401 Million Dollars on Cardboard Foundations
There exists a category of business that fascinates analysts for the wrong reasons: those that display extraordinary numbers precisely because they have outsourced everything that could cost them money to the periphery, including the truthfulness of their sales channels. Medvi, the telehealth startup powered by artificial intelligence, fits this category with almost didactic precision.
The facts are striking in their simplicity. The company, founded by Matthew Gallagher, operated during its peak growth period with just two employees. It generated $401 million in transactions last year, with $65 million in profits, and projects $1.8 billion in sales by 2026, according to The New York Times. The initial investment was a mere $20,000, allocated between advertising and software licenses for AI tools, including ChatGPT, Claude, and Grok—tools Gallagher utilized to build the website, serve customers, and generate content. From a pure efficiency perspective, the ratio is spectacular. From a business architecture standpoint, it raises significant alarms.
The issue isn't that Medvi used artificial intelligence to scale. The problem lies in what aspect of the model it replaced.
The Channel as Achilles' Heel
When reviewing the plans of a commercial building, the first question isn't how many stories it has but what columns bear the weight. In Medvi, the column that supports nearly all of the customer acquisition load is its affiliate network. Gallagher acknowledged that approximately 30% of the company's advertising operates through this system. In absolute terms, this translated to over 5,000 active campaigns on Meta simultaneously, according to the platform's public ad registry.
Outsourcing the sales channel to an uncontrolled network of affiliates is, in itself, an architectural decision with predictable consequences. The affiliate optimizes for conversion, not compliance. Their incentive is to generate clicks and sales; the legal responsibility for the claims they make rests, at best, in a gray area. When Business Insider analyzed Meta's ad library, it found profiles of supposed medical professionals recommending Medvi's products, several showing clear signs of having been created by artificial intelligence: watermarks from Gemini on profile pictures, illegible text in images, Angolan phone numbers linked to accounts that previously belonged to gospel musicians or real estate agents.
At least six pages of alleged doctors actively promoted the company's weight loss and sexual performance products at the time of the investigation. Between the Friday before the article and the following Monday, active campaigns fell from 5,000 to approximately 2,800, a reduction suggesting that media pressure—not internal controls—stopped the machinery.
This isn't a minor execution failure. It's a structural failure in the element generating 100% of the flow of new customers.
When Regulation Becomes a Cost Variable
In February, the FDA sent a letter warning that representations on the site medvi.io were "false or misleading" due to comparisons with approved drugs like Wegovy and images suggesting that the company directly composes the drugs it sells. The Federal Trade Commission received a formal request for an investigation in September, including Medvi among six telehealth companies cited for questionable advertising practices. This is further compounded by lawsuits with a scope that isn't fully documented in available sources.
What is analytically relevant isn't the list of regulatory actions per se, but what it reveals about the financial architecture of the model. Medvi built its growth curve on artificially depressed acquisition costs. If affiliates can make unverified medical claims, attribute them to non-existent doctors, and launch thousands of campaigns without oversight, the cost per acquired customer collapses. The margin of $65 million on $401 million in sales partially reflects the postponement of compliance costs to the future, when those costs manifest as fines, litigation, and operational restrictions.
Gallagher's official response to the evidence was to point out that the company has a disclosure policy for AI actors or representations and that it works to remove affiliate ads that violate this policy. The website includes a general notice stating that certain materials are generated or enhanced by artificial intelligence, and the company disclaims responsibility for their accuracy. This architecture of disclaimers, without operational verification mechanisms prior to publication, is functionally equivalent to installing a firewall on the top floor of a building without sprinklers.
Two employees cannot audit 5,000 active affiliate campaigns. That arithmetic is the most critical audit of the model.
The Model Promising $1.8 Billion No Longer Has the Same Engine
The projection of $1.8 billion in sales for 2026 relies on a growth rate sustained by the same channel now facing simultaneous restrictions from multiple fronts. Meta has already reduced the visibility of campaigns under pressure. The FDA has documented precedents for escalating from warning letters to more severe enforcement actions. The FTC has the authority to impose civil penalties, demand corrective advertising, and ban specific practices via injunctions.
Each of these pressures directly impacts the cost of customer acquisition. If affiliates can no longer make claims that led to conversions, they need distinct creatives, possibly featuring verifiable physicians and explicit disclosures. This increases the cost per click, reduces conversion rates, and compresses margins. Should advertising platforms implement credential verification for telehealth advertisers, the pool of eligible affiliates shrinks.
The model doesn’t automatically die under these conditions, but the piece that made it extraordinarily efficient ceases to function under the same circumstances. Replacing it requires what the company deliberately avoided building: a compliance layer, verifiable healthcare professionals, and an operational structure capable of sustaining that process at scale. This incurs a fixed cost that transforms the business’s economics from the ground up.
The $65 million in profit from last year may well be the last snapshot of a model that worked exactly as long as it wasn't audited. Companies don't collapse for lacking ideas to grow; they collapse when the elements driving their operational efficiency become incompatible with the conditions in which they must operate, and there exists no replacement ready to bear the same weight.
