The Rhyme: AI's First Credit Crisis & Its 2015 Echo

"History doesn't repeat… but it rhymes." — Mark Twain

◉ THE PRESENT

CoreWeave's unsecured bonds fell below 80 cents on the dollar this week, their yields blowing past 15 percent, and the credit default swaps on the company's debt nearly doubled to 643 basis points in a matter of days. The neocloud provider has taken on more than $51 billion in debt and obligations to buy Nvidia GPUs and install them in data centers, and now the market is asking whether any of that money is coming back. The contagion has already spread: chip stocks have fallen for eight straight sessions, the Philadelphia Semiconductor Index has lost roughly 20 percent from its 52-week high, and the top twenty AI and semiconductor names worldwide have shed more than $1 trillion in market value this week alone. Elon Musk posted one word in response to the bond chart — "Yikes" — and Chamath Palihapitiya warned of a "private credit bubble" forming beneath the entire AI buildout.

The last time a distressed-debt blowup in a single leveraged sector sent fear rippling through a market that was otherwise doing fine, it was December 2015. The commodity wasn't GPUs. It was oil.

◉ THE ECHO — AUGUST 15, 2019

The Fund That Couldn't Give Money Back.

The email went out after the market closed on a Thursday evening. It came from Third Avenue Management, a New York fund shop founded three decades earlier by Martin Whitman, one of the most respected distressed-debt investors alive. The message was short and brutal: the Third Avenue Focused Credit Fund, which held $788 million in high-yield bonds, had frozen all redemptions effective immediately. The board had approved a plan of liquidation. If you had money in the fund, you could not get it out. The doors were locked.

It was the largest mutual fund failure since the 2008 financial crisis, and it caught Wall Street completely off guard. Not because junk bonds had been quietly cracking for months — they had — but because this was an open-end mutual fund, the kind of vehicle that retail investors counted on to be liquid every single day. A bond fund that refuses to give you your cash back is supposed to be unthinkable. Third Avenue made it thinkable. The S&P 500 closed that night at 2,052, about three percent below its May high of 2,130. Nobody panicked yet.

The problem under the surface was enormous. American shale companies had borrowed more than $200 billion in high-yield bonds over the previous five years to drill wells, buy acreage, and build out infrastructure across the Permian Basin, the Eagle Ford, and the Bakken. They had done it while oil was above $100 a barrel, and the math looked bulletproof. But by December 2015, WTI crude had fallen below $40 and was still dropping. The wells were still pumping. The debt was still compounding. S&P warned that 50 percent of all energy junk bonds were "distressed," showing elevated risk of default. The high-yield OAS spread had widened to nearly 700 basis points by year-end, the widest since the financial crisis, and the energy sector alone was yielding close to 14 percent.

Six days after Third Avenue froze its fund, the Federal Reserve raised interest rates for the first time in nine years, lifting the funds rate from near zero to 0.25-0.50 percent on December 16th. Chair Yellen called it a vote of confidence in the economy. The bond market disagreed. High-yield spreads kept blowing out. Oil kept falling — past $35, past $30, all the way to $26.68 on January 20th. The S&P 500 followed. It slid from 2,052 on the night of the Third Avenue freeze to 1,829 by February 11th, a drop of nearly eleven percent in two months. The energy default rate eventually hit 15 percent. The E&P subsector's default rate reached 29 percent.

But here is the part that mattered most: the credit crisis in energy did not become a credit crisis in everything. The contagion reached the borders of the high-yield market and stopped. Non-energy spreads widened, but they widened from tight, not from normal. The economy didn't roll over. The Fed paused after a single hike. And by the middle of 2016 the S&P had clawed back every point it lost, finishing the year up about nine and a half percent.

◉ THE RHYME — WHAT'S IDENTICAL

A single overleveraged sector cracks in the credit market first, while equities near all-time highs still pretend nothing is wrong. Credit breaks first. It broke first in 2015. It is breaking first now.

◉ THE DIVERGENCE — WHAT'S DIFFERENT THIS TIME

The rate level is radically higher. In December 2015 the Fed was lifting off from zero. Today it sits at 3.75-4.00 percent with a 10-year yield above 5 percent. That means the cost of servicing all that AI infrastructure debt is multiples of what shale companies faced. CoreWeave's bonds were issued into a rate environment that was already expensive. The shale drillers at least had the excuse of issuing when rates were near zero.
The collateral is different, and that matters. Shale companies owned wells, pipelines, and mineral rights — physical assets with salvage value even in a bust. CoreWeave's collateral is Nvidia GPUs that depreciate every time a new chip generation ships. If GPU resale values collapse the way rig counts did in 2016, the recovery rate on those bonds could be far worse than what energy creditors eventually got back.
Customer concentration is a new risk. CoreWeave derives a large share of its revenue from a small number of AI labs. When OpenAI reportedly told investors its annualized revenue hit $50 billion but implied a slower growth trajectory, CoreWeave felt it immediately. Shale companies at least sold into a global commodity market with millions of buyers. CoreWeave's revenue depends on a handful of companies that are themselves burning cash.
The equity market is broader than it was in 2015. The Dow is actually up on the week as money rotates into banks, healthcare, and consumer staples. In late 2015 the market was narrower and the rotation options were fewer. That breadth could act as a cushion — or it could mask the severity of the tech credit problem until it's too late.

◉ THE RECKONING — WHAT HAPPENS NEXT

After Third Avenue froze its fund on December 10th, the immediate reaction was muted. The S&P 500 actually rose the next day. People assumed it was an isolated problem — one eccentric fund run by an aging legend who'd taken too much risk. The market wanted to believe it was contained. It was wrong, but only partially.

What happened over the following nine weeks was a slow bleed, not a crash. The Fed hiked on December 16th. Oil kept sliding through January, dropping from $37 to $26. High-yield spreads widened from 695 basis points at year-end to 839 by February 11th. The S&P lost eleven percent, bottoming at 1,829. Trading desks in New York and London started marking down anything connected to energy, then anything connected to commodities, then anything that smelled like credit risk. Chesapeake Energy's stock dropped below $3. Linn Energy filed for bankruptcy. The words "systemic risk" started appearing in research notes from banks that had underwritten the very bonds that were now blowing up.

But the people who kept their heads made money. The smart trade in early 2016 was not to sell everything. It was to figure out whether the energy credit crisis would jump the firewall into the broader economy. It didn't. Non-energy corporate earnings were fine. The labor market was solid. The Fed paused after one hike. By mid-February, the selling exhausted itself. By June, the S&P had recovered everything. By December it finished up about nine and a half percent for the year.

The pattern laid out a simple two-question framework. First: is the credit stress in the leveraged sector spreading to investment-grade bonds and the banking system? In 2015, the answer was no — and the market recovered. Second: is the Fed going to keep tightening into the blowup, or pause? Yellen paused after one hike. If the current Fed pauses after its September hike, the playbook says this pain has a shelf life. If it hikes again in December, the shelf life gets a lot longer.

Right now, CoreWeave's CCC-tier distress is screaming at 643 basis points while the broad high-yield market sits at a sleepy 316. That gap — the distressed tier cracking while investment-grade barely flinches — is the exact signature of December 2015. The question is whether it stays contained there.

The edge: In 2015, the credit crack in one leveraged sector dragged the S&P down eleven percent over two months before the market figured out the contagion had walls. Broad high-yield spreads at 316 basis points today are telling you the contagion hasn't jumped yet. Watch CCC spreads and watch the next Fed decision. If CCC spreads widen past 1,400 basis points and the Fed hikes again anyway, this stops being 2015 and starts being something worse. If spreads stabilize and the Fed pauses, the dip has a floor.

◉ TOMORROW’S WATCH

Netflix reports earnings Tuesday after the bell, and Q3 earnings season accelerates into the week with dozens of S&P 500 names. But the real signal is whether AI infrastructure names stabilize or whether the credit stress spreads to other debt-heavy tech buildout companies like Equinix and Digital Realty. If neocloud and data center bonds start widening in sympathy with CoreWeave, this starts to look less like Third Avenue in December 2015 and more like Bear Stearns' hedge funds in June 2007 — the canary that sang twice before the mine collapsed.