer can be monetized immediately, but in medicine, the answer has to survive biology.
Regulatory Breakthroughs Can Reprice a Biotech Overnight
Wall Street does not need to wait for Anthropic to cure cancer in 10 years before it can make money from biotechnology. Sometimes the catalyst is considerably simpler like a regulator who changes its assessment, and a company the market had discounted suddenly looks commercially viable.
Replimune (NASDAQ: REPL) provided a dramatic example on July 31. Its shares more than doubled in premarket trading after an FDA advisory committee voted 10–3 in favor of the evidence supporting its experimental melanoma treatment, RP1. The company had been worth approximately $455 million at the previous close then topped out at $1.25 billion a week later when the drug received its approval. One favorable regulatory development was enough to trigger a wholesale reassessment of the business.The company had moved from arguing that its evidence deserved consideration to preparing to sell its first approved product.
uniQure (NASDAQ: QURE) demonstrated that even an improved route to an application can unlock substantial value. On June 17, shares surged approximately 80% in premarket trading after the FDA reversed an earlier position and opened a path for the company to seek accelerated approval of its Huntington’s disease gene therapy, AMT-130. By midday, the stock was still up nearly 80%, around $48. This was not a drug approval—it was a major reduction in the uncertainty surrounding how the company could pursue one.
An oncology platform does not have to become the next Anthropic to create substantial value. It may need to generate an asset that a pharmaceutical company cannot afford to ignore. That normally gets the job done.
CAPR: Another Test of Biology Over Narrative
Capricor Therapeutics (NASDAQ: CAPR) offers a different near-term test of what some would call regulatory overreach except of course the hack analyst Adam Feuerstein and Pharma Bro, Martin Shkreli who continues to predict their demise.
Its investigational cardiac-derived cell therapy, deramiocel, remains under FDA review for Duchenne muscular dystrophy (DMD), with a November 22 target action date. The company reported statistically significant slowing of upper-limb disease progression in its randomized HOPE-3 study. Deramiocel is a cell therapy but it holds great promise of ushering in a new class of biologic drugs called exosomes. The underlying mechanism behind their DMD treatment is based on heartlike cells lodging in the lungs and producing exosomes that keep the heart healthy. This mechanism pushes beyond the simplistic idea of cells becoming replacement tissue but bioreactors for exosomes. Capricor demonstrated that the cells release exosomes that influence immune activity and fibrosis. Deramiocel itself remains a cell therapy, not an approved purified-exosome product, but an approval will increase awareness in the space.
Aruna Bio & NurExone Biologic: Aruna Bio leads neural stem cell-derived exosomes (AB126) capable of naturally crossing the Blood-Brain Barrier (BBB) to treat stroke, ALS, and neurodegenerative inflammation. Similarly, NurExone Biologic Inc. (TSXV: NRX / OTC: NRXBF) is advancing its ExoTherapy platform, utilizing targeted siRNA-loaded exosomes to treat acute spinal cord injuries and optic nerve trauma, earning valuable FDA Orphan Drug Designations.
A number of pure exosome plays are likely to benefit like Pluris (NASDAQ: PLUR) and ZEO ScientifiX (OTCID: ZEOX) who combined to start selling commercially available exosomes for the regenerative medicine market.
One interesting private company with technology to make cGMP exosomes at a fraction of the cost of all the big players like Lonza and Rion in the space using a machine that solves the manufacturing problem that stymied the commercialization of exosomes for the past decade. ProgeniX Bio, a privately held biotech company, is developing ExoFlow, an automated platform designed to lower the cost of continuous, scalable exosome production. The company is working toward cGMP-compatible manufacturing and is also advancing a preclinical exosome-based therapy for glioblastoma. After a CAPR approval they might be sitting on some of the hottest IP in the world.
Next to peptides, exosomes represent some of the hottest science in biotech, but until there was a solution for manufacturing at scale these therapies are sidelined no matter how much computing power ANT can muster. There's a simple reason why AI hasn't figured out how to make exosomes at scale. Exosomes are on the fringe of science and don’t naturally fit in the LLM narrative until of course CAPR gets an approval.
Can Small Biotechs Actually “Beat” Anthropic?
One of the most influential CEO’s Dario Amodei claimed AI will have an answer for cancer in the next decade. The thing is that means the cure for cancer is already here because it takes that long for the regulatory process to catch up. This means the small biotechs are beating Anthropic right now. It also means that companies like Helomics that have biotech development and compute have an edge over traditional biotechs and can accelerate their regulatory path.
We keep throwing research dollars at cancer for example, but that's not what is needed. The Diagen model showed us that the way we screen molecules matters. Raw compute doesn’t necessarily mean an advantage. There are so many Biotechs out there developing marginal drugs and the truth is they are suffering slow death. If you look at small cap biotechs it looks like a COVID ward during the height of the pandemic. Patients are done with taking okay drugs and taking drugs to treat their side effects. They are looking for real answers and the iterative process Diagen is using can be applied to all biotechs. Drugs need to improve by orders of magnitude if not they should be eliminated from consideration.
Shrinking a tumor is only part of the battle.Metastasis which is the spread to distant organs is the cause of most cancer deaths.CytoDyn (OTCQB: CYDY) is investigating whether leronlimab, its CCR5-targeting antibody, can disrupt signaling involved in cancer-cell migration and the tumor microenvironment. Think of it as targeting the tumor’s communications network rather than just its largest visible mass.
In cancer drug screening a well-characterized tumor biobank offers significant advantages compared to a biotech testing candidates against patient-derived models. The sooner that weak compounds are found and discarded the more efficient the development process becomes. It's very important to identify responsive tumor types, and bring better-supported candidates into clinical trials. There is no shortcut around safety testing or FDA review, but the acceleration comes from making fewer wrong turns before reaching the regulator. While compute can generate possibilities; a biobank can help determine which ones deserve the time and capital required to become medicines.
The Scarce Asset Is Not Another Prediction
Anthropic can help scientists work faster, but computing power alone does not provide access to patient tissue or replace the experiments and clinical trials needed to prove a treatment works.
That is the opportunity behind MEDS and the Helomics–Diagen collaboration. They are combining specialized AI with a curated tumor biobank and laboratory capabilities to generate evidence competitors cannot simply download.
A biobank cannot guarantee approval, but it can help researchers discard weak candidates earlier and concentrate resources on stronger ones. The potential advantage is fewer wrong turns because there are no shortcuts to clinical testing
Investors do not need to bet against Anthropic to recognize that opportunity. As models become more capable, proprietary biological inputs and experimental validation may become more valuable. Wall Street is paying for the machine that generates possibilities. The overlooked opportunity may be the company that can prove which ones work.

