Market Overview | October 13, 2026

Summary

• Shanghao Jin sees the source of the liquidity shock starting to heal, but expects the downstream recovery to lag. Rates and rate volatility were the source; drivers included the Fed’s stance that it would not manage the long end, the link between oil and rates, MBS convexity hedging and a swaption gamma squeeze. Friday’s data he reviewed showed convexity hedging easing and the swaption squeeze nearing its end, though he had not checked the latest data that day. Long-end yields have not fallen, but rate vol has.
• The rate spike was driven by the Fed’s stance on the long end and a once-stronger link between rates and oil; the Brent-rates correlation has since eased. Rates rose from about 4.5%–4.6% to 5.3% in a very short time, a move of more than half a percentage point, and nobody wanted to catch the falling knife. Higher rates slow mortgage prepayments and extend MBS duration, prompting pension funds, insurers and others to hedge in stages, potentially by selling bonds. Short swaption positions were forced to close and buy back options as both rates and volatility rose. Rates moving above 5.5% could trigger another round of hedging, but a squeeze-driven move higher in yields would be harder to sustain.
• Energy remains a tail risk, with little margin for supply-side error. Oil exports have recently fallen but are broadly back to prewar levels; natural gas exports have recovered from near zero to roughly one-quarter to one-third of prior levels—he recalls about 27%. Winter heating demand in Europe, concerns about European gas supplies, US concerns about diesel and crack spreads could all feed into perceived inflation through freight, airfares and heating costs. Details of diesel supply releases by Europe and France have not yet been announced.
• The downstream recovery is uneven: volatility in G3 currencies such as the euro and yen has eased as dollar rate vol fell, while carry-trade volatility in EM FX ex-RMB has not fully subsided. Sovereign credit spreads in France, Germany and Spain have yet to recover. CDS investors focus on spreads, while many buyers of high-yield or IG bonds focus on absolute yield; when rates are volatile, they pull back, widening spreads. US CCC has been hit hardest, while IG has held up relatively better. Hyperscalers and CSPs have issued about $230B–$250B in bonds this year, with issuance largely finished since August and September.
• The equity outlook is mixed, but the market is resilient and investors risk being left behind. Fundamental long/short funds are underweight; if indexes rise, they may chase. Higher skew and stronger call than put demand suggest investors are reluctant to take much macro risk but fear underperforming the S&P. The setup resembles March, when the Fed eased bank SLR requirements and liquidity was better; today liquidity is poorer and more tightening is priced in. Valuations are around 19x P/E, and he expects multiples could return to tariff-war lows after earnings, but multiple compression does not mean stocks cannot rise.
• AI financing has a risk mismatch: a compute center might recoup its investment in about 10 months by selling tokens, or in about 2 to 2.5 years by selling compute directly, while creditors face risks closer to equity risk if AI demand is disproven. In an Apollo project-finance example, 75% of the funding is debt and 25% equity. If AI demand does not materialize, the chips’ use and the debt’s protection are both in question. His view: if you do not believe in AI, bond yields do not justify taking this risk; if you do, consider equities. AI still needs substantial funding.
• 2027 will test K-shaped funding demand, with financing needs and refinancing risk converging from 2028; the key is whether CSP cash flow improves. About $140B of debt matures in 2026, while 2027 financing needs are about $700B versus roughly $400B in 2026; those needs do not overlap with the refinancing wall until 2028. Microsoft’s earlier guidance suggested it could reach positive free cash flow next year, including financing, but the market has not fully priced that in. Shanghao Jin is relatively constructive on cloud demand, while recognizing that market skepticism is high and volatility will create opportunities.

Deep dive

1. The source is healing: the dam in rates and rate volatility is holding

• Picking up where he left off, Shanghao Jin likened the liquidity shock to a breached dam: the flood spread from one market to the next, hitting FX, credit and even sovereign risk. Plug the source, though, and downstream markets still take time to settle.
• He thinks the upstream rates market is close to stabilizing. The Friday flow data he reviewed showed convexity hedging easing; rates had moved little the previous week, while rate vol had come down. He had not checked the latest data that day. Long-end yields have not fallen, but lower volatility matters more to markets; rising vol also hurt markets in February and March this year.
• He is bullish on rates—meaning he expects yields to fall and volatility to ease—while oil remains a tail risk. Another gamma squeeze looks unlikely. If rates rise above 5.5%, a new wave of hedging may emerge, but a squeeze-driven rise in yields would be difficult to sustain.

2. What drove the rate-vol spike: convexity hedging meets a swaption gamma squeeze

• Factors driving rate volatility included the Fed’s clear stance that it would not manage the long end, the link between oil and rates, convexity hedging and a gamma squeeze in rate options. Since the war began, Brent’s correlation with rates had risen steadily, but he said it has now eased. Brent is the better lens than WTI because WTI is also affected by US production, while Middle East developments may show up more clearly in Brent.
• Rates jumped from about 4.5%–4.6% to 5.3% in a very short time, a gain of more than half a percentage point. Nobody wanted to catch the falling knife, but sellers kept coming. As rates rise, mortgage borrowers prepay less, extending MBS duration—for example, from an expected 3 years to 6 years. Pension funds and other real-money investors holding MBS can hedge by paying fixed or selling bonds.
• These investors are not dealer desks trading mechanically off models every day. They tend to hedge in stages after large market moves or once certain thresholds are crossed—for example, if duration exceeds 10 or 15 years, they may sell long-end bonds to bring it back to 10 years. Shanghao Jin thinks the scale of this hedging matters more than whether the Treasury issues or buys back a little more or less long-end debt. Only the Fed can truly control long-end rates, and under the Taylor rule, it is unlikely to focus on that.
• At the same time, short swaption positions came under pressure as both rates and volatility rose. Sellers were forced to close out and buy back options—effectively buying volatility and pushing it higher. He compared the dynamic to a gamma squeeze in equities.

3. Oil is the swing factor: energy supply has little room for error

• Rates currently have no clear anchor and may be partly tied to oil. Oil exports have recently fallen but are broadly back to prewar levels; natural gas exports have recovered from near zero to roughly one-quarter to one-third of prior levels, which he recalls may be about 27%. Supply remains tight, and winter heating demand in Europe could drive gas demand higher.
• The energy market has little room for error. Europe is more concerned about natural gas, while the US is more concerned about diesel. The pass-through to perceived inflation includes crack spreads, vehicle fuel, freight, airfares and heating costs. Over the weekend, markets worried that a hurricane could hit US refineries; Europe and France also mentioned releasing diesel supplies, but details have not been announced. Winter weather remains uncertain.

4. Downstream recovery is uneven: G3 has eased, while EM carry and European sovereign spreads remain under pressure

• The more peripheral and riskier the market, the longer the effects take to fade. Volatility in G3 currencies such as the euro and yen has eased as dollar rate vol declined. The Australian dollar appeared in his chart, but he said the focus was really on the euro and yen. Excluding the RMB, a managed currency, carry-trade volatility in EM FX has not fully subsided.
• The impact is more visible in credit: sovereign spreads in France, Germany and Spain have yet to recover. Investors do not all buy bonds for the same reason. CDS investors focus on spreads; many direct buyers of high-yield or IG bonds focus on absolute yield. When buying Walmart or Starbucks bonds, for example, they ask whether they can earn 5% over a year. If rates are volatile and bond prices and yields keep moving, these real-money investors pull back; fewer buyers mean spreads widen quickly. He sees European government-bond spreads as a relatively large risk, potentially more serious than high yield, though not likely to trigger another liquidity problem on their own.
• US CCC has been hit hardest, while IG has held up better. One reason is that hyperscaler and CSP bond issuance in the US has largely wrapped up since August and September, with about $230B–$250B issued year to date; IG supply should be limited over the next few months. Lower issuance from large cloud providers, stronger earnings expectations and their role in index allocation and market liquidity have also supported the market.

5. Equities are mixed but resilient: low positioning, elevated skew and a risk of being left behind

• Shanghao Jin said he was surprised by the equity market’s resilience over the prior 2 weeks; at the time, investors thought corporate earnings would not be especially bad. Risk appetite weakened and the market de-risked on Monday, with factors including liquidity concerns, Hock Tan’s comment that AI needs more funding and rumors that Vera Rubin shipments could be 5,000 units rather than 10,000.
• Fundamental long/short funds are relatively underweight. He stressed that this is not the same as hedge-fund positioning; prime-broker net margin is roughly where it was a year ago. Goldman’s chart shows low fundamental positioning, leaving funds at risk of underperforming their benchmarks. If indexes do better, they may chase, creating upside risk around earnings.
• Skew has risen over the past few months, with more investors buying calls than puts. Macro uncertainties include the midterm elections, war, rising rates and what Kevin Warsh and the FOMC do next. Investors may not want broad market exposure but still fear underperforming the S&P, so they buy S&P calls. One-month at-the-money implied volatility has barely moved recently, while skew has risen—a sign that fear of being left behind is increasing.
• The setup resembles March this year: low positioning, plenty of macro risks and investors worried about missing a rally. The difference is that the Fed eased bank SLR requirements then, providing more liquidity to the interbank market. Liquidity is worse now, more tightening is priced in and valuations are lower. The market is currently around 19x P/E; he expects multiples could return to tariff-war lows after this earnings season, but multiple compression does not mean stocks cannot rise, as earnings growth can still lift expectations. Overall, his equity view is mixed.

6. AI needs a “bagholder”: creditors are taking equity-like risk

• He reiterated his view from the start of the year: AI financing will need a “bagholder.” A compute center itself could be highly profitable: selling tokens might pay back the investment in about 10 months, while selling compute directly might take about 2 to 2.5 years. But the interest and ROI are mismatched.
• One example is Apollo providing 75% of the debt in project financing, with the other 25% in equity. If AI demand is disproven, the debt has little protection; without inference demand, the chips have no use, and even an Nvidia backstop would not solve the problem. Shanghao Jin’s view: if you do not believe in AI, bond yields are not worth taking this risk for; if you do, you should invest in equities.
• He said Hock Tan and Charlie Munger may also have said that AI will ultimately need capital, perhaps even more than electricity. He agrees.

7. 2027 is a K-shaped funding test, followed by refinancing risk and the question of CSP cash flow

• There are 2 funding streams to watch: rapidly growing AI-related bond issuance, and bonds issued in the low-rate environment of 2020–2022 that are coming due and need to be refinanced. Much of the debt originally due this year and next has been extended to 2028 and 2029; those extensions were largely completed this year.
• About $140B matures in 2026. Debt to be refinanced in 2027 is mostly IG, with a large share in software-company bonds. High yield, leveraged finance and leveraged loans competing with AI data centers for funding are not yet under as much pressure in 2027: total financing needs are about $700B, versus roughly $400B in 2026. At that point, financing needs have not yet collided with the refinancing wall; the two converge in 2028. Shanghao Jin sees 2027 as a K-shaped test: heavy funding needs do not necessarily trigger immediate defaults, but when refinancing is due, borrowers have to find a way to raise the money.
• Heavy funding needs and poor liquidity weigh on valuations. He noted that the market is around 19x P/E and expects multiples could return to tariff-war lows after earnings; but multiple compression does not mean stocks cannot rise, as earnings expectations may keep being revised higher.
• When refinancing pressure arrives in 2028 and 2029, the market will also need to see whether CSP cash flow improves between next year and the year after. Based on cloud providers’ current expectations, they should not still need to raise this much money by then; they should already be generating cash. If cash flow does not improve this year or next, credit markets will run into trouble first—there is no need to wait until 2028. Microsoft had already signaled in an earlier Q3 that a company like it should not be cash-flow negative next year and could reach positive free cash flow, including financing; Google and Meta may be somewhat worse. The market has not fully priced this in.
• He thinks 2027 will be the decision point for whether AI can earn back these investments; if it cannot, problems may follow. His personal view is relatively constructive: cloud-provider demand is not coming only from Anthropic and OpenAI, but also from demand and expansion they see within their own businesses. Anthropic and OpenAI’s ARR is also constrained by compute supply, and he has heard that Google teams are having trouble renting enough cloud compute for training. Because his view differs significantly from market consensus, he expects substantial volatility—but also sees opportunities in it. Equities are mixed in the near term; downstream markets such as credit and FX still need to recover, though he thinks convexity hedging and the swaption gamma squeeze at the source have largely stopped bleeding.