# Who Actually Makes Ozempic? The GLP-1 Supply Chain, Mapped

> We mapped the GLP-1 supply chain behind Ozempic and Zepbound, from amino acid to injection pen — and found the molecule was never the real bottleneck.

- Author: Barebone Research, Barebone AI
- Published: 2026-05-05
- Canonical: https://barebone.ai/resources/who-makes-ozempic-glp1-supply-chain-mapped
- Publisher: Barebone AI (https://barebone.ai)

---

## Two Factories, One Drug Class

Inside a Novo Nordisk plant in Denmark, the world's best-selling weight-loss drug is brewed a little like beer. Vats of genetically engineered yeast ferment, the cells secrete the semaglutide backbone — the molecule inside Ozempic and Wegovy — and a fatty-acid chain is bolted on afterward by chemistry.

A few thousand miles away, in Eli Lilly's reactors, the rival drug is built the opposite way. Tirzepatide — the active ingredient in Mounjaro and Zepbound — is assembled one amino acid at a time on a solid resin bead, a method called solid-phase peptide synthesis. The molecule runs 39 amino acids long, two of them synthetic building blocks that don't exist in nature. Those two non-natural links are exactly why Lilly can't brew it in yeast the way Novo brews semaglutide.

Same drug class. Opposite factories. And that split is the first clue to a story most investors get backwards.

The popular version goes like this: GLP-1 drugs are easy to make, so all the money flows to whoever owns the patents — Lilly and Novo — and everyone else is a rounding error. The patents part is right. The "easy to make" part is where it falls apart.

We used Barebone AI to trace the GLP-1 supply chain end to end — from the amino acids that start the synthesis to the injection pen in a patient's hand — across earnings calls, capacity announcements, and the filings of the contract manufacturers nobody talks about. What we found is a chain where the molecule was never the bottleneck. **Building a billion doses a year was.** In 2025, four GLP-1 products pushed roughly **$67 billion** through it.

## The Two Questions That Decide Who Wins

To see why "easy to make" is the wrong read, start with the two questions that decide who captures the money in any drug: how strong are the patents, and how hard is it to manufacture at scale?

Those two questions make a grid.

| | Easy to make | Hard to make at scale |
|---|---|---|
| **Strong patent** | Drugmaker takes almost everything | Drugmaker shares with specialist manufacturers |
| **Weak patent** | Distributor wins | Whoever can manufacture wins |

The top-left is the dream: a strong patent on something simple to produce. Pfizer lived there with Viagra, which peaked above **$2 billion a year** — until the patent expired, the molecule (sildenafil) turned out to be trivial to make, and the box dropped into the bottom-left, where telehealth sellers like Hims and Ro now move it for pennies. Same pill, completely different winner.

COVID vaccines sat top-right: patent-protected and genuinely hard to manufacture, which is why Pfizer and Moderna had to share the spoils with specialist fill-finish firms like Catalent — the company that filled the vials.

The script most investors are running puts GLP-1 firmly in the top-left: strong patents, easy to make, drugmakers take it all. The patents are real, and the drugmakers do capture most of the value. But GLP-1 is sliding rightward, toward *hard to make at scale* — and that drift is the entire reason a supply chain exists to map.

## The Map at a Glance

Here is the chain, from rock to injection.

| Layer | What it is | Who controls it |
|---|---|---|
| **1. Building blocks** | Protected amino acids, resins, coupling reagents, solvents | Many suppliers; CDMOs like Bachem make their own |
| **2. The API** | Turning a sequence into a GMP-grade peptide | Drugmakers in-house, plus a handful of CDMOs |
| **3. Contract manufacturers** | Outsourced API and intermediate capacity at scale | Bachem, PolyPeptide, CordenPharma, WuXi, AmbioPharm |
| **4. Fill-finish** | Sterile filling into cartridges and vials | Catalent (now Novo's), Vetter, Thermo Fisher, a few others |
| **5. Devices** | Pens and autoinjectors | SHL Medical, Ypsomed, BD |
| **6. Drugmaker + distribution** | The patent owners, then pharmacies and telehealth | Lilly and Novo; then LillyDirect, Hims, retail |

<Chart name="PeptideChokepointChart" />

Read the chart against the map. The layers that matter — API at scale, fill-finish, the pen, the branded drug — are each held by single digits of companies. The chain doesn't narrow smoothly toward the top; it starts narrow and collapses to two names at the end. Walk it from the molecule out.

## Layers 1–2: The Molecule Was Solved Years Ago

The chemistry of these drugs is not the constraint. Semaglutide and tirzepatide were designed, patented, and proven in trials years before they were household names. What took the years — and the billions — was learning to make them by the ton without defects.

That is harder than it sounds. Novo's recombinant route ferments semaglutide in yeast, then purifies it through chromatography and attaches the fatty-acid chain. Lilly's chemical route stacks tirzepatide's 39 residues one coupling reaction at a time, each step needing protected amino acids, coupling reagents, and rivers of solvent. Both routes are well understood. Neither is easy to run at metric-ton scale with regulators inspecting every batch — which is why, by industry estimates, only about five companies can synthesize GLP-1 peptides at commercial volume at all.

That is the real meaning of "the molecule was the easy part." The sequence is public. The factory that can make 855 metric tons of it — roughly where industry forecasts put annual peptide demand by 2031, up from about 265 tons in 2025 — is not.

## Layer 3: The Picks-and-Shovels Nobody Names

When demand for a hard-to-make ingredient explodes, the contract manufacturers — CDMOs, for contract development and manufacturing organizations — are the ones who scale it. Two of them trade publicly as pure plays.

**Bachem**, the Swiss peptide specialist, grew 2025 sales to **CHF 695 million**, up roughly 15%, at a **30.9% EBITDA margin** — fat margins for a manufacturer, the signature of a layer where customers have nowhere else to go. **PolyPeptide**, its smaller rival, lifted 2025 revenue to about **EUR 389 million** and dragged its EBITDA margin from 7.5% to the low teens as its large-scale GLP-1 lines finally ramped. Both are pouring money into capacity: across the top five peptide CDMOs, committed investment since 2022 runs past **$3 billion**.

These are the names that benefit no matter which drug wins, because Mounjaro and Ozempic both need the same reactors. The peptide CDMO market has compounded at roughly 20% a year — a very good place to stand, with one catch we will get to.

## Layers 4–5: Where the Real Bottleneck Lives

Here is the part the "easy to make" crowd never reaches. A purified peptide is not a product. It has to be filled, under sterile conditions, into a cartridge — and then that cartridge has to go into a device a patient can press against their own stomach.

Fill-finish is the chokepoint that actually bound. Sterile filling lines are scarce, capital-intensive, and slow to qualify; at the peak of the shortage, lead times stretched past a year. That single fact explains the most aggressive move in the story: in December 2024, Novo's controlling shareholder bought the contract manufacturer Catalent for **$16.5 billion**, mainly to seize three fill-finish sites and point them at Wegovy. You do not pay that for capacity that is easy to come by.

Then the pen. The least-discussed layer on this map is arguably the tightest: the autoinjector and pen market is dominated by **SHL Medical**, with around 60% share, and **Ypsomed**, with roughly 20%, in a market worth on the order of €1.3 billion. A weight-loss drug without a device is a vial nobody can use. Two companies largely decide how fast the world's injectable GLP-1s reach arms.

## Layer 6: Where the Money Actually Lands

Now the top of the funnel — and the script gets this part right. Value concentrates with the patent owners.

<Chart name="PeptideDemandChart" />

In 2025, Lilly's Mounjaro and Zepbound booked a combined **$36.5 billion**, and Novo's Ozempic and Wegovy roughly **$30 billion** more. Lilly's Q1 2026 revenue hit **$19.8 billion**, up 56% year over year, and management raised full-year guidance to **$82–85 billion**. The market has paid for it: Lilly carries a market value near **$1 trillion** as of the May 5 close.

But notice the asymmetry inside the duopoly. Novo's market cap sits around **$200 billion** — less than a quarter of Lilly's — and its stock is down double digits in 2026, hit by compounded knockoffs and a slower US obesity market. Owning the patent is necessary. It is not sufficient. Even at the richest layer of the chain, one of the two kings is being out-competed.

## The Bottleneck Was Never the Molecule

Step back and the map reads cleanly. The science was finished years ago; the binding constraints were physical — reactors that run peptides at the ton, sterile lines that take a year to stand up, two device makers, two drugmakers. The scarcity premium didn't sit only with the patent owners. It got distributed *down* the chain and turned into the largest capacity build-out in the industry's history.

That is the supply chain the "easy to make" thesis can't see. It is right that Lilly and Novo capture the value. It misses that doing so forced a multibillion-dollar scramble through six layers of suppliers — and created a set of pricing-power chokepoints at the API, fill-finish, and device layers that didn't matter when the drugs were small.

## Where the Map Breaks

An honest map marks its own hazards. This one has four.

<Chart name="PeptideCapexChart" />

**The customer is becoming the competitor.** Lilly already outsourced part of tirzepatide synthesis to CDMOs like WuXi — and is now insourcing hard. Its February 2025 plan commits **$27 billion** to four new US plants, three of them for active ingredients, lifting Lilly's US manufacturing spend past **$50 billion** since 2020. Novo makes its API in-house and bought its fill-finish outright. Put the drugmakers' build-out next to the CDMOs' and the chart above is the bear case in one image: the suppliers' biggest potential customers are building their own capacity. That is why the pure-play CDMO stocks have been volatile despite booming demand.

**The next molecule routes around the entire chain.** On April 1, 2026, the FDA approved Lilly's orforglipron — branded Foundayo — the first oral GLP-1 for weight loss with no food or water restrictions. It is a *small molecule*, not a peptide: made by ordinary chemical synthesis, like any pill — no yeast, no solid-phase reactors, no peptide CDMO, no injection pen. If oral drugs take meaningful share, layers one through five get designed around, exactly the way they were built up.

**Capex widens chokepoints.** Every dollar in the chart above is funding tomorrow's supply. Today's scarcity premium at the API and fill-finish layers is precisely what is paying to erase it. Chokepoint maps have a shelf life of quarters.

**Don't confuse the two peptide stories.** In April 2026, the FDA pulled 12 research peptides — BPC-157, semax, and the rest of the biohacker shelf — off its Category 2 "do not compound" list, with advisory meetings starting in July. That is a story about compounding-pharmacy access, not the branded GLP-1 supply chain. The molecules, the makers, and the rules are different. They share a word, not a market.

## What This Means

A supply-chain map is not a buy list. It exists to show where scarcity actually binds — because pricing power lives where the supplier count is smallest, and disappears where capital floods in.

Four signals will tell you whether this map still holds through 2026:

1. **Oral share.** How fast orforglipron and other small-molecule pills take volume from injectables is the single biggest threat to layers one through five.
2. **Insourcing pace.** As Lilly's and Novo's new plants come online, watch whether CDMO order books soften — the tell that the customer-as-competitor risk is real.
3. **Fill-finish lead times.** Still measured in quarters? The bottleneck holds. Back to weeks? Capital has caught up.
4. **Capex prints.** Each new plant announcement is the chain telling you where the makers think the next constraint is.

The molecule was the easy part. The chain that turns it into a billion injections a year is the hard part — and for now, most of it still runs through a handful of hands per layer. The question is not whether the drugs sell. It is whether the chokepoints survive the capital being thrown at them.

---

*Data: Barebone | Sources: Eli Lilly Q4 2025 and Q1 2026 results (SEC filings, Feb 4 and Apr 30, 2026), Novo Nordisk FY2025 report, Bachem and PolyPeptide FY2025 results, Novo Holdings–Catalent acquisition (Dec 2024), FDA orforglipron approval (Apr 1, 2026), company capacity announcements | Prices as of May 5, 2026 close*