$290 million. That's what a group of investors just handed a company that has never put a drug into a single human being.

The company is Anew Labs, and until this summer it wasn't a company at all. It was a research team living inside ByteDance — the same ByteDance that owns TikTok and keeps your thumb moving. On September 16, Reuters reported that the newly spun-out unit had closed its first outside funding round at a $1.5 billion valuation, with ByteDance holding onto a 56% controlling stake (The Next Web / Reuters).

So a social-media giant now co-owns a billion-dollar drug shop with exactly zero medicines in the clinic. The pitch isn't a molecule. It's that the machine-learning muscle built to rank your For You page can be pointed at proteins instead.

🧠 Why This Matters

Anew didn't appear from nowhere. Its people spent years inside ByteDance as an "AI-for-Science" team, running since 2021 before the formal spin-out in June 2026 (pharmaphorum). Along the way they shipped something real: Protenix, an open-source model that predicts the 3D structure of biomolecules. It's an open reproduction of DeepMind's AlphaFold3, published on bioRxiv in early 2025, and the v1 release in February 2026 claimed AF3-level accuracy on standard benchmarks.

"Advancing Structure Prediction Through a Comprehensive AlphaFold3 Reproduction." — the Protenix team, bioRxiv (Jan 2025)

That matters because structure prediction is the expensive, slow part of early drug design. If you can guess how a protein folds and how a candidate molecule latches onto it — in software, in minutes — you skip a lot of wet-lab grind. The reason ByteDance could build this at all is the same reason it can serve a billion video feeds: compute and talent, at scale. Its cloud arm, Volcano Engine, still supplies Anew's computing power.

📊 Deep Dive

Here's the anatomy of the deal, by the numbers:

  • Raised: $290 million — the company's first external financing round.
  • Valuation: $1.5 billion post-money.
  • ByteDance stake: 56%, a controlling position, even after selling shares to outsiders.
  • Lead investors: HSG (formerly Sequoia China), IDG Capital and Hillhouse, with 5Y Capital co-leading; Gaorong, Primavera, Boyu, the state-backed Shanghai Future Industries Fund and SBP Group also joined (Finimize).
  • People: roughly 50 core staff moved over in the spin-out, from a compact team of around three dozen.
  • Footprint: offices in Shanghai, Singapore and San Jose, California.
  • Pipeline: about four programs, led by an IL-17 small-molecule inhibitor aimed at autoimmune conditions like psoriasis and ankylosing spondylitis, plus IL-4 receptor work — all still in lab optimization, none in human trials (Crypto Briefing).

As a standalone business, Anew now markets a platform stack — protein structure prediction, antibody design, and an LLM-style reasoning assistant for chemists — built on the same lineage as Protenix. It's led by Liu Kai, who came out of the venture world at IDG Capital.

ByteDance released Protenix-v1 as "a new open-source model achieving AF3-level performance in biomolecular structure prediction." — MarkTechPost, Feb 2026

⚠️ The Catch

Now the cold water. A $1.5 billion valuation on a company with nothing in clinical trials is a valuation on a promise. Predicting a protein's shape is not the same as proving a pill is safe and works — and the graveyard of drug development is stacked with molecules that looked perfect on a screen and failed in a body.

The AI-drug field as a whole is still waiting for its first blockbuster designed start-to-finish by these tools. AlphaFold won a Nobel Prize and reshaped biology, yet no approved medicine yet carries "designed by AI" as its origin story. Anew's lead program is in lead optimization — years, not months, from an approval decision, and that's if the trials go well.

There's also the ownership wrinkle: with 56% still in ByteDance's hands, the "independent" startup answers, ultimately, to a company whose main business is short video. And ByteDance itself sits in a complicated spot — recent secondary sales have valued it anywhere from about $480 billion in a share auction to roughly $550 billion in later stake sales (Music Business Worldwide; PE Insights) — a spread that tells you how hard the whole group is to price.

🎯 What Happens Next

Watch three things. First, the clinic: Anew needs to move that IL-17 program into an actual trial. Filing to test a drug in humans is the milestone that turns a software story into a drug story. Second, the money trail — whether ByteDance keeps writing checks or lets outside investors carry more of the burn. Third, distance. Spin-outs are often the first step toward a company that can raise, partner, or list on its own without a parent's shadow.

Expect partnership announcements with established pharma before you see trial data. That's the usual path: license the platform, bank the milestones, and let a big drugmaker shoulder the trial costs.

🧩 Bigger Picture

This is a pattern now, not a one-off. Alphabet has its own drug-discovery outfit in Isomorphic Labs, spun from DeepMind. A wave of AI-first biotechs has raised billions on the same wager — that computation shortens the decade-long, billion-dollar slog of making a medicine. ByteDance is the latest tech heavyweight to decide that the muscles it built for consumer apps might also crack open biology.

The honest read: the tooling is genuinely good and genuinely useful, and it is also unproven where it counts, at the bedside. A $290 million round buys years of runway to find out which. For now, Anew has the compute, the models, the money — and a burden of proof measured in clinical readouts, not benchmarks.

The algorithm learned to fold proteins. Whether it can fold one into a cure is a much longer download.

Sources