Government debt is at record highs, and it isn't just America's problem. France, Japan, and the UK are in a similar situation. At the same time, corporations are issuing debt at a pace we haven't seen before. The companies that made their name being asset-light are now borrowing heavy in order to fund a massive AI infrastructure build.

Somebody has to buy all that credit, and buyers are charging more for it. That's what a yield is: the return a lender earns for parting with money, or the price the borrower has to pay. In the US and UK, 30-year government debt now costs over 5%. Five years ago it was closer to 1%.

Which brings us back to businesses. For fifteen years, debt was almost free, so balance sheets barely mattered. All of a sudden, they matter again. A big debt burden can very quickly turn into a big liability and blow-up risk is higher than it was just 5 years ago.

So when you're looking at a business now, three questions worth asking.

Here is an example of a company we would be watching closely.

Moderna was up 177% on Wednesday, after it and Merck said their personalised mRNA cancer vaccine hit its goals in a Phase 3 melanoma trial. Intismeran, given alongside Merck's Keytruda to 1,137 patients whose high-risk melanoma had been surgically removed, delayed recurrence and reduced the risk of the cancer spreading.

The wider pharma industry is having a moment, too: most pharmaceutical industry-themed ETFs are handily beating the broader S&P 500 index this year to date. Some of that is the industry starting from a multi-decade valuation low; more of it is the tariff and drug-pricing overhang lifting earlier this year.

But if you’re thinking about investing in pharma, just remember what you're buying. This is an industry that’s notoriously difficult to get right. Patents run twenty years from filing, but a decade of that is eaten by trials, leaving roughly eight to ten years of real selling time. Deloitte puts the industry's return on R&D investment at 7%, which is below what these companies pay for capital. Growth comes mostly from buying other people's pipelines (and we know most managements overpay for acquisitions), and one government policy can rewrite the income statement.

Here is an example of a company we’d take over any drugmaker.

OpenAI reports they grew revenue to $6.7 billion at the end of Q2, up 18% compared to the previous quarter. Anthropic, on the other hand, doubled its revenue to $11.6 billion in the same period. As if that was not enough, OpenAI’s operating margins continue to worsen as the operating losses outpaced the revenue growth.

Why it matters beyond bragging rights: OpenAI made hundreds of billions of dollars in capital expenditure commitments with companies like Oracle, Microsoft, Broadcom, AMD, CoreWeave, Amazon… There were already fears about them not being able to follow through on these payments, and this news is not re-assuring the market.

Large language models are seeing more competition now than ever, as Chinese firms catch up and enterprises realise the latest and greatest models are expensive for every-day use. If you have any AI-related investment today, it might be worth taking note of how much of today’s valuations depends on “future commitments” by models like OpenAI and Anthropic.

Here is a company with lots of AI-themed exposure and low reliance on any singular model.

Upgrade to unlock full access to the RatedA Portfolio (with weekly updates on every buy/sell and rationale), and the complete database of companies + ratings.