They are funding the artificial-intelligence build-out with paper, and they are doing it with your retirement account before the companies have earned a dime. U.S. companies have issued about $54 billion in convertible bonds so far this year, up 43 percent from 2025 and the highest volume since the early days of the pandemic, with artificial-intelligence firms leading the charge — many of them paying zero percent interest. Wall Street has been funneling record capital into the AI build-out across equity and debt for months, and the convertible market is the newest spigot. The question that should trouble anyone who still means something by the word conservative is: growth for whom, and governed by what?
The strongest honest version of the argument deserves a hearing, and I will grant it cleanly. CoreWeave is pouring concrete somewhere — it recently issued $4 billion in convertible bonds at one and three-quarter percent, and the data centers it is building require real workers, real steel, real electricity. Microchip Technology is fabricating the semiconductors that run the thing. Akamai Technologies, the cybersecurity and cloud-computing firm, issued $3.5 billion in zero-coupon convertible notes and sold them in a single offering. These are companies with employees and balance sheets and physical plant. The convertible bond market gives them cheap capital, and in some cases free capital, to build what they say they are building. Michael Youngworth, head of global convertibles strategy at BofA Securities, summed up the issuers’ wind at their backs: “high stock prices, tight credit spreads, and well-supported stock volatility. We have all three of those things right now.” The convertible index has returned more than 20 percent this year, outpacing both the S&P 500 and the Nasdaq. That is supposed to be the beauty of it: the market channels money to where the innovation is.
Now let me tell you what that sentence actually means when you translate it from the language of the trading floor back into the language of the county where I live.
I used to trade agricultural futures — paper claims on corn that had not been planted yet, for buyers who had never seen a field, priced by men who would never eat the crop. The mechanism was elegant. The mechanism always is. What it did to the people who grew the corn was something else.
The convertible bond is the same machine running on a new commodity. Strip away the Wall Street vocabulary and here is what happens: an investor lends money to a company. In return the company pays him nothing — literally nothing, in the case of Akamai’s zero-coupon notes. The investor accepts this because he is not really lending money. He is buying a side bet. If the company’s stock rises to a certain price, the bond converts into equity, and the investor gets to ride the wave. The bond is not a loan. It is a call option dressed in a loan’s clothing.
The tell is the zero. When a company can borrow money at no cost, it is not because the lender believes in the enterprise. It is because the lender believes in the stock price. The bond investor does not care what CoreWeave builds in your county or mine, or whether the workers who build it can afford to live there afterward, or whether the electricity it swallows comes from a grid that was already straining. He cares whether the stock moves. The data center is a line item in a prospectus. The county does not appear.
A zero-coupon convertible note sold into a market driven by AI mania is not a prudent allocation of savings to productive enterprise. It is a speculative bet that the share price will rise far enough, fast enough, to make the conversion worth more than the interest the bond never paid. The funding is cheap not because the underlying business is sound but because the share price is volatile — and in this market, the volatility is being manufactured by the same wave of enthusiasm that has driven AI revenue projections to historic extremes. Some analysts have begun asking whether the revenue projections behind these builds match any historical base rate, whether the money going in will come back at these valuations. The bond market is not waiting for the answer. It never has. A machine that can generate code or summarize a legal brief is a marvelous thing, but the capital flooding into it now is on a scale that requires these firms to become the most valuable enterprises in human history — and to stay there — before the conversion prices are even reached.
The deepest dishonesty of this moment, though, is not about the odds of a speculative bubble. It is about what kind of property we are building with all this cheap paper, and who will end up holding it.
The same investors who buy these bonds already own the common stock; the conversion feature sweetens their bet while insulating them from the downside. The risk that the stock price will never hit the conversion level falls on the bondholders, many of whom are pension funds and insurance companies — the pooled savings of ordinary workers — that chase yield in an ocean of near-zero real returns. If the AI dream materializes, the early investors get the equity upside. If it doesn’t, the losses are borne by the same retirement accounts that already carry the risk of the entire financialized economy, and by the communities whose pension systems bought the bonds because they had no better place to put their money. Credit spreads — the premium investors demand to hold riskier corporate debt — are near decade lows, which is another way of saying investors have stopped asking to be paid for risk.
We have done this before, and we know where it ends. In 2024 the convertible-debt market was flooded by cryptocurrency firms — Strategy, the bitcoin accumulation company founded by Michael Saylor, led the way — before issuance collapsed along with crypto prices. The machine does not learn. It changes the label on the can and keeps selling. The same investors who now explain, with perfect charts from BofA, why this time the volatility is different are the ones who, two years ago, explained why bitcoin was a new asset class that would never lose its correlation advantage. It will be different this time, they always say, and it is never different. The underlying business either generates real free cash flow or it doesn’t, and when the music stops, the bondholders who collected no coupon will discover that they were holding equity in disguise, at the worst moment to hold it.
This is what the modern conservative movement now means by the free market: a company that has never earned a dollar of profit issues four billion dollars in bonds at below-inflation rates, and the financial press celebrates it as growth capital. I know the argument. I made it myself, once, from the other side of the desk: the market is pricing risk, the market is allocating capital, the market is doing what the market does. And that is precisely the problem. The market is doing what the market does — treating the particular as fungible, the place as a container, the worker as a cost line, the community as an externality. The town where the data center lands will bear the water, the power draw, the strain on the school and the road and the grid. The bond investor in Stamford will bear nothing. He will collect his conversion premium or he will not, and either way he will never set foot on the concrete he financed.
And when I read the names — Akamai, Microchip, CoreWeave — I see something else: the abstraction of capital away from place. None of these firms is rooted in a particular town; none of them answers to a local board of directors who live down the road