AI in drug discovery is starting to look like a teenager in a growth spurt: lots of hype, messy execution, and a tendency to burn through allowance faster than results appear. From $30 million in 2013 to $1.8 billion in 2021, funding for AI-driven drug discovery has skyrocketed. Yet in all that time, not a single AI-discovered drug has crossed the approval finish line.

TL;DR: Billions poured into AI drug discovery, but no approvals yet. About 31 drugs are in human trials, success rates in early phases look inflated, and industry consolidation is underway. The real promise for vitiligo lies in AI spotting biomarkers, stratifying patients, and repurposing drugs—not inventing miracle creams from scratch.

Why the Disconnect?

Biology doesn’t work on investor timelines. Designing a molecule on a screen is easy; getting it to hit the right target in a human body—without collateral damage—is a different beast. Daphne Koller of Insitro put it bluntly: "we’re trying to fix something we don’t understand, because of the complexity of human biology."

That gap explains why early poster children like BenevolentAI collapsed (shares down 99%), and why Exscientia — once a $2.9 billion darling — just sold to Recursion for $688 million. Think less Wall Street rocket ship, more yard-sale bargain.

Where Things Really Stand 

According to the Progress, Pitfalls, and Impact of AI‐Driven Clinical Trials paper, the scoreboard looks like this:

• 31 AI-discovered drugs are in human trials across eight leading companies.
• 17 in Phase I (one terminated)
• 5 in Phase I/II (one discontinued)
• 9 in Phase II/III (one already showing weak results)

Broader industry counts put the number closer to 67 molecules in trials.

And the approvals? Zero.

One repurposed generic has launched, but no genuinely novel AI drug has made it.

Business Models: Choose Your Poison

Most AI biotechs fall into three camps:

• Repurposers: Apply AI to old drugs and hope they work in new diseases. Faster to Phase II, but risky if the biology doesn’t line up.
• Best-in-class hunters: Use AI to design sharper molecules for proven targets. Less risky on the biology side, more brutal on the chemistry side.
• Moonshots: AI-designed, first-in-class molecules for novel targets. Brave, costly, and occasionally impressive — a Phase IIa trial in idiopathic pulmonary fibrosis showed safety and tolerability in just 18 months.

None of these are slam-dunks. Each trades risk in one area for exposure in another.

The New Twist: Big Money and Bigger Machines

A few developments in 2025 have reset the tone:

• Isomorphic Labs raised $600 million in Series A funding this March, led by Thrive Capital—the largest single round in AI drug discovery. Demis Hassabis is still betting AI can cut timelines “from years to months,” with deals from Lilly and Novartis reportedly worth up to $3 billion.
• Recursion’s BioHive-2 supercomputer now sits at #35 on the global TOP500 list. Pharma companies are suddenly building infrastructure once reserved for national labs and Silicon Valley giants.
• Phase I outcomes look rosy: AI-discovered drugs have shown 80–90% success rates compared to the traditional ~40%. But that may reflect smaller, handpicked trials, not true superiority. 

And then there’s consolidation. The Recursion-Exscientia merger signals that the industry is maturing. The days of “AI will cure everything” startups may give way to fewer, larger, better-capitalized players.

Novartis has quietly positioned itself as one of the most aggressive adopters of AI in pharma. In January 2024 it signed a multi-target deal with Isomorphic Labs (Alphabet/DeepMind), worth $37.5 million upfront and up to $1.2 billion in milestones, now expanded to cover six targets using next-generation AlphaFold models. Unlike many peers focused on faster screening, Novartis is betting on AI to crack “undruggable” targets and probe entirely new chemical space. Beyond Isomorphic, the company partners with Generate Biomedicines on generative AI for protein drugs and with Oxford’s Big Data Institute on statistical machine learning, blending internal capacity with “best of breed” external platforms. The result: Novartis sits alongside Lilly and Pfizer in the AI innovation race, but with a distinct focus on bold, high-risk biology.

The Roadblocks Nobody Likes to Talk About

• No benchmarks: Companies brag about speed but rarely publish timelines or costs.
• Data silos: Massive, publicly funded datasets remain disconnected.
• Culture clash: Drug hunters distrust algorithms; AI teams underestimate clinical slog.
• Funding inefficiency: Big checks flow into platforms before experimental proof. 

In other words: the hype machine runs ahead, while the science plods along.

What It Means for Vitiligo Drug Discovery

AI’s sweet spot is pattern recognition. That means the near-term potential in vitiligo is less about inventing pigment-restoring molecules from scratch, and more about:

• Identifying biomarkers that predict who responds to therapy.
• Stratifying patients into trial-ready subgroups.
• Spotting overlooked drugs or combinations worth repurposing.

The fantasy of a fully AI-designed cream that reboots melanocytes? Not yet. But AI may help us get smarter, faster, and leaner with the tools we already have.

The Bottom Line

AI in drug discovery is moving past the hype and into its awkward adolescence. Yes, progress is slower than promised. Yes, money is being burned. But glimmers of real efficiency are emerging. Supercomputers are running in pharma basements, Phase I success rates look suspiciously high, and consolidation is beginning to separate signal from noise.

For vitiligo, the message is simple: don’t wait for an algorithm to hand us a miracle cure. But do expect AI to quietly improve how we design trials, match patients, and repurpose compounds. The breakthroughs will come — but biology, as always, refuses to be rushed.

CEO VRF, Professor | Author A No-Nonsense Guide To Vitiligo

Dig deeper:

• 📌 Vitiligo Drug Pipeline Analysis and Market Insights
• Biologic Therapies for Vitiligo: A New Era of Hope
• Sixty Years of Vitiligo Research: Where We’ve Been and Where We’re Going

Or listen to podcast:

• Vitiligo Pipeline 2025: Winners, Watchlist, What’s Next (Ep. 41)

Our podcast Deep Dive In Vitiligo is available on all digital platforms, like Apple Podcasts, Spotify, Amazon, YouTube Music, Podcast Addict, iHeart and elsewhere.