The AI infrastructure boom's appetite for capital showed up again overnight. Nebius Group priced a $5 billion offering of convertible senior notes, increased from the $4.5 billion it announced initially, with settlement expected August 24. A convertible note is debt that the holder can exchange for shares at a preset price, so investors collect interest while keeping a claim on the stock's upside.

The terms

The deal comes in two tranches. A $3 billion slice pays 0.50% and matures in February 2030, converting at about $313.46 per share, a 40% premium to the $223.90 closing price; a $2 billion slice pays 4.50%, matures in February 2034, and converts at about $324.65, a 45% premium. Both tranches accrete, with principal growing to 110% of face value at maturity for the 2030 notes and 125% for the 2034 notes. Net proceeds are roughly $4.94 billion, or up to about $5.68 billion if underwriters exercise their option for additional notes.

The premiums are the tell. Investors accepting conversion prices 40% to 45% above the market, on a stock that has already run hard, are effectively underwriting continued explosive growth in demand for AI computing.

What the money buys

Nebius says the proceeds will finance business growth, including data center construction, AI cloud development and GPU procurement. The Amsterdam-based, Nasdaq-listed company, built from the international assets left when Yandex split its Russian and foreign operations in 2024, has become one of the fastest-scaling independent AI cloud providers, renting GPU capacity to customers who cannot build their own fast enough, including a $1 billion multi-year compute deal with AI startup Reflection signed in July.

That growth is exactly why it needs the money: data centers and GPUs are paid for years before the contracts they serve are fully collected, so even fast-growing infrastructure companies are structurally cash-hungry.

The bigger pattern

Nebius is not an outlier; it is the template. Convertible debt has become the AI build-out's favorite financing instrument because it fits both sides: issuers get cheaper coupons than straight debt and defer dilution to much higher share prices, and investors get downside protection with a call option on the sector's momentum. The structure works beautifully while AI-linked stocks rise. The risk sits in the other scenario: a sector-wide slowdown would leave companies with billions in debt that never converts, due back in cash from businesses whose spending was premised on perpetual growth. This week's other market story, bond investors questioning how much AI-related issuance the market can absorb, is the same question asked from the other side of the table.