The pool at Resorts World Las Vegas is, in the end, just a pool. Fifty metres of chlorinated water under a temporary roof, walled in by a custom-built complex on the Strip, ringed by 2,500 invited guests, two hundred journalists flown in from twenty-five countries, and — should anyone doubt the fifty metres really measured fifty — tape measures, available on request. Last weekend, some fifty athletes swam, sprinted and lifted weights in the inaugural Enhanced Games — the first international sporting event in which the use of performance-enhancing drugs was not merely tolerated but advertised. The total prize pot was $25 million. The Killers played the closing party.
It is tempting to dismiss the whole thing as a freak show — and many sports administrators, including the World Anti-Doping Agency, have done exactly that.2 The results, when they came, did not obviously help the organisers' case. Across an evening of swimming, sprinting and lifting, with every competitor free to use whatever America's drug regulator permits, exactly one world record fell19 — and even that, within hours, was being disputed on social media, with viewers pulling the race footage apart frame by frame. The organisers called the criticism "internet drivel".22 The Greek swimmer Kristian Gkolomeev — a four-time Olympian who finished fifth in Paris and had never stood on a global podium — swam 50 metres of freestyle in 20.81 seconds, a full tenth of a second inside the record César Cielo set in 2009, and quicker still than the 20.89 he had clocked in a North Carolina time trial in February.3 He was 31, ten pounds of lean muscle heavier than his Olympic self, on a cycle of drugs he declined to name, and a million dollars richer by the end of the night.
Almost everyone else fell short. Ben Proud went 22.32 in the 50 metres butterfly, agonisingly outside the record; Hafthor Björnsson's 475-kilogram deadlift was thirty-five kilos shy of the mark.19 The doping, it turned out, was not the button the marketing had implied. And the night's neatest irony was saved for the marquee event: the 100 metres, the race the whole format exists to sell, was won by the American sprinter Fred Kerley in 9.97 seconds — racing clean, by his own account, against a field that was not.20 James Magnussen, who had promised to "juice to the gills" and break a record, finished fourth in the 100 freestyle and went home with a participation cheque.21 A spectacle built to advertise pharmacology delivered its headline victory to the one man who declined it.
Mr Gkolomeev's record will not, strictly speaking, count. World Aquatics will not ratify it; the suit he wore is illegal; the substances are banned. But this is to miss the point. The Enhanced Games are not really competing with the Olympics. They are competing with the idea that there is a clean line between treatment, prevention and enhancement — and with the technology that polices it.
The interesting question is not whether a man on testosterone can shave a tenth off a sixteen-year-old swimming record. It is what is dissolving the line between treatment, prevention and enhancement altogether — and that, it turns out, has remarkably little to do with the chemistry on display in Las Vegas. The short answer is artificial intelligence, applied to a class of drugs called peptides, manufactured, as it happens, mostly in Switzerland.
Neither sick nor well
Aron D'Souza, the Australian lawyer behind the Enhanced Games, is fond of a phrase he uses on every podcast: that the Games exist to "enhance humanity". His backers — Peter Thiel and Christian Angermayer among them — have been more candid.4 Mr Angermayer calls himself a "human maximalist". Mr Thiel, when he is being colourful, talks about not dying.
Beneath the swagger is a regulatory observation that is, if not original, then under-appreciated. Every medicines regulator in the rich world is set up to evaluate treatments for disease. Swissmedic, in Bern, approves drugs that demonstrate efficacy against a specific indication: heart failure, breast cancer, type 2 diabetes. The European Medicines Agency does the same in Amsterdam; America's Food and Drug Administration does the same in Silver Spring. None of them is set up to approve a drug intended to make a healthy 31-year-old swim faster. Nor is any set up to approve a drug intended to make a healthy 65-year-old live another decade in good health. The categories are treatment and prevention. The thing that does not fit is enhancement.
This is awkward, because the molecules used for one are remarkably close to the molecules used for the other. Consider semaglutide, the active ingredient in Ozempic. It is a 31-amino-acid peptide engineered from the gut hormone GLP-1, originally approved for diabetes, then obesity, and now under investigation for Alzheimer's, addiction, cardiovascular events and chronic kidney disease.5 A meta-analysis published earlier this year suggested it might reduce all-cause mortality. In about five months, semaglutide's patent will lapse in India, China, Brazil and Canada, and the price will collapse from hundreds of dollars a month towards a generic floor of perhaps six.6 A drug invented for diabetes will, within a decade, look more and more like a longevity intervention.
The loop
Until recently, finding a peptide like semaglutide took twenty years and a great deal of luck. That has now changed. Generative AI — the same family of models behind ChatGPT, repurposed for biological sequences — is producing peptide and protein candidates at industrial scale.
The pace is startling. Profluent, a Californian firm that raised $106 million in December, was the first to show that large language models trained on protein sequences can design functional proteins from scratch.7 Generate:Biomedicines, which signed a deal worth up to $1 billion with Novartis this year, took its AI-designed antibody GB-0669 into the clinic in seventeen months against an "undruggable" target — roughly a tenth of the historic norm.8 Isomorphic Labs, the DeepMind spin-out, closed a $2.1 billion Series B on 12th May and now has offices in London, Cambridge MA and Lausanne.9 AlphaFold 3 was open-sourced for academic use last year and has been cited more than nine thousand times.10
What ties this together is not any single model but a loop. AI designs a molecule. Robots synthesise it. Wet labs test it. The data goes back into the model, which designs better molecules. Roche is now running this loop on 3,500 Nvidia Blackwell GPUs — the largest hybrid-cloud AI infrastructure in the pharmaceutical industry — through Genentech's "Lab-in-the-Loop" programme.11 Aviv Regev, Roche's chief scientist, has said that almost every small-molecule programme in the company now incorporates this approach.
For a chief technology officer, the relevant observation is that drug discovery has become, in about four years, a compute-and-data problem rather than a craft. The implications for capital allocation, supply chains and Swiss competitive advantage are not subtle.
Which is to say, here
The company that turns most of these designs into actual drugs sits half an hour's drive from Basel, in the village of Bubendorf. Bachem is the world's largest contract manufacturer of therapeutic peptides, and the practical bottleneck through which much of the GLP-1 boom is now passing.12 If Profluent or Generate designs a peptide in California, there is a reasonable chance that Bachem, or one of its handful of competitors, will end up making it.
This is the part of the story that should interest Romandy directly. The Biopôle campus in Lausanne raised CHF 756 million in its 2025 cohort.13 EPFL labs — Bruno Correia's group, in particular, with its open-source BindCraft platform — are producing some of the year's most-cited papers on AI protein design.14 Isomorphic Labs has a Lausanne office.9 Two of the world's three largest pharmaceutical companies are headquartered within an hour of Lake Geneva. The infrastructure of the AI-bio era is not somewhere else. A great deal of it is here.
The longevity case follows from the engine. Ageing is the largest single risk factor for cancer, dementia and most of what kills people in rich countries. Slowing it, even modestly, might do more for human health than every individual disease programme combined. The trouble has always been that regulators do not recognise ageing as an indication, and that no one knew how to design drugs for healthy biology in any case. AI is starting to solve the second of those problems, which will eventually force the first.
The FDA's hand has already been forced, oddly enough, by dogs. In January, Loyal, a startup developing a daily pill called LOY-002 for canine ageing, became the first company in any species to have a lifespan-extension drug accepted by the agency's veterinary division on the safety endpoint.15 Conditional approval is expected this year. Ageing-as-an-indication has walked in through the back door, on paws.
The caveats one should not skip
A great deal of what is offered in the longevity-and-performance space remains unproven. The biological-age "clocks" sold to consumers are, as a recent npj Aging paper put it, of uncertain incremental value over routine risk scores.16 Bryan Johnson's much-mocked Blueprint protocol is a brilliant marketing exercise, not a clinical trial. The Enhanced Games themselves carry real risks: EPO and anabolic steroids do not become safe because a doctor signs off, and broadcasting all this to an audience that includes teenagers is, as critics say, irresponsible.
The inequality concern is also genuine. If a class of expensive AI-designed interventions turns out to add five healthy years of life, those years will not be evenly distributed. "Enhancement capitalism", as the philosopher Nick Agar calls it, is already visible in cosmetic dentistry and designer handbags; it would have a sharper edge applied to lifespan.17
And AI-designed drugs are not yet approved drugs. Isomorphic Labs slipped its first-in-human trials by a year.18 The clinical readouts that will validate or sink the AI-bio thesis are still mostly ahead of us.
The Killers played the closing party.1 Nobody, as far as we know, was killed in action. On the contrary: under the lights, in full view of an invited crowd, a small group of athletes were quietly helping the rest of us work out what it takes to live a little longer.
Further reading: Natasha Loder, "Dope and glory: inside the Enhanced Games", The Economist 1843, 21 May 2026 — the long-read this essay was provoked by. And STAT, "Longevity startup Retro Biosciences says latest fundraising values it at $1.8 billion", 22 May 2026 — the most recent valuation milestone in AI-adjacent longevity.