Obesity medication has something of a troubled past. Fen-phen, a weight-loss drug combination popular in the 1990s, was pulled after it was found to cause heart valve problems. Sibutramine, sold under the brand name Meridia, was prescribed until it was discovered to lead to adverse cardiovascular events including strokes in 2010.  

But the market for an effective weight-loss drug is too big and the potential profits too high for pharmaceutical companies to give up. More than one in eight people around the world live with obesity. In the United States, it’s more than two in five. Though many clinical trials of weight-loss drugs over the past decade ended in failure, it was only a matter of time until a successful drug emerged. 

GLP-1 medications ¹ like Ozempic appear to be that drug. Estimates suggest GLP-1s can reduce body weight by at least 15% when taken regularly — and perhaps even more as newer drugs come to market. And though evidence is still being gathered, they may have benefits beyond weight loss: potentially curbing drinking, treating sleep apnea, and reducing risk of stroke. They’ve been called, in many places, a miracle drug, and as such, the category is poised for massive growth. Gallup estimated that 15.5 million Americans have tried them, and half as many are currently using them.

But to date, the supply has been plagued by shortages so severe that there’s a website that tracks which pharmacies have it in stock. Though new products continue to come to market, and companies are doing what they can to increase supply, it’s extremely unlikely that demand will be met in the near term. So what does growth for GLP-1s really look like, and how many people stand to benefit? 

In this forecast, I’m going to look at this over the next six years.

Question: How many GLP-1 agonist medications will be sold in 2030 in the United States? Since a substantial portion of people discontinue their use, we’ll measure this by the number of one-year supplies sold.

Our projections for the future will be based on growth relative to the present, so it’s important to have a good idea of what GLP-1 sales look like today, which we’ll approach in a few different ways.

Polling is one approach. A Gallup poll from March estimated that 3% of American adults (approximately 7.75 million people) are currently taking GLP-1s. The poll has some issues for our purposes, ² but its flaws bias the results in opposing directions, so approximately eight million is a fine starting point. ³

Another approach is to take company revenue for GLP-1s and divide that by the estimated net price to get monthly supply. Almost the entire supply of GLP-1s in the United States is manufactured by either Eli Lilly or Novo Nordisk. Looking at their revenue for 2023 implies enough supply for 5.4 million patients each month: 3.5 million for Novo Nordisk and 1.9 million for Eli Lilly. 

But sales are growing fast, and we want to base our estimate as much as we can on 2024 data. Novo Nordisk expects approximately 23% sales growth for 2024, or $7.8 billion at current exchange rates. Lilly expects revenue growth of about $8.9 billion. If those projections are right, and we assume 100% of Novo’s growth and 90% of Lilly’s ⁴ is from GLP-1s, that gives us growth of $15.8 billion (54%) in the GLP-1 market, for a total of $45 billion. 

The United States represented 72% of GLP-1 sales globally in 2023 — if that stays roughly constant, we should expect $32 billion in US sales for 2024. If, for simplicity, growth is also evenly spread between brands, then implied supply for 2024 would be 5.4 million * 154% = 8.3 million — similar to the poll results. 

So enough supply for eight million people on average in 2024 seems like a safe base to work from.

I’ll begin by taking an “outside view” approach — that is, ignoring the details of specific GLP-1 drugs like Ozempic and coming up with an estimate based on other drugs in roughly the same reference class. This approach is less accurate than a more detailed “inside view” model — which we’ll get to— but it also requires much less information. Taking the outside view allows me to come up with a rough estimate that I can adjust as I do additional research. It also helps me account for factors I might not explicitly think to include when building a detailed model of GLP-1s in particular. From there, I can tweak only the factors that I think are significant and different for Ozempic, rather than trying to establish each and every one from scratch. 

So what’s the right reference class for GLP-1 medications? I chose to look at the uptake of other recent multibillion-dollar “blockbuster” pharmaceutical drug categories. These are new classes of drugs created in the past 40 years or so that, like GLP-1s, experienced rapid expansions in sales and production. The drug categories I used are:

There are hundreds of other types of drugs I could’ve used. I want to address some drugs I didn’t use, and why.

The first omitted drug is insulin. Insulin is similar to obesity drugs in many ways. It addresses a large patient population, it’s a biologic (or drug made from a living organism), it’s an injection used on an ongoing basis, and it treats similar indications — Ozempic was initially approved as a treatment for diabetes. But market conditions now are very different from those of 1923, when commercial insulin was first introduced. Furthermore, when insulin was discovered and commercialized, it had to be extracted from pig and cow pancreases. Scaling production meant building an operation to collect animal pancreases and developing processes to better refine the pancreatic extract — a much different process than building a modern factory. 

I also left out Hepatitis C drugs like sofosbuvir, which came out in the 2010s. Though this is another blockbuster release that had several billion dollars in sales, it is a cure, rather than an ongoing treatment, and so its sales trend downward rather than up. 

That leaves the four drug classes listed above, which have their own limitations. First, obesity drugs have a larger potential market than any of them. Almost 150 million American adults have diabetes or obesity. About 45 million are eligible for statins, ⁵ the next largest population, and the markets for DOACs and TNF-α inhibitors are likely in the single-digit millions. The US government also has more incentive to make GLP-1 drugs available — both because promoting weight loss would lead to savings for Medicare down the line and because it would (probably) please voters. Both factors suggest faster growth.

On the other hand, GLP-1s currently receive different insurance treatment in the United States when prescribed for weight loss, and they currently have lower patient adherence rates when compared with comparable drugs. Those factors point in the other direction.

Finally, and most importantly, obesity drugs have already shown a different growth path than our base rate drugs. The other drugs gained popularity very quickly after launch. But GLP-1s were initially prescribed as a treatment for diabetes and then gained popularity for weight loss later on thanks to off-label use. Novo Nordisk then underestimated demand for Wegovy and ran into supply problems, which meant sales didn’t truly take off until late 2022 or 2023. 

There are two different benchmarks we can get from the reference class: sales and prescriptions. Each approach has its benefits and drawbacks, but both should hopefully give similar estimates. If they do, then we’ll be more confident that our base rate forecast is robust.

For each category, I collected annual net sales data for each drug from published research and drugmakers’ annual reports. ⁶ I then calculated year-over-year changes in sales relative to the drug class’s first full year of significant sales. The growth rates over time for the drugs selected follow  similar exponential decay patterns. Growth rates start high when sales are small. PD-1/PD-L1 sales grew by about 75% in their second year, and DOACs more than doubled. But by their eighth year on the market, sales for all classes grew between only 9% and 18%.

Looking at cumulative growth relative to the fourth year of sales, we can see very roughly the type of growth we should expect for GLP-1s over the next few years. If 2024 is year four, we expect somewhere between 180% and 240% growth in total sales by 2030.

To implement the sales model, I built a Monte Carlo model and ran it 10,000 times. The goal of a Monte Carlo simulation is to model future sales while accounting for uncertainty and variability in the possible growth trajectory of GLP-1s. This works by setting some parameters ourselves and randomly sampling others for each run. The cumulative result of this process should tell us which outcomes are most likely. To get a forecast for GLP-1s using that information, we’ll need to first answer one question: What year of sales are we in for GLP-1s?

This matters because, while sales grow each year, the rate at which they grow slows down. Picking the wrong year would give us the wrong rate of growth, and ultimately very different results. We could take the simple approach and define year one as the first full year in which a modern GLP-1 drug, Ozempic, was widely sold (2019), ⁷ making 2024 what we call “index year” six. This isn’t a bad approach. In 2019, Ozempic was approved only as a treatment for diabetes, but that’s still a patient population of 38 million Americans. But GLP-1s have exploded into the larger obesity market much more recently. Indeed, if we look at growth in revenue for Novo Nordisk and Eli Lilly from 2023 to 2024 and compare it with our growth rate model, it suggests that 2024 should really be seen as approximately index year 3.5. I handled this discrepancy by making it one of the parameters I randomly varied between runs of my model.

After converting sales numbers to patient supply, this produces the following distribution:

Our median estimate at this point suggests a patient-year supply of around 18 million.

That’s what looking at sales gets us, but we can also look at prescription volume directly. Data on prescription fills is not widely publicly available, especially going back multiple decades. However, there is some limited data from the long-running Medical Expenditure Panel Survey. The survey measures 18 respondents reporting a TNF-α prescription in year four and 38 in year 10. In the same period, DOAC respondents went from 132 to 527. Data doesn’t go back far enough to include the early years of statins.

Still, this is a data point worth considering. Our outside view using sales predicts approximately 100% growth in US supply. Prescription growth is higher than those estimates, so we may want to adjust them upward. 

The outside view approaches above are a useful starting point. Now we’ll consider information specific to GLP-1s. 

We should start with supply constraints. After that, there are three other considerations that, in my opinion, are both significant to and different from the reference class: insurance treatment, patient adherence, and pipeline developments. 

Supply constraints and manufacturing expansion

Demand for GLP-1s is sky-high. A major limiting factor for sales right now, according to both Eli Lilly and Novo Nordisk, is supply. So why is supply so hard to scale up? 

Right now, the constraint appears to be in production facilities. Entirely new facilities need to be bought and repurposed or even built to meet demand. At baseline, much of pharmaceutical manufacturing is contracted out to third-party companies. Normally, this system would provide some flexibility for Novo and Lilly to expand production, but current capacity is too small to meet the historic demand for GLP-1s. Novo Nordisk recently agreed to acquire three factories by purchasing Catalent, a contract manufacturer which also produces products for Eli Lilly, in order to maximize production of its own products. Additionally, injector pens for Mounjaro are apparently unusually complex and may not be able to use existing assembly lines from other factories, which further limits Lilly’s options. 

There are also geopolitical concerns. The United States is actively cracking down on Chinese contract manufacturers, including WuXi AppTec, who reportedly produces much of the active ingredient in Mounjaro. All of this suggests that Novo and Lilly will need to invest in building new facilities or expanding existing ones, rather than utilizing existing sites. This isn’t like building a new Costco. Such advanced manufacturing facilities are multiyear, multibillion-dollar projects. 

What’s especially notable about the current situation is that it’s very distinct from the reference class drugs. We didn’t see large spikes in capital expenditures after the launch of drug classes in our base rate. However, we see it very clearly already for Novo and Lilly. Lilly recently committed at least $13 billion for factories in North Carolina, Wisconsin, Indiana, and Germany, which are expected to come online at varying points by 2028. That’s a nearly 70% increase relative to their gross plant, property, and equipment (PP&E) holdings in 2021 of about $19 billion. Novo Nordisk has similarly announced at least $19 billion in investments. $11 billion of that is for Catalent’s factories, with the rest going to facilities in Kalundborg, Denmark, and Chartres, France. This would more than double their gross PP&E holdings of about $14 billion in 2021.

How much should we update based on this information? 

Not as much it may seem we should at first glance. First, some investments would have happened regardless. Lilly averaged approximately $1.2 billion annually in PP&E purchases from 2015 to 2021, and Novo Nordisk averaged approximately $1.1 billion. The large sums announced recently likely include some investments they would have made anyway. 

It’s also already partly priced in. Our model predicts doubling sales, and while it doesn’t explain how, that capacity has to come from somewhere. We already expected that sales would be much larger than TNF-α inhibitors and DOACs, so unusually large investments are reasonable.

There are a lot of unknowns here that make it very difficult to rigorously factor this information in. It’s clearly important, but I don’t know, for example, to what degree the base rate factors this type of growth in, and I don’t know how much these investments actually increase production, whether they’ll be completed on time or in budget, what level of growth would be possible without them, or what additional investments are yet to come. 

In the end, I tried to estimate the GLP-1-relevant manufacturing assets Novo and Lilly have now and then estimated how much of an increase their new investments might represent. 

The announced investments across both companies total $32 billion. ⁸ GLP-1s were 71% of Novo’s revenue in 2023, 16% of Lilly’s in 2023, and 26% of Lilly’s in 2024Q1. If these sales are proportional to the manufacturing capacity used to create those drugs, then about 40% of Novo and Lilly’s combined estimate of $45 billion in gross PP&E is for GLP-1s, for a total of $18 billion; $25 billion would then mean a 140% increase in GLP-1-relevant PP&E. However, a 140% increase in company-owned PP&E does not necessarily mean a 140% increase in production capacity, because much capacity currently is contracted out. If company-owned production grows 140% but contracted production is unchanged, then total growth will be much lower. Contractors may scale up in parallel, or the new company-owned facilities may displace them, which seems likely for WuXi AppTec. Lastly, we’re also likely to see even more investment by Novo and Lilly announced between now and 2030.

Ultimately, the companies are investing more in PP&E than I would’ve expected, which means I need to update. Increased investment is significant both in itself and for what it tells us about the companies’ future intentions. In light of this, I increase the base rate estimates for US volume by a pretty sizable 30%, with a standard deviation of 5%. This means I’m estimating most likely a 20%–40% increase in volume in light of this information.

This would tentatively bring median estimated patient-year supply up from 18 million to around 24 million.

Insurance treatment

Next let’s consider insurance treatment. In the United States, insurers are typically required to cover all FDA-approved medications that are “medically necessary.” While diabetes falls into that category, obesity does not. Medicare is actually prohibited by law from covering medication prescribed for weight loss. This makes the drugs much more expensive for patients when used only for weight loss, which should lead to lower use. 

About 140 million Americans are obese. 38 million have diabetes, but most of them — approximately 31 million — are obese as well. This makes the total patient pool for GLP-1s approximately 147 million people, a bit over a quarter of whom can get GLP-1s prescribed for diabetes and covered through their insurance.

And insurers have a strong incentive to keep coverage limitations in place. Right now, an annual supply of Ozempic or Mounjaro costs roughly $3,000. If our supply estimate of 22 million patient-years is accurate, that would mean a total annual spend of $66 billion — more than 10% of current spending on all pharmaceuticals in the United States combined. Insurance companies and Medicare may seek to restrict access to these drugs to keep premiums down.

However, expanding coverage for these drugs is a popular idea. In one poll, 61% of people stated Medicare should cover GLP-1s for weight loss. There’s also the chance that the government repeals Medicare’s ban on weight-loss drugs, which increasingly feels antiquated now that we have safe and effective treatments. And finally, we shouldn’t underestimate pharma’s famously strong lobby — or advocacy from the millions of patients who stand to benefit. 

So currently, insurance coverage is quite restricted for GLP-1s, which should limit sales even as shortages become less severe. However, it seems like the desires of patients and most lobbyists are aligned in wanting to remove those limits. Medicare could also be able to negotiate the price of Ozempic as soon as 2025, and lower prices may make coverage more palatable. However, it’s still possible that prices will remain stubbornly high in the commercial (non-Medicare) market and that insurers will be able to restrict coverage enough to slow down demand. 

Combining all of those pieces, I’d estimate (very roughly) that there’s a 50% chance that nothing changes, which reduces volume by 10% relative to the base rate drugs with no restrictions, a 40% chance of GLP-1s getting coverage parity, so no change from the base rate, and 10% chance that insurers restrict coverage even more, reducing volume 20%. On average, this is a 7% adjustment downward relative to the base rate.

Patient adherence

The second consideration is that so far, GLP-1 drugs have had low adherence rates. One study showed 68% of patients who started taking a GLP-1 drug for weight loss weren’t taking it within a year. Some percentage of this is due to supply shortages — there’s simply not enough to go around. But there are concerns that people are discontinuing the drugs once their weight loss has plateaued, once they’re no longer able to afford it, or due to side effects like nausea, diarrhea, vomiting, and loss of muscle mass. This discontinuation rate is much higher than those of the base rate drugs. Statins and TNFa inhibitors have one-year adherence rates of 59% and 73% respectively. 

GLP-1s’ net turnover rate ⁹ may be considerably higher than that of comparable drugs, meaning it may be harder to reach new GLP-1 patients. However, if turnover is driven by seasonality or people hitting weight-loss plateaus, then lapsed patients may become eligible again in the near future. On the other hand, if side effects become more well known as time goes on, there’s a chance that puts a dent in demand. Still, demand is so high relative to supply, though, that this probably won’t be a major issue. I’d say there’s a 20% chance of an 8% decrease in volume resulting from these concerns.

Pipeline drugs

The last factor I considered is other obesity drugs currently in development. There are several ongoing phase 3 clinical trials for obesity drugs that could lead to subsequent FDA approval if successful. It’s likely though not certain that these will be roughly comparable to existing treatments in terms of efficacy. And most of them are produced by Eli Lilly or Novo Nordisk, so we should expect their introduction to mostly cannibalize sales and manufacturing capacity from existing drugs, rather than widen the market. 

However, innovations in the pipeline might help raise adherence or improve production efficiencies. Orforglipron, for example, is a daily oral pill, as opposed to a weekly injection. Rybelsus, which is daily oral semaglutide made by Novo Nordisk, showed up to 17.4% weight loss at 68 weeks in a Phase 3 trial but is not yet approved for weight loss. Orals don’t depend on the complex injection pen manufacturing process, but they also use much larger doses of active ingredients than injectables do. A typical maintenance dose of the injectable Ozempic is 0.5 mg per week. The equivalent dose for Rybelsus is 7 mg per day. This means a manufacturer can create many more doses of injectables than orals with the same ingredients. Daily oral alternatives are likely to be available by 2030. But given their comparable efficacy, higher ingredient requirements, and the fact that the manufacturers overlap, I don’t believe that the current pipeline merits an adjustment.

In sum, there’s three adjustments to be added. Supply investments: a 30% increase with standard deviation of 5%. Insurance coverage: 50% chance of 10% decrease, and a  10% chance of a 20% decrease. Side effects and adherence: a 20% chance of an 8% decrease. 

Applying those adjustments to the original estimate, we get our final prediction:

My forecast predicts that the supply of GLP-1s will increase from eight million patient-years to roughly enough for approximately 23 million Americans by 2030. Still, this is only enough supply for about 15% of the 147 million Americans with diabetes or obesity. 

Most patients who want them will not have access to these drugs in the near future. The need is vast, especially abroad, where rollouts are being delayed until demand in the United States can be met. And as indications continue to be added for conditions ranging from heart disease to potentially even alcoholism, demand will only grow. It will likely be many years before these drugs become widely available to the patients who need them. It takes a long time to make a miracle.