Nine in Ten Drug Trials Still Fail Despite the AI Hype, and That's Just One Crack in a Broken Health System

The AI drug discovery promise hasn't delivered where it matters

Pharma spent the last decade selling a story: AI would compress years of molecular hunting into months and end the brutal failure rates that make new medicine so expensive.

Investors bought it. Billions of dollars poured in. Thousands of startups launched.

But according to TechRound, the numbers that matter most haven't moved. Nine out of ten drug candidates entering Phase I clinical trials still fail before ever getting approved. Developing a drug still takes an average of 9.1 years. More than half of trial protocols need at least one amendment, and a single Phase III modification can cost anywhere from $140,000 to $535,000, per TechRound's reporting.

AI-discovered molecules do show Phase I success rates of 80 to 90 percent, roughly the industry average. But push those same candidates into Phase II and success rates sink back down to the standard 40 percent mark, TechRound reports. AI sharpened the early search. It hasn't solved the wall these drugs hit later.

Proponents can rightly say the technology was never claimed to fix clinical trial design, patient stratification, or the biological complexity that causes most late-stage failures. Those are execution problems, not discovery problems. TechRound's own analysis backs that distinction, noting the failures are "biological complexity failures, patient stratification failures, trial design failures" rather than computational ones.

The capital allocation gap is telling. TechRound points out there are almost no AI-native public companies focused on clinical trial execution, regulatory submission, or drug safety monitoring after approval. The money went to the flashy part of the pipeline, not the expensive, messy part that actually determines whether a drug reaches patients.

A commentary piece from Downunder Voices put the human cost more bluntly: "The damage is hard to assess: the trials that were never run, the breakthroughs that never happened. What we do know is that the search for new cures is broken."

Cuban: it's not just the science, it's the money

Mark Cuban has been making a related argument for years, and he laid it out again in a recent interview with Stanford Department of Medicine's "Future of Medicine" podcast, hosted by Dr. Euan Ashley.

Cuban didn't mince words. "The whole health care system is so f---ed up," he told Ashley.

His core complaint isn't the drugs themselves. It's opacity. "Trust equals transparency divided by self-interest," Cuban said, arguing that patients and doctors alike are kept in the dark about what things actually cost and why.

Cuban has also argued doctors should be paid more and, in some cases, work less, a position he raised in earlier conversations referenced in the interview. His broader pitch is a market solution, not a government one: force pricing into the open and let competition do what regulation hasn't.

Transparency and competition, not more federal mandates, are the tools Cuban says would actually move prices. Whether insurers, hospital systems, and drugmakers ever agree to that kind of exposure remains an open question none of the available reporting answers.

The Clancy trial and the gap in postpartum care

A third crack in the system showed up far from a lab or a boardroom: in a Massachusetts courtroom.

Lindsay Clancy, a former labor and delivery nurse, stood trial for killing her three children and attempting to take her own life in 2023. She does not deny the killings, according to PBS. Her defense argued she was suffering from postpartum psychosis at the time and that the care she received fell short. Prosecution experts testified to a different diagnosis and argued she is criminally responsible. Closing arguments took place in Plymouth, Massachusetts in late August 2026, per PBS.

Postpartum psychosis affects roughly one in 1,000 women, typically developing in the weeks after birth with symptoms including hallucinations, delusions, agitation, and confusion, PBS reports, citing Dr. Susan Hatters Friedman, a reproductive and forensic psychiatrist at Case Western Reserve University.

Maternal mental health advocate Meghan Cliffel, who experienced postpartum psychosis herself in 2015, told PBS the case triggered a wave of recognition among women who saw their own struggles reflected in it. "It feels infuriating to watch the hyper interest in this case without any proactive support for women to prevent this highly treatable, temporary illness," Cliffel said.

Dr. Hatters Friedman said the deeper problem is that providers often don't recognize the condition when it's in front of them. "I really feel for these women because it would be such a scary thing to be going through, and then there's not a system that necessarily recognizes what it is going on," she told PBS.

None of the reporting establishes whether Clancy's specific care fell below a professional standard, that's a contested factual question the jury, not this article, is positioned to answer. What isn't contested is the gap advocates describe: limited provider training on a rare but severe condition, and a system built around a six-week postpartum checkup that critics say misses warning signs entirely.

Three different failures, one pattern

A drug pipeline burning capital without matching results. A pricing system built on opacity instead of competition. A maternal mental health system that misses a rare but severe illness until it's too late.

None of these problems will be solved by the same fix. The common thread, according to the sources here, is a health system where the money and attention go to the visible, marketable parts of care while the harder, less glamorous operational work, whether it's clinical trial execution or postpartum screening protocols, gets left behind. Whether Massachusetts jurors, federal regulators, or pharma investors respond to any of this remains to be seen.