Everything About AI
NextFin News - The US convertible bond market has blown through its annual record with months to spare, as artificial-intelligence spending drives issuance to USD 85.5 billion across 127 deals by early June 2026 - already 61% of the USD 140.5 billion record set for all of 2025, according to Dealogic data compiled by ION Analytics. The surge is not a broad credit boom; it is a concentrated financing wave running through the AI value chain, and it is rewriting how technology companies pay for growth.
The market did not ease into the record. Issuers raised USD 49.5 billion across 52 transactions in the second quarter alone, surpassing the USD 35.9 billion raised across 75 deals in the first quarter. June was the strongest month of the year by volume at USD 21.4 billion - achieved on just seven transactions, the lowest deal count of any month, which means the average deal size swelled as the biggest names came to market. Globally, convertible issuance reached USD 84.6 billion across 96 deals in the second quarter, a 67% increase over the same period a year earlier and the largest quarterly total ever recorded, with North America accounting for USD 64.2 billion of that volume.
The AI Concentration Is the Story
Two-thirds of the market now belongs to one theme. Computers and electronics companies raised USD 56.8 billion across 43 deals, accounting for 66.4% of total US issuance volume in 2026. Utilities and energy issuers followed at a distant USD 7 billion across nine deals, and healthcare at USD 6.6 billion through 26 transactions. The technology share is not a rounding difference; it is a structural dominance unseen in previous convertible cycles, which were typically led by biotech and speculative growth names rather than the largest companies in the world.
The deal tape reads like a map of the AI buildout. Alphabet priced two separate tranches totaling USD 9.625 billion in June. Alibaba and Oracle each raised USD 5 billion. CoreWeave, the GPU cloud provider at the center of AI infrastructure, sold USD 4 billion of convertibles. DoorDash added USD 2.75 billion, Akamai USD 3 billion in zero-coupon notes, and Cloudflare USD 2.175 billion, also zero-coupon. Information technology led all sectors in the second quarter with USD 34.4 billion across 31 deals, driven by AI infrastructure, semiconductor demand, and data center expansion.
Roughly half of 2026 issuance is tied in some way to AI, according to Bank of America Global Research and Barclays Research. Companies are using the proceeds to fund data centers, power infrastructure, and cloud expansion while also rolling over pandemic-era debt.
A lot of it is to build out capital expenditure, particularly AI, and that's unusual and not something we've seen in previous cycles.
Michael Youngworth, managing director and head of global convertibles at Bank of America Securities, put it plainly: the financing is funding a buildout unlike any prior convertible cycle.
Why Convertibles Beat Debt and Equity
The convertible's appeal is mechanical, not fashionable. A convertible bond pays a coupon below what straight debt would demand because it embeds a call option on the issuer's stock - an option that gains value when equity volatility rises. In May 2026, with the benchmark 10-year Treasury yield at a 16-month high, that coupon relief was the difference between affordable and prohibitive capital for issuers facing enormous, front-loaded spending plans.
For AI-heavy companies, the choice narrows to three options, each with a different pain point. Straight debt at these yields loads the balance sheet with cash interest before the new assets earn their keep. Equity issuance dilutes existing shareholders immediately, often at a moment when management believes the stock does not yet reflect the AI payoff. The convertible defers both pains: minimal cash interest today, and dilution only if the stock actually rises above the conversion strike.
A lot of these issuers share the same situation: a massive expansion program ahead and substantial capital needs. They've realized that equity-linked instruments are a very effective tool to finance this kind of growth.
Ivan Nikolov, head of convertible bonds at Fisch Asset Management, sees the pattern across the issuer spectrum.
The terms confirm how much pricing power issuers hold. Cloudflare's notes, priced August 11, 2026, carry a conversion premium of approximately 60% - meaning the stock must rally more than 60% from its USD 310.59 reference price before conversion begins to pay off. Akamai's zero-coupon notes, priced May 19, 2026, convert at premiums of 42.5% and 35.0% to its USD 141.34 closing price. CoreWeave's April 2026 notes priced at a 30% premium to its USD 92 share price, with capped call transactions costing USD 492 million to blunt dilution. Tempus AI sold a six-year convertible with zero coupon and zero principal increase at maturity, converting only if shares reach USD 69.26, about 40% above the sale price.
These are issuer-friendly terms by any historical measure. Investors accept them because the embedded optionality gives them exposure to AI upside without taking the full credit risk of the underlying company. Hedge funds and large asset managers dominate the convertible investor base, harvesting relative value from the implied volatility embedded in the bonds. Long-only investors are "buying for exposure to semiconductors - the hottest part of the market right now, driven by AI capital spending," said Venu Krishna, managing director and head of U.S. equity strategy at Barclays.
I think deals are trading well and investors are happy to get exposure to AI through the convert market, so I'd expect that to continue. Right now everything is very strong and there's a lot of optimism. There needs to be the right balance: terms that are attractive for issuers but that also work for investors.
Bryan Goldstein, head of convertible bonds origination at Matthews South, endorsed the trend while flagging the one condition that could break it.
Cyclical Wave or Structural Shift
This is the question that determines how the story ends, and the answer has two layers. The surge is cyclical in one clear sense: it is driven by a discrete, time-bound capital-expenditure buildout. When the data centers are built and the GPU orders are delivered, issuance tied to that specific program will slow. Convertible arbitrage flows are also notoriously mean-reverting; when implied volatility compresses or the equity tape turns, the hedge-fund demand that supplies this market's liquidity can withdraw quickly. Historically, convertible arbitrage has accounted for as much as 80% to 85% of the market's issuance at peaks, and the strategy's 2004-05 downturn emptied the market for years.
But the financing behavior underneath is structural. Technology giants that funded almost their entire expansion through free cash flow in the 2010s and early 2020s have crossed a threshold: planned AI investment is now too large to self-fund. Hyperscalers are projected to need more than USD 1 trillion for capital expenditure in total by 2028 - roughly 3% of current annual US GDP. Once a company shifts from internal funding to external finance at that scale, it rarely shifts back; the capital structure absorbs a permanent new layer of debt.
The evidence for a regime change is in the scale, the breadth, and the durability of the driver. Morgan Stanley expects AI-related global debt issuance to more than double to a record approximately USD 570 billion in 2026. Goldman Sachs Research estimates nearly USD 500 billion of AI-related debt issuance so far this year, with hyperscalers accounting for only 40% of it. BNP Paribas data puts AI hyperscaler debt issuance at USD 220 billion in 2026 as of August 10, versus USD 12.5 billion in the comparable 2025 period - a seventeen-fold increase in twelve months. And the breadth now extends well beyond investment-grade balance sheets: WhiteFiber, which went public in August 2025 with a negative forward price-to-earnings ratio of about 36, raised USD 230 million through a five-year convertible in January, pricing its enterprise value at roughly 19 times forward EBITDA.
The verdict: a structural shift in how the technology sector funds itself, riding on top of a cyclical AI capex super-cycle. The buildout will end; the reliance on external capital will not.
The Second-Order Risk Nobody Is Pricing
The first-order story is simple: AI companies need money, convertibles are cheap, so they issue convertibles. The second-order implication is more uncomfortable. Convertible issuance at this scale is a signal that even the strongest tech balance sheets are reaching the limits of what free cash flow and straight debt can carry. Goldman Sachs Research projects roughly USD 250 billion of direct supply from the hyperscalers in 2026 alone, leaving project finance aside. That is a single cohort of issuers absorbing a chunk of global credit capacity that used to belong to entire industrial sectors.
When one theme commands that much financing, the risk transmission changes. Credit default swap costs on hyperscaler debt have risen, and spreads have widened - most for Oracle and Meta, less so for Alphabet, Amazon, and Microsoft. The convert market's embedded dilution is not a theoretical concern for existing shareholders; it arrives precisely when the stock is rising, but it also means that if AI returns lag the capex, the equity cushion that makes these bonds palatable evaporates at the same moment refinancing needs peak.
There is also a crowding-out dynamic that the record volumes mask. Some European issuers reportedly timed their issuance to avoid days when hyperscalers tapped the market, trying to preserve investor demand. When the largest borrowers in the world set the issuance calendar for everyone else, the market has stopped being a niche and become infrastructure - with all the systemic weight that implies.
The Counter-Thesis, and What Would Kill It
The strongest case against the structural read is that convertible markets are, at the margin, volatility-arbitrage markets, not corporate-finance markets. Hedge-fund arbitrage desks trade implied volatility against realized volatility and the underlying credit; they do not hold convictions about AI's terminal value. When implied volatility is rich and equities rally, converts print. When volatility compresses or the tape rolls over, the flow reverses - and it reverses fast, because the same delta-hedging that supports issuers in good markets becomes forced selling in bad ones. On this view, the 2026 surge is a function of market conditions, not a permanent change in how technology is financed, and the market will revert to its pre-2025 footprint once the AI capex cycle peaks.
That argument has force on the investor side but misses the issuer side of the trade. Arbitrage demand explains the liquidity; it does not explain why issuers are willing to pay for it. The issuer motivation - funding a multi-year, cash-flow-negative buildout without crushing near-term earnings per share - is a balance-sheet reality that survives any single volatility regime. Even if hedge-fund flows retreat, the financing need does not disappear; issuers would simply migrate to a more expensive channel, which is itself evidence that the cheap-capital era, not the need for capital, was cyclical.
The falsifying signal is concrete: if annual US convertible issuance falls back below USD 60 billion - roughly the pre-surge footprint - for two consecutive years after AI capital-expenditure growth slows, and if the technology sector's share of issuance returns to below 40%, the structural-shift thesis is wrong. A second confirming signal would be conversion premiums compressing back toward the 20% range on new deals, which would indicate that issuer pricing power has evaporated with the cycle.
What Comes Next
Issuance is likely to stay elevated through the summer despite the seasonal slowdown that usually accompanies corporate blackout periods. "Typically we'd expect some slowdown in the summer," Nikolov said. "But given the pace of these capex programs, I don't think this will be a typical summer." Beyond that, the path splits by time horizon.
In the short term, the market's direction depends on volatility and the equity tape. As long as implied volatility stays rich and AI-linked stocks hold their gains, new deals will clear at issuer-friendly terms. The watch item is the conversion premium on new issuance: a sustained move above 50% across the cohort signals term exhaustion, not strength.
In the medium term, the question is whether AI revenue begins to cover the capex. If it does, converts convert cleanly, shareholders absorb dilution on a higher stock price, and the market graduates into a permanent financing channel for technology capital expenditure. If it does not, spreads widen, terms deteriorate, and issuers with the weakest credits - the WhiteFibers of the market - find the door closing first.
In the long term, the structural shift to external finance is durable regardless of which AI winners emerge. The sector's capital intensity has changed permanently; the financing mix will reflect that even after the current cycle turns.
Three scenarios frame the outlook. The base case: full-year 2026 issuance surpasses the 2025 record of USD 140.5 billion, with terms staying issuer-friendly but gradually tightening into year-end. The upside case: AI revenue accelerates, stocks rally through conversion premiums, and the convert market becomes the default funding channel for tech capex. The downside case: AI returns disappoint, volatility collapses, convertible arbitrage exits, and the market freezes - forcing issuers back into straight debt at much higher coupons just as their refinancing needs peak.
The record US convertible market is not an AI bubble financed by cheap optionality. It is the first place a permanent change in how the technology economy is funded became visible - and the bill for that change will come due in dilution, not in cash.
Explore more exclusive insights at nextfin.ai.
