Iambic vs Isomorphic Labs 2026: Which AI Drug Discovery Bet Wins?

Iambic vs Isomorphic Labs is the AI drug discovery matchup to watch in 2026. Iambic Therapeutics is heading for a US IPO at a valuation of up to $806 million, just after another AI listing met weak demand and left public investors nervous about AI funding. Isomorphic Labs, Alphabet’s AlphaFold spinout, is still private, well funded and partnered with big pharma. The public market will soon price one company every trading day, while Alphabet’s balance sheet shields the other. That gives us a real test of which AI drug discovery model is more likely to get drugs to patients and pay back its investors.

What’s new in Iambic vs Isomorphic Labs: Iambic files for an IPO

The San Diego company has filed to list in the US at a valuation of up to $806 million. That price is modest for an AI-first biotech, and the timing is poor. A recent AI-linked IPO drew weak demand, and investors are asking whether the AI funding boom has run ahead of revenue. Iambic pitches itself as a clinical-stage drug company that happens to use AI, not as an AI platform that might someday produce drugs. Public investors care about that difference.

Iambic’s core technology is NeuralPLexer, a generative model that predicts how proteins and small molecules bind. Instead of docking a molecule into a fixed structure, it models the full complex, including how the protein changes shape when a ligand binds. Iambic pairs it with property-prediction models and an automated chemistry and biology lab, so design-make-test cycles take weeks instead of months. Its lead asset, IAM1363, is a selective HER2 inhibitor for HER2-altered cancers now in early clinical testing. Other oncology programs follow.

Google DeepMind spun out Isomorphic Labs in 2021, and Demis Hassabis runs it. The company co-developed AlphaFold 3, which predicts structures and interactions across proteins, DNA, RNA, ligands and modifications. It has signed research deals with Eli Lilly and Novartis worth up to roughly $3 billion in combined milestones, and it raised $600 million in 2025, its first outside round. Isomorphic says it is preparing to take internal programs into human trials, but it has not disclosed its pipeline the way an S-1 forces Iambic to. That is the core asymmetry: Iambic has to show its cards, and Isomorphic does not.

Why it matters

• The IPO puts a price on “AI-native biotech.” If Iambic prices well and trades up, other AI drug discovery companies get a template for going public in 2026. If it breaks issue, expect more private rounds, mergers and quiet pivots to services revenue.
• Clinical data beats model benchmarks. NeuralPLexer vs AlphaFold 3 is an interesting technical debate, but no investor gets paid on a pose-prediction score. The asset that reads out clean Phase 1/2 data wins the narrative, whatever its architecture.
• Two business models are colliding. Iambic is a vertically integrated drug developer that owns its assets and its risk. Isomorphic runs a hybrid of Lilly and Novartis partnership milestones plus internal programs. Partnerships bring in cash sooner. Owned assets offer bigger upside.
• Value is moving from structure prediction to the full loop. Open and semi-open structure models such as the AlphaFold 3 code, Boltz and Chai are commoditizing the prediction layer. What stays defensible is proprietary data, automated wet labs, and how fast a team turns predictions into molecules that work.
• AI biotech stocks are still a thin category. Recursion is the main public comparable, and its stock has been volatile. A second credible listing gives generalist investors a benchmark, and a second company to sell if sentiment turns.
• Timelines are the real claim. Both companies say AI shortens the path from target to clinic. The only proof is how long it actually takes, which trial registrations and filings will show over the next 24 months.

How to use it today

You can’t buy into Isomorphic directly, and Iambic isn’t trading yet. You can still use the same class of tools both companies build on and track the investment case with public data.

1. Run AlphaFold 3 predictions for free on AlphaFold Server. Go to alphafoldserver.com, sign in with a Google account, and paste in protein sequences plus supported ligands, ions or nucleic acids. Non-commercial use is free, with a daily job quota. For a quick protein-ligand test, use a JSON job like this:

   [
     {
       "name": "her2_kinase_atp_test",
       "modelSeeds": [],
       "sequences": [
         {
           "proteinChain": {
             "sequence": "PASTEKLLE...PASTE_FULL_KINASE_DOMAIN_SEQUENCE",
             "count": 1
           }
         },
         {
           "ligand": {
             "ligand": "CCD_ATP",
             "count": 1
           }
         }
       ],
       "dialect": "alphafoldserver",
       "version": 1
     }
   ]

   Get the real sequence from UniProt (HER2 is P04626) and trim it to the kinase domain. AlphaFold Server only accepts ligands from a fixed list. Arbitrary small molecules need the local model.

2. Run AlphaFold 3 locally for arbitrary ligands (academic use). DeepMind released the inference code, but you must request the model weights separately under non-commercial terms. You need a recent NVIDIA GPU with plenty of VRAM.

   git clone https://github.com/google-deepmind/alphafold3.git
   cd alphafold3
   # Follow docs/installation.md to build the Docker image and fetch databases
   docker build -t alphafold3 -f docker/Dockerfile .
   
   docker run -it \
     --volume $HOME/af_input:/root/af_input \
     --volume $HOME/af_output:/root/af_output \
     --volume $HOME/af_weights:/root/models \
     --volume $HOME/public_databases:/root/public_databases \
     --gpus all \
     alphafold3 \
     python run_alphafold.py \
       --json_path=/root/af_input/fold_input.json \
       --model_dir=/root/models \
       --output_dir=/root/af_output

   The local input format accepts custom ligands as SMILES strings. That is the feature that matters for drug design.

3. Read the NeuralPLexer paper and try the open alternatives. Iambic described NeuralPLexer in Nature Machine Intelligence (2024) and published an earlier research codebase. For hands-on protein-ligand co-folding under permissive licenses, Boltz is the easiest to get running. Install it:

   pip install boltz -U

   Save this as input.yaml:

   version: 1
   sequences:
     - protein:
         id: A
         sequence: PASTE_PROTEIN_SEQUENCE
     - ligand:
         id: B
         smiles: 'CC(=O)Oc1ccccc1C(=O)O'

   Then run the prediction:

   boltz predict input.yaml --use_msa_server

   Run the same complex through AlphaFold 3 and Boltz and compare the poses. You will quickly see why every serious AI drug discovery company relies on wet-lab validation instead of trusting a single model.

4. Track the Iambic Therapeutics IPO filing directly. Read the S-1 on SEC EDGAR, not the headlines. Pull the filing index with the EDGAR full-text search API:

   curl -s -H "User-Agent: YourName your@email.com" \
     "https://efts.sec.gov/LATEST/search-index?q=%22Iambic%20Therapeutics%22&forms=S-1,S-1/A"

   Focus on the risk factors, cash runway, pipeline table and any collaboration revenue. Those four sections tell you more than any pitch deck.

5. Monitor clinical progress for both companies. ClinicalTrials.gov has a free API. Check it monthly for new registrations:

   curl -s "https://clinicaltrials.gov/api/v2/studies?query.spons=Iambic&fields=NCTId,BriefTitle,Phase,OverallStatus" | python -m json.tool
   
   curl -s "https://clinicaltrials.gov/api/v2/studies?query.spons=Isomorphic&fields=NCTId,BriefTitle,Phase,OverallStatus" | python -m json.tool

   Isomorphic’s first registered trial will be a major milestone. Until it appears, its internal pipeline is a promise.

6. Use an LLM to stress-test the investment thesis. Paste the S-1 risk factors into any capable model with a prompt like this:

   You are a skeptical biotech analyst. Using only the text below,
   list the 5 risks most likely to cause a >50% drawdown in the
   first 18 months after IPO. For each, cite the exact sentence,
   estimate probability (low/med/high), and name the observable
   signal (trial readout, cash runway, partner exit) that would
   confirm it. Do not speculate beyond the text.
   
   [PASTE RISK FACTORS HERE]

How it compares

Dimension	Iambic Therapeutics	Isomorphic Labs
Core model	NeuralPLexer (generative protein-ligand complex prediction) plus property models	AlphaFold 3 lineage plus proprietary design models
Backing	Venture-backed; US IPO at up to $806M valuation	Alphabet, plus a $600M external round in 2025
Business model	Owns and develops its own assets	Pharma partnerships plus internal programs
Major partners	Mostly internal; limited big-pharma dependence	Eli Lilly, Novartis (up to ~$3B in combined milestones)
Clinical status	Lead HER2 inhibitor IAM1363 in clinical testing	Preparing first internal programs for the clinic
Transparency	High once public (S-1, quarterly reports)	Low; private with selective disclosure
Wet-lab integration	Automated in-house design-make-test lab	Growing internal lab capacity plus partner labs
Can you invest?	Yes, after listing	Only indirectly through Alphabet
Biggest risk	Single-asset clinical readouts and post-IPO sentiment	Long time to clinical proof; value mixed into Alphabet

Our take: on technology alone, Isomorphic has the stronger brand and more compute and talent behind it. AlphaFold 3 changed the field. Iambic, though, is closer to what pays off: human data on a molecule it owns. If you want the bet that produces a clear, checkable answer sooner, pick Iambic. If you want the bet more likely to still be around in ten years, Isomorphic has the deeper pockets and the pharma partners.

What’s next

For Iambic, the next 12 months come down to two things: how the IPO trades and the first meaningful IAM1363 data. Biotech IPOs usually face their first real test at lock-up expiry and the first post-listing readout. A clean safety profile with early efficacy signals in HER2-altered tumors would support the “AI makes better molecules faster” story. A messy readout would hit the stock and the whole AI biotech category. Also watch for a post-listing pharma partnership, which would bring non-dilutive cash and outside validation of the platform.

For Isomorphic, the milestone is its first internal program entering human trials. Every month without one makes the company look more like a well-funded research lab and less like a drug company. Watch for Lilly and Novartis milestone payments, any expansion of those deals, and whether Isomorphic publishes models beyond AlphaFold 3 that push into binding affinity, ADMET or generative chemistry. Alphabet could also eventually spin it out or take it public, which would give retail investors a second direct way into AI drug discovery.

For AI drug discovery companies broadly, the open question in 2026 is whether public markets will pay for platforms or only for assets. The Iambic listing is the first real-money answer this year. Watch the first-day pop, the 90-day performance and the follow-on financings that come after it.

Frequently Asked Questions

What is the main difference between Iambic and Isomorphic Labs?

Iambic is a vertically integrated, clinical-stage biotech that uses its NeuralPLexer platform to design drugs it owns and develops. Isomorphic Labs is an Alphabet-owned company built on AlphaFold 3 technology. It combines large pharma partnerships with internal programs that have not yet reached the clinic. Iambic is further along clinically. Isomorphic has more resources.

How does NeuralPLexer compare to AlphaFold 3?

Both predict the 3D structure of protein-ligand complexes. AlphaFold 3 covers a broader range of biomolecules, including DNA, RNA and modifications, and is widely treated as the reference model. NeuralPLexer focuses on how proteins change shape when a small molecule binds, which matters for drug design. Neither model alone determines success. The data, lab automation and chemistry team around the model matter more.

Can I invest in Isomorphic Labs?

Not directly. Isomorphic Labs is private and majority-owned by Alphabet, so Alphabet stock is the only public exposure, and Isomorphic is a tiny fraction of Alphabet’s value. Iambic will be directly investable once its IPO is complete.

Is the Iambic Therapeutics IPO a good investment?

That depends on your risk tolerance. This is not financial advice. Early-stage biotech IPOs swing hard on clinical readouts, and Iambic’s value depends heavily on its lead HER2 program. Before you put money in, read the S-1 risk factors and cash runway and know when the next data readout is expected.

Which AI biotech stocks exist besides Iambic?

Recursion Pharmaceuticals is the most prominent publicly traded AI-first drug discovery company. Schrödinger, which sells physics-based computational chemistry software and develops its own drugs, is often grouped with it. Several smaller names have listed or merged into public shells. Liquidity and analyst coverage across the category are still thin.

Has any AI-designed drug been approved yet?

AI-discovered molecules have reached mid- and late-stage trials, but none from this new generation of AI-native companies has won a clear regulatory approval that settles the question. The first approval of a drug designed largely by AI will be a defining moment for the sector, and Iambic and Isomorphic each want to be the company that gets there.