Project Jupiter's loans are trading at stressed levels. Six AI groups now carry more spread risk in US investment-grade than the six largest banks combined. This is no longer just an equity story.
Two desks are watching the same junction this morning. The Dark Side Of The Boom and ING Think both have rates at the centre, though they come at it from different angles. The Dark Side Of The Boom is focused on the 10-year yield approaching 5% and what holds it there. The MOVE Index has stayed contained even as the level climbed, which means the rise has been persistent without becoming disorderly. ING Think, covering oil and the dollar separately, is reading the same yield environment through the lens of how central banks are pricing their options.
On energy, both desks see the Trump-Xi summit as a variable. ING Think published on oil cooling from recent highs. The Dark Side Of The Boom frames the summit explicitly as an energy event: a credible path to de-escalation could reduce oil's inflation impulse and give Treasuries room to rally. Both desks read the diplomatic story through its macro consequence rather than its geopolitical surface.
The equity market is treating the AI buildout as a growth story. The credit market is treating it as a financing problem. That gap is where the tension lives today.
The Dark Side Of The Boom has published three pieces on AI credit in this window. Project Jupiter's loans are trading at stressed levels. Healthy loans trade near par. That discount does not mean default is near, but it does mean credit investors are demanding a larger cushion before accepting exposure to a project facing construction delays and questions about the credit quality behind it. Separately, The Dark Side Of The Boom notes that six major AI groups now account for more spread risk in the US investment-grade bond market than the six largest banks. Bloomberg's John Authers argues that the credit market is the first place genuine doubt about the AI buildout should appear. Spreads have not yet signalled serious stress. But the test is getting harder as long-dated yields rise and project timelines slip.
The thing about a 10-year yield near 5% is that it changes the arithmetic on almost everything else simultaneously. Higher rates are manageable when nominal growth and profits are accelerating alongside them. They are far less forgiving when the source is an inflation shock, which raises financing and operating costs without guaranteeing stronger demand to offset them. The Dark Side Of The Boom makes that distinction clearly, and it is the right one.
The Project Jupiter situation crystallises it. The structure is project-finance debt, notionally ring-fenced from Oracle's balance sheet. But The Dark Side Of The Boom notes that Oracle is the campus tenant, OpenAI is the ultimate computing customer, and the economics still depend on Oracle delivering capacity under its computing agreement. The structure that was meant to isolate the risk has instead concentrated it. The rising real cost of capital, as The Dark Side flags, is the most credible brake on AI capital expenditure right now.
The Trump-Xi summit adds a wildcard. The Dark Side Of The Boom notes that crude has already priced some diplomatic relief, which means the upside response could be larger if negotiations disappoint than if they succeed. That asymmetry matters for where the 10-year yield settles, which matters for every other number in this picture.
The 5% level on the US 10-year yield is the test the Dark Side Of The Boom names explicitly. Watch whether the MOVE Index stays contained as the level holds or breaks, because a disorderly rise is a different market than a persistent one. On the AI credit side, watch whether Project Jupiter's loan prices stabilise or continue to widen, and whether the Trump-Xi summit produces any change to the energy and inflation outlook that would shift the yield path.