Every month a new press release claims AI is about to end cancer. Every month the survival rates for pancreatic, brain, and ovarian cancer barely move. According to the National Cancer Institute, the five year survival rate for pancreatic cancer sits at just 13%. AI has been in oncology labs for five years now. That number has not budged.
What Is Actually Happening in Cancer Research Right Now
The hype is real. The results are not keeping up.
In 2026, over 400 AI drug discovery companies are operating worldwide, according to Deep Pharma Intelligence. Venture capital has poured more than $50 billion into AI biotech over the past four years, according to CB Insights. Yet the FDA approved just two drugs last year that were developed primarily through AI assisted processes. Two.
The problem is not the technology. The problem is what the technology is being pointed at.
Xaira Therapeutics, a startup that raised $1 billion at launch in 2024, is making a different argument. They say AI cannot cure cancer the way most companies are trying to use it. You cannot feed a model bad data and expect a miracle. You cannot skip the biology. And you cannot pretend that 12 to 15 years of clinical trial time disappears because your model runs fast.
According to a 2025 report from Deloitte, the average cost to bring one approved drug to market is $2.3 billion and takes an average of 12.5 years. AI can compress parts of the discovery phase. It cannot compress FDA review. It cannot compress phase three trials with 50,000 patients. That process exists because cancer is not one disease. It is over 200 diseases, and they behave differently in every single body.
The Real Problem With How AI Is Being Used in Cancer Research
Here is my honest take. Most AI cancer companies are not trying to cure cancer. They are trying to raise a Series B.
There is a difference between finding a molecule that looks good in a model and finding a molecule that keeps a human alive. According to the MIT Center for Biomedical Innovation, over 90% of cancer drugs that enter clinical trials fail. That number has barely changed in 20 years. AI has not fixed this yet.
What the best companies are doing differently is asking a different question. Instead of “can AI find a new drug faster,” they are asking “can AI help us understand why drugs fail.” That is the angle that matters.
Companies like Owkin and Recursion Pharmaceuticals are building models trained on real patient data, including pathology slides, genomic sequences, and treatment outcomes. They are not trying to replace chemists. They are trying to find the patterns that chemists miss after year ten of staring at the same data.
The rich versus poor version of this story goes like this. The average investor sees “AI” and “cancer” in the same headline and clicks buy. The sharp operator asks who controls the data, how many drug approvals they have, and what their burn rate looks like. One of those investors is going to get rich. The other is going to get a press release and a tax loss.
If you are building content around biotech investing or trying to explain complex research to a broader audience, InVideo AI lets you turn dense scientific reports into short video breakdowns in minutes. Worth keeping in your stack if biotech is a beat you cover regularly.
What This Means for You
Let me tell you what I would actually do with this information.
First, stop treating AI cancer headlines as investment signals. A press release is not a clinical result. A funding announcement proves nothing. The only numbers that matter are phase three trial results and FDA approval status. Everything else is storytelling.
Second, watch the companies that have drug approvals, not just drug candidates. Recursion, Exscientia, and Insilico Medicine all have drugs in or through clinical trials. That is a different category than most of the noise in this space. Candidates are cheap to make. Approvals are not.
Third, if you or someone close to you is dealing with cancer, the AI clinical trial matching tools are genuinely useful right now. Companies like TrialSpark and Antidote are using AI to connect patients with trials faster than the old system ever could. That application is working today, not in five years.
Fourth, watch the data ownership story. The companies that control large, clean, diverse patient datasets are building a moat that most competitors cannot cross. Owkin’s partnership with 25 major cancer centers is worth more than any single algorithm they run on top of it.
If you want to track biotech and AI research companies without paying enterprise prices, AppSumo regularly features research and productivity tools that independent investors and analysts use to stay ahead without the Bloomberg terminal budget.
The Bottom Line
AI will play a major role in cancer treatment. I believe that. But “a major role” is not the same as “a cure.” Anyone telling you we are five years away from defeating cancer with machine learning is either confused or selling something. The biology is hard. The trials take time. The data is messy. The best companies in this space know that. The ones raising money on hype do not, and they will not last long enough to matter. Watch who controls the data. That is where the real money is hiding.
Frequently Asked Questions
Is AI making real progress in cancer drug discovery?
Yes, but more slowly than the headlines suggest. According to the MIT Center for Biomedical Innovation, AI has improved the speed of early stage drug discovery by 30 to 50% in some cases. The bottleneck is still clinical trials, and no algorithm speeds those up.
Which AI cancer research companies are furthest along?
Recursion Pharmaceuticals, Insilico Medicine, and Owkin have drugs in or past phase two clinical trials. They work with real patient data and have partnerships with major medical institutions. They are past proof of concept, which puts them in a different tier than most competitors.
Can AI replace oncologists in diagnosing cancer?
Not yet, but it is getting close in specific areas. According to a 2025 study published in The Lancet Oncology, AI matched or outperformed radiologists in detecting early stage lung cancer in two out of three test conditions. Diagnosis and treatment are still different problems, though.
When will AI actually help cure a major cancer type?
Realistic estimates put AI assisted treatment protocols making a measurable difference in five year survival rates for some cancers within the next eight to twelve years. That is real progress. It is not the press release timeline, but it is not nothing either.
What is the biggest barrier to AI curing cancer?
Data quality and cancer’s biological complexity. Cancer is not one disease, and the datasets needed to train models across all 200 plus cancer types are fragmented across thousands of hospitals worldwide. The companies solving the data access problem are the ones worth watching closely.
