Peptides: A Once in a Lifetime Trillion Dollar Revolution in All of Public Health

The software of the human body

Peptides work by mimicking the body's hormones, binding to specific receptors to regulate processes like appetite, metabolism, and cellular behaviour. GLP-1 peptides slow digestion, increase satiety, and improve insulin response, effectively reshaping how the body manages energy and disease risk.

Because they act across multiple biological pathways, their impact extends beyond weight loss into cardiovascular health, inflammation, and even neurological signalling, positioning peptides as a programmable layer of human biology.

Blood-brain barrier peptides targeting Alzheimer's represent a c.$30bn+ untapped market, and peptide-drug conjugates are opening a new frontier in oncology, delivering toxic payloads directly to cancer cells while leaving healthy tissue intact.

Two structural shifts are converging simultaneously:

• Oral transition is already under way, and this will accelerate manufacturer revenues sharply. Oral Wegovy launched earlier this year and is setting weekly new script records, with GLP-1 prescriptions moving from 200,000 to 300,000 per week in a matter of months. Oral formulations require ten to twenty times more peptide per dose than injectables. Manufacturer revenues can accelerate sharply without a single new patient entering the market.
• Regulatory change will kill the grey market this year. On 23-24 July 2026, the FDA's Pharmacy Compounding Advisory Committee finalises the 503A Bulks List, widely expected to shut the grey market of unregulated compounding clinics supplying peptides outside the licensed channel. The compounding episode proved something important: patients were willing to accept unknown manufacturing risk just to access the benefits at half the price. That demand does not disappear when the grey market closes. It redirects into the regulated channel, and only GMP-certified manufacturers can absorb it.

Peptides are not only the backbone of GLP-1s, but are fast becoming foundational to the longevity movement, telling cells when to repair, grow, reduce inflammation, or die. The Enhanced Games in Las Vegas this May will draw significant publicity, both good and bad, for peptides.

The Cognition Tax

"They are doing it again, and it's so damn obvious. When you start to think of Chinese intelligence in the same way you think about Chinese solar panels and cars, the world of AI starts to look very different as cheap Chinese tokens flood the market."

The numbers

Chinese AI tokens trade at roughly one-sixth the cost of their American equivalents, at comparable performance. In a single week in February, Chinese models processed 4.12 trillion tokens versus 2.94 trillion for US models. On OpenRouter, Chinese models account for 61% of top-10 global usage. Four of the five most-used models in the world are Chinese. Silicon Valley isn't just aware of this; it's becoming dependent on it, with 80% of US startups now building on Chinese base models.

The most innovative tech ecosystem in human history is quietly, systematically offshoring its cognitive infrastructure to a geopolitical rival.

A token gesture: what tokens actually matter

A single API call now involves multiple layers: input tokens, output tokens, reasoning tokens, cached tokens, tool-use tokens, and video tokens. Each is priced differently, and each consumes compute at varying rates. The generic "cost per token" is already obsolete. What matters is the mix, and how efficiently you convert watts into useful work.

When you're attempting to automate entire business functions at scale, the cost structure of intelligence becomes mission-critical. Jensen Huang said it plainly at GTC 2026: if a $500,000-a-year engineer spends only $5,000 in tokens, he'll "go ape." Token budgets are joining salary and equity as a measure of how talent is deployed.

The electron gap

Why is Chinese compute so cheap? Two reasons: electricity and efficiency. By 2030, China is projected to carry 400 gigawatts of spare capacity, triple the projected demand of the entire global data centre industry. America, meanwhile, has a 241 GW data centre pipeline with no credible plan to power it, and an enormous skills gap in areas like nuclear that might help close the shortfall. At CERAWeek in March, grid operators told developers to "get more flexible." This is why stocks with excess capacity, like WhiteFiber and Iren, have been on a march.

c.40% of 2026 US data centre projects are delayed by more than three months, and over 60% of 2027 projects show no construction activity as of April 2026, despite a c.50GW pipeline. Labour shortages, equipment constraints and permitting delays are driving costs up to 30% higher. The gap between announced capacity and physical progress is widening.

Three consequences nobody wants to say out loud

1. Sanctions created the competitor. Washington's chip export controls were designed to kneecap Chinese AI development. They did the opposite. Denied access to Nvidia's best silicon, Chinese engineers became ruthlessly efficient. China's leading model now trails the US frontier by just 2.7% on benchmarks, while costing 23 times less to run.

2. Distribution is flipping. For five consecutive weeks, Chinese models have outprocessed American models on OpenRouter. The default inference layer of the global AI economy is being set in China, not San Francisco.

3. A new digital ecosystem in China's hands. Every business on earth pays a tax to American platforms just to reach its own customers. Now ask yourself: what happens when intelligence works the same way? When every knowledge worker runs on ten AI agents, and those agents bill by the token, token pricing becomes the corporate tax rate. A sixfold cost advantage, compounded across every knowledge business on earth, is an economic gravity well. Capital flows toward cheap cognition the same way water flows downhill.

The verdict: the AI race was never just about who had the best model. It was always about who had the cheapest intelligence at scale. That is a question answered in gigawatts, not parameters. America has the brains. China has the muscle. The electrons. The spare capacity. The structural cost advantage. In a commoditised token economy, muscle wins. The rent we once paid for US distribution in advertising is becoming the rent we pay for cognition. The only question is who collects it.

What drove the Collective Ideas this week

WoW: +2%

This week was more earnings-driven than thematic headlines suggested, but the earnings tape still mapped cleanly onto the themes we follow through the Conviction list. For context, the Conviction ideas are a concentrated list of publicly-listed companies we believe best express the big investment themes that we as a collective support: AI infrastructure and its second-order demand pull (power, grids and industrial capacity), power scarcity and reliability, a turning industrial and capex cycle, real assets and commodity-linked inflation hedges, security and resilience spend, and the institutionalisation of digital assets. Against that backdrop, the week's action was still coherent. AI infrastructure remained the organising principle, with capex commentary pulling demand through power, industrial capacity and supply chains behind compute. Cyclical green shoots were confirmed in capital equipment and mining, with stronger orders, an easing of destocking and a higher service and aftermarket mix supporting multi-quarter visibility rather than a one-quarter spike. Capital return became a signal too. Buybacks topped up mid-cycle, not at peak earnings, read as management conviction in cash generation. The net effect was a week where individual prints moved stocks, but the ideas' exposure still pointed to infrastructure bottlenecks, tangible cashflows and the physical constraints defining the next phase of AI.

Best content

🎧 AI and the End of Traditional Institutions

Why Watch? Raoul Pal, Peter Diamandis and Salim Ismail break down how AI is accelerating fast enough to force a full reset of companies, labour markets and governance structures. The discussion moves beyond hype into second-order effects, from the collapse of traditional business models to the risk of social disruption, while making the case that this shift could ultimately create a more decentralised and abundant system. Watch here.

📚 Nature as Critical Infrastructure

Why Read? This reframes climate risk from an environmental issue into a core economic and national security concern, arguing that nature should be treated as critical infrastructure. With biodiversity collapse, food system fragility and climate shocks converging, the piece highlights how current agricultural and policy frameworks are undermining long-term stability, and why resilience will depend less on new technology and more on recognising and protecting the systems we already rely on. Read here

Stock of the week

Atome and the Fertiliser Bottleneck

Atome is a project-led fertiliser story sitting in the middle of a tightening global supply market, where geopolitical disruption and transport chokepoints are exposing how fragile nitrogen supply can be. Rather than trying to win as a commodity producer, Atome is building a low-carbon, locally anchored fertiliser platform designed to benefit from exactly that constraint.

The key question is execution. The company has now pushed its Villeta project into the build phase with funding in place, but the investment case still depends on converting a strong strategic narrative into delivered capacity. That makes it less a near-term earnings story and more a catalyst-driven infrastructure play.

If the project scales as planned, Atome offers leveraged exposure to one of the most important bottlenecks in global food production: fertiliser supply.

Read full investment case here

New or updated showcases

Diginix (DGNX US, $54m mkt cap)

Diginex provides ESG, climate and compliance software that enables organisations to collect, manage and report sustainability and supply chain data, and is evolving into a fully integrated platform following acquisitions including Plan A, Matter, The Remedy Project and Resulticks. This shift expands the story beyond compliance into a broader AI-driven data and analytics opportunity, with early SaaS traction and a growing customer base positioning the business at the intersection of regulation, data and capital markets. Read the full investment case here.