China’s Petrochemicals Glut Turns to Crunch After Months of War
The actual shortages are happening now.
China has burned through much of a petrochemicals glut built up by years of overproduction as the US-Iran war chokes oil flows, setting the stage for higher prices for an array of industrial materials.
The squeeze is most acute in the so-called aromatics chain and the downstream products made from these chemicals, according to JPMorgan Chase & Co. and commodities data provider ICIS. That includes rubber for car tires and electrical insulation, resins, and plastic bottles, as well as various paints and coatings. (…)
“Right now, inventory levels are basically near the bottom,” said Parsley Ong, head of Asia energy and chemicals research at JPMorgan. The tightness in Chinese markets has already affected consumers with “inflation happening across most products,” she said.
The price rises have fed through into China’s producer price index, which tracks inflation at the wholesale level, and has gone from below zero in February to a faster-than-expected 5% in September. These price gains are also being exported, given the country’s role as a major supplier of petrochemicals and downstream products to global markets. (…)
Inflation going global
Rising inflation isn’t just a US problem. It’s happening globally, but the extent varies. This chart shows the latest CPI inflation across top countries and regions. The large European and North American economies are mostly in the 3%-4% range.
While less-advanced economies generally have higher inflation rates, Turkey and Argentina are in a class of their own. (Mauldin Economics)
- The energy threat persists:
Data: FactSet; Chart: Courtenay Brown/Axios
THE US HOUSING PICKLE
While aspiring homeowners can’t afford a new house,
existing owners do not want to sell, since moving would mean swapping a cheap mortgage for a dearer one. If rates stay at 7.4%, our analysis suggests, this “lock-in” effect could prevent a fifth of sales that would otherwise have taken place. (…) unlike borrowers in Britain and Canada, Americans cannot generally “port” their mortgages to a new property, forcing those with cheap loans to pay higher rates when they move. (…)
In previous cycles no more than 40% of borrowers had mortgages even one point below the prevailing rate. Today 87% do, and 76% pay at least two points less. (…)
How much does moving cost existing mortgage-holders? To find out, The Economist analysed data from the Federal Reserve’s latest Survey of Consumer Finances. Conducted every three years, it includes some 1,700 mortgage borrowers.
Using their incomes, their outstanding balances, the number of years left on their loans and their interest rates, we estimate that the typical borrower would pay an extra $335 a month in interest, or 3.2% of their pre-tax household income, to sell their home and buy an identical one at today’s market rate of 7.4%.
In dollars, mortgage-holders in the top fifth of earners would face the biggest bill ($558) but, relative to income, middle earners would be hit hardest (4.4% of income). (…)
By discouraging homeowners with cheap mortgages from selling, it reduces the supply of homes on the market, putting upward pressure on prices. The FHFA economists estimate that between the second quarter of 2022 and the same period in 2024 higher rates pushed down house prices by 5.6%, but the resulting lock-in pushed them up by 7%. The net effect was to make homes more expensive.
What to do? Making mortgages assumable (allowing buyers to take over a seller’s loan) or portable (allowing homeowners to transfer their mortgage to a new property) could help. The FHFA says it is evaluating both ideas. But such reforms would largely benefit future borrowers, doing little for those already locked in. For them, the only real solution is a substantial fall in mortgage rates. And, for now, that looks a long way off. (The Economist)
The Weak Points in AI Bonds Are Becoming Clearer Traders are revealing where they see the biggest pitfalls in debt issuance related to artificial intelligence
The market seems increasingly focused on risks to some of the players in the race for AI computing power. It isn’t quite as worried about the data centers that will host that computing.
While stocks are about upside, bonds are about what can go wrong. So when the bond market takes different views on different parts of the AI financing market, it can be a good guide to what investors view as the most salient risks.
“You should not view the AI trade in credit as one big trade,” says Lotfi Karoui, multiasset strategist at Pimco. “There is going to be differentiation.”
This came clearly into focus this past week when bonds issued by CoreWeave, which sells computing power to AI companies and others, sold off sharply. (…)
In effect, investors have focused on CoreWeave’s only corporate unsecured bond. But, as Goldman Sachs explains,
CoreWeave’s primary funding method for scaling AI infrastructure is via DDTLs (delayed draw term loans) – these are asset-backed project financings rather than traditional corporate debt.
Rather than underwriting CoreWeave’s standalone credit profile, lenders are primarily underwriting the combination of a specific long-term customer contract, the underlying GPU infrastructure, and the cash flows generated by that deployment. These facilities are typically secured by the associated GPUs, servers, contract revenues, and related assets, allowing lenders to leverage the credit quality of the end customer in addition to the collateral value of the infrastructure itself.
The structures have historically been largely self-amortizing, with debt maturities designed to align with contract terms so that customer payments repay principal over the life of the agreement. As a result, by the time the contract expires, the associated debt has often been substantially or fully repaid, leaving CoreWeave to capture any residual value from the GPU fleet or potential re-leasing opportunities without the burden of ongoing interest expense.
More recent transactions, particularly DDTL 5.5, represent an evolution of the model, as lenders appear increasingly active in underwriting not only contracted cash flows but also the future re-contracting potential and residual value of NVIDIA GPU infrastructure, effectively extending financing beyond the duration of the initial customer agreement.
From a risk perspective, DDTLs shift much of the financing risk from CoreWeave’s corporate balance sheet to specific customer contracts and infrastructure assets, reducing refinancing risk and limiting the amount of equity required to fund growth. The key investor debate is increasingly around residual value risk: whether NVIDIA GPU clusters will remain highly utilized and capable of being re-contracted after the initial customer agreement expires.
Back to the WSJ article:
The bonds issued to build data centers that CoreWeave will be leasing actually trade at narrower yield spreads over Treasurys than CoreWeave’s own bonds. (…)
CoreWeave bonds due in 2030 now yield close to 15%, according to MarketAxess data. A bond issued last year to finance Applied Digital’s build-out of a North Dakota data center leased to CoreWeave, however, currently yields around 8%. (…)
Effectively, investors are betting that overall demand for data infrastructure means the data center will be of use to someone, even if the current tenant runs into trouble. (…)
Earlier this year, CoreWeave reworked a lease for an Applied Digital center. The lessee became a separate entity, with its own investment-grade credit rating, which is backed by a CoreWeave computing contract with Meta Platforms. Spreads on those bonds fell sharply when that happened.
(…) if data centers become harder to build in the future, there might also be more value in one that does get done.
“With a data-center bond, you’re underwriting assets, and our view has been that when it gets completed, that can be a meaningful catalyst for spreads to compress,” says Andrew Keches, who is tech, media and telecom credit analyst and head of high-yield research at Barclays.
The bigger picture is that as the market for data center financing matures and a wider variety of bonds and business models come to market, investors might be differentiating more between credits. (…)
One critical factor is the assumed life of the compute assets. Nvidia’s contribution has been to secure lenders on the residual value of its GPUs, offering a backstop if its chips decline in value beyond a certain threshold.
The concern is not whether CoreWeave and other neoclouds are survivors but, crucially, whether a particular data center can be built and financed at reasonable costs, and on time.
The recent Microsoft exit from Nscale’s Loughton data center project in the UK is a real-world case study for why the “protection” of these massive compute contracts is conditional. It exposes the ultimate kryptonite for a neocloud operator: The Delivery Failure Clause.
Hyperscale contracts are structured as “take-or-pay” agreements, meaning Microsoft must pay for the capacity even if they don’t use it—but only if the site is operational and energized.
Microsoft pulled out of its agreement to anchor a 50 MW site because local grid officials stated that power would not be fully supplied until the 2030s.
Nscale is now hunting for alternative expansion sites across the UK that already possess unutilized, localized power hookups, hoping to re-allocate the 23,040 Nvidia GB300 chips that were originally destined for Loughton
Just a day after the Microsoft UK setback, we learned that Anthropic is in talks with Nscale for a second $45 billion computing deal involving a second facility at Nscale’s massive Monarch campus in West Virginia.