AI Junk Bond "Blast Radius": $17.3 Billion in Bonds Now in the Danger Zone
nashnova research
Of the 20 AI data-center junk bonds Morgan Stanley tracks, seven totaling $17.3 billion are now flagged as high-risk — down roughly 5 points since late September, about four times the drop of the rest. A blue-wave midterm result and a closed construction window are squeezing these bonds on both policy and timeline.
A $45.6 billion AI junk-bond pool — how concentrated is the risk?
Morgan Stanley's 20-bond universe totals $45.6 billion, backed by 468 MW of delivered compute. But roughly 2,512 MW is still scheduled for delivery by end-2027, with another 744 MW slotted for 2028.
This means → delivered capacity is only a fraction; the vast majority remains on paper. The bonds' repayment logic rests on those projects coming online on time.
Seven of the 20 have dropped about 5 points since late September — roughly four times the decline of the remainder. In plain terms = the market is already pricing a distinct risk premium into this subset.
After the blue wave, which bonds are directly exposed?
Of the 20 bonds, nine sit entirely in states that just flipped to Democratic governors: six in Texas (Black Pearl, Stingray, Galaxy, Cipher, Flash, and SoftBank's SE Cosmos), one in Georgia (Meta's Clearspark), and two in Storey County, Nevada (Tract bonds).
A tenth — CoreWeave (ticker CORZ) — has a 590 MW project spanning Texas, Georgia, Oklahoma, and North Carolina.
Together, these 10 bonds total about $24.3 billion — 53% of the pool — and correspond to roughly 2.1 GW of compute. This means → more than half the AI junk-bond exposure lands squarely in territory where the policy winds just reversed.
Virginia — is the 'data-center corridor' safe?
Two more bonds sit in Virginia: DDC's Richmond campus and Yondr's Loudoun County project.
Jefferies notes that moratorium-related legislation has already advanced ahead of Virginia's January 13 legislative session — and Virginia is the core of America's "data-center corridor."
This reflects a risk that extends beyond the newly blue states: even the most data-center-friendly jurisdiction in the country is tightening its stance.
The Texas legislative timeline — why is May 2027 the deadline?
The core risk is a timing mismatch. Wells Fargo's analysis shows the Texas election result converted what had been a governor-led audit pause into a pause tied to the legislative calendar — and the Texas 90th Legislature does not adjourn until May 31, 2027.
In plain terms = approval authority shifted from one governor's desk to a full legislative process — and that process doesn't wrap up until late May 2027.
Morgan Stanley had already flagged October as the last possible construction start to ensure projects go live by end-2027. October has passed, and the election reshaped the political landscape.
Strip out completed projects and protected bonds — what's still exposed?
After removing projects that will be finished by then and the one bond with a "rent certainty" clause, roughly $17.5 billion in bonds in newly blue states have completion dates beyond the Texas adjournment — Galaxy, both Tract bonds, Clearspark, and CORZ.
Governor-elect Gina Hinojosa's proposed moratorium targets new data-center permits and grid-interconnection approvals; projects already under construction could theoretically be exempt. But her platform also requires large power loads to bear their own generation, transmission, and interconnection costs, and phases out data-center tax breaks worth at least $1 billion a year.
This means → even if projects are not directly halted, the cost structure could be rewritten — no more tax breaks, grid costs internalized. The economics of every affected project need recalculating.
For investors holding these bonds, what is the most immediate question?
Morgan Stanley's report states plainly that project-level financing is most sensitive to interest rates: "where lease economics are not strong enough, the current environment may pose additional near-term challenges to supply."
In plain terms = if a project's rental income cannot cover its financing cost, a rate increase or a policy shift can stall it — and bondholders' principal takes the hit directly.
The most immediate question: can these projects secure enough policy certainty before the Texas legislature adjourns in May 2027? The window is already narrow, and the margin for the market is thin.
市场有风险，内容仅供研究参考，不构成投资建议。