Hello, this is Dean Kurnut, and welcome to the Alpha Exchange, where we explore topics in financial markets associated with managing risk, generating return, and the deployment of capital and the alternative investment industry. Greetings and happy holidays, Alpha Exchangeers. I hope you've had a chance to unwind and share some quality time with your families, as I did with mine. Recently, we had to say goodbye to our dear family Pat Griffin. Dog truly can be a man's best friend, and the Griffin I were side by side for 11 years. I'll miss him, and it was a great run. On the positive side, I've been lucky enough to have all three of my children home for the break, and increasingly rare occurrence. They say that you'll have spent 90% of all your time with a child up until the age of 19. Most of you are younger than I am, so some advice from this old timer. Be present, I say, and enjoy these times. If you can, also take 10 minutes to check out a recent TEDx talk from my dear friend, Allegra Cohen, on a concept that she calls "microjoy." You'll find yourself re-centered. On the market's front, we are as widely expected ending the year on a quiet note. I asked Chatchee BT to calculate the percent moves of the S&P over the last seven trading days of the year. I did have to tell it not to write any Python code, but it got the job done anyway. We all recall the doubt 2.7% Christmas Eve and up 5% day after Christmas Caper in 2018. But almost always, there's nothing going on towards the back end of the year. Since 2010, the average of the absolute value of the daily moves of the last seven trading days is just 56 basis points. Eliminate 2018 and you're at just 47 basis points. That is skinny. Zoom out, and it's not just a holiday inspired decline involved. One month realized on the S&P is 8.8. Even considering a three month window, which captures the three week 5% S&P drawdown that began as October ended, realize VAL is just 12.4. And it's not only equity VAL that suffers from George Costanza-like shrinkage. The risk-free asset class, the U.S. government bond market, is actually living up to its name. The daily moves have narrowed dramatically in the TLT, where one and two month realized are 6.7 and 7.8 respectively. But I hope to present to you over the next 35 minutes is some version of closing thoughts. As I do so, I want to look back on 2025 with an eye towards pointing out its unique characteristics from a market risk perspective. I probably say this too frequently, but these are fascinating times with a lot at stake. Embracing the notion that absolutely anything can happen in markets is a good starting point for risk management. But for now, the aforementioned absence of meaningful daily moves on the important macro assets is imposing downward pressure on option prices. The S&P and TLT are recently at least accident-free, and that's all the market cares about in pricing options. Count May is pretty excited that we'll start 2026 with market insurance that feels reasonably priced. It's a nice offset to the car, health, and homeowners' insurance affordability mess. Most folks are navigating. Let's start this exercise in closing thoughts by highlighting what I consider to be 2025's three most interesting days from a vol and risk perspective. And that must begin with the chaos that ensued post-april second liberation day. Previously, there were only two other instances when the VIX surpassed 50, the GFC, and the COVID crash, will exclude 1987. The tariff tantrum is the third. On April 7, the Monday after the S&P experienced a two-day 10.5% meltdown, we were all forced to ride the VIX roller coaster hands-free, I might add. April 7 was truly a wild day in the market, as it was caught in the crosshairs of unreliable information that hit the tape on the severity of tariffs. On 9.45am, we saw the VIX fall from 54-38 in just 30 minutes, and then rise back to 53 over the following 14 minutes, only to fall to 44 over the next 30 minutes. There's only one word that comes to mind, absurd. Markets simply cannot absorb that level of risk for too long, things break. And it became clear to me that the VIX gelatinies would require a full-blown nevermind from Trump on April 9 to restore order, he obeyed orders from the market. This was a good lesson in thinking about trading the policy response. That is, anticipating how market prices would force the administration's hand, and how those prices would react once the retreat occurred. I wrote a fair amount about this, using the Bill Gross GFC strategy to "shake hands with the government and buy what they're buying." Only this was to sell what they're selling, in this case the VIX. You knew that the VIX could not be allowed to remain in the 50s because the GFC and COVID Presidents told us that it would be 50 on the way to 80. Hopefully, Don Jr. got his VXX short off at the April 8 peak, just kidding, maybe. The next most interesting day, for me at least, was the September 10th surge in the share price of Oracle. Recall this was the earnings date for the company, and it came with a forecast of a tremendous revenue increase along with a tie-up with OpenAI. The stock price jacked higher by 36%, Larry Ellison briefly overtook Elon Musk as the world's richest man. But the stock is down 40% since, and, most interestingly, the five-year CDS spread has risen 100 basis points from 45 to 145 since September 10th. Does the market's judgment of Oracle's credit have information content, and is it some shorthand for whether AI financing aspirations have become too ambitious? Zooming out, Oracle is up 19% on the year, and its CDS is more than 100 basis points wider. Its two-month implied vol has nearly doubled from 25 to 45. The equity is being treated more like an option than a stock. Oracle may be unique in just how aggressive its borrowing and capex plans are relative to its market cap. But if we look at a five-year CDS for a basket of Google Amazon, Apple, Microsoft, and Broadcom, that's up on average from 24 to 36 basis points this year. The five-year corporate IG, by contrast, is flat on the year at 50 basis points. Too early to derive any strong conclusions, but put this on your dashboard of metrics to watch for market warning signs. The last most interesting day is the October 21st Gold Meltdown. As its cousin Silver delivers epic vol on moves both higher and lower, let's recall the dramatic spiral up and one day unwind that the GLD experienced in late October. One of my little sayings is that "risk on" and "risk off" are curious cousins. It's a nod to the way in which profits from a trade invariably draw attention and lower-end fresh capital eroding the margin of safety in the process. When the success of a risk on episode is significant enough, it paves the way for a sharp unwind. In the limit, like a GME, it's a certainty that it will occur. While timing is never easy, it wasn't difficult to see the giant one-day unwind of very extended positioning in the GLD coming. The GLD had rallied 10.3% over just seven trading days from October 9th to October 20th. That's just way too much for a 15-vall asset and the FOMO nature of gold led to its chase. All of the classic signs were there. A spike in implied vol, the GVZ, the gold vix, reached 32.8. An inverted vol term structure, an inverted call skew, and massive call volume. I shared the following on Twitter on October 8th, quote, "The strength of the recent gains in gold paradoxically due to things at once. First, the rising price is the advertisement compelling folks to buy. There's no Graham and Dodd valuation framework to do. As Soros said, "When I see a bubble forming, I rush into buy, further adding fuel to the fire. The rising price is a source of new demand. The rate of change of upside moves is accelerating. Since 2023, there are 14 days when the GLD has moved up 2% or more. Eight of them have occurred since April. Second, as the, quote, "sky is the limit narrative" builds, implied vol rises, reflecting the market's understanding that the risks are becoming more two-way. That is, for folks wanting to play the upside, using call options may be preferable as the recent strong gains could quickly reverse. The call option permits you the right to walk away if wrong. Two month implied vol on the GLD has gone from 15 to 18 over the last two months, even as realized vol has fallen from 15 to 13. This is not about how the options are carrying. It's simply about one-way demand for options. You wind up in a situation where the strength of the risk on creates the vulnerability for the risk off as those investors in early take profits and those in late try to limit losses. It's a sharp unwind that clears out positioning. It may be good for a 3-5% decline over a few days. The option dynamics may accelerate it. If the buyers of all the calls that have traded in the GLD are outright and the sellers are hedging, you might get some feedback as these hedgers need to rebalance their deltas by selling into a falling market. When an asset experiences a stock up, vol up event that is substantial enough, there's really no way for it to unwind except a stock down, vol down, reversal. Thank GMA in 2021, MSTR in 2024, and now even in silver. And that naturally leads to the next part of this review, which is to highlight two main themes in risk, the first of which is that stocks are behaving like options. As the price of many companies rise, the market assigns their option a higher implied volatility. This is completely antithetical to the relationship between the S&P and VIX, which have a consistent correlation of around -80%. Take Google for example, up in astounding 65% this year. At one point in late November, two year, 120% of spot strike call options traded at a vol of 39, up 12 on the year. This is a massive increase. To give you a sense, a two year, 120% strike call at 39 vol costs 64% more than it does at 27 vol, the level we saw at the start of the year. This is the market's way of assigning a considerably wider degree of potential outcomes to the stock. The relationship between Google VAL and Spot actually isn't atypical these days. It's just a good example of the VAL characteristics common to today's high flyers. Stock returns and implied VAL are very often positively correlated these days. It's a reflection of a winner take all market in which speculation and taking upside convex bets has been rewarded. The other side of this stock up, vol up dynamic is the seller of VAL. Hedging upside calls used to be easier. The stock would rise, typically gently, and implied VAL would fall in the process. Now, upside price shocks underpin volatility by a considerably greater degree than in the past. Over the second half of 2025, Google is realizing 32.6 vol on updates and just 20.7 vol on down days. The seller of upside calls is having to contend with this new kind of return distribution and account for hit in his or her hedging protocol. Consider two year implied VAL on both Google and Nvidia at 36 and 46 respectively. That's $8.3 trillion of combined market cap and both stocks have double A2 ratings from Moody's with tons of cash and free cash flow. At risk, often a driver of volatility and inequity is not a thing that comes to mind for these money printing enterprises. Nvidia's market cap is 40 times that of GM and Ford, yet their two year implied VALs are around 32. The car makers are all rated triple B, the bottom rung of investment grade. For these companies, unlike the tech mega caps, debt can be an issue. Why the lofty, long-dated implied VALs and option prices for Google and Nvidia, even as their stock prices are doing so well and their credit ratings are gold plated? My take is that the market cap of the tech behemoths is so large and has increased so quickly that the options market is struggling to provide insurance against loss on them. The option price may clear at a high level because there's not enough natural capital to bear the risk of loss. All else equal a higher premium is needed to bring sellers to the table. There's almost an option market equivalent of what's happening in the broader insurance industry. Premiums are higher and it's not necessarily simply the result of risks that are materializing today. It's more about compensation for future uncertainties and related a shortage of risk-bearing capital. The tech stocks aside if there were an annual stock up VAL up award and I will argue that perhaps there should be. It has to go to silver in 2025. Let's take a look. The SLV is up 150% on the year. It's 2 month implied VALs started the year at 25, it's ending it in the 60s. The correlation between price and 2 month implied VAL is running consistently north of 90%. Realized VAL in updates is 32.4 versus just 29.4 on down days. And since November, Realized VAL in updates is 53.6 versus just 26.9 half of it on down days. There are 8 moves of greater than 4% up in 2025 and just 2% 4% down moves. There's massive call volume, far outstripping put volume. There's a highly inverted VAL term structure, that is the market prices short dated implied VAL higher than further out VAL. And lastly, there is a deeply inverted call skew, that is the market is paying a 17 VAL premium for a 1 month 10 delta call versus a 1 month 10 delta put. As silver spiked, there are lots of takes on whether to be in the mean reversion or momentum camp in positioning long or short. With respect to the latter, as I've said, Soros told us he's going to rush in to buy a bubble when he sees it and he's going to add fuel to the fire. You've got to be careful when you see stock up VAL up to this extent. My framework suggests that when a stock up VAL up event is this protracted, it's more likely than not that lower prices and lower VAL will eventually emerge. But here's the thing about a market dislocation. As you think about capitalizing on it, you've got to respect the forces that created it in the first place. Market prices don't stray far from fundamental value without very good reason. And those same forces could very likely push it even further away. Think the 29 and a half year versus 30 year US treasury bond basis in 1998 due to long-term capital management's leverage position gone wrong. Think about the 2008 Volkswagen squeeze, the 2009 implosion of the dividend swap market, the 2010 blow up and long dated S&P variants, the 2020 crude meltdown, the previously mentioned 2021 spiral in GME, the 2022 short squeeze in nickel, the UK guilt crisis in 2022. In each of these, the VAL and correlation assumptions that investors, credit risk officers and exchanges had assumed proved remarkably wrong, suddenly the existing trades underwritten at much lower VALs and correlations became much larger in terms of value at risk. The process of finding the right sizing can amplify an already unstable situation. All of this is to say, be careful. If you see a trade that looks compelling and as a result of a large dislocation, commit only a small amount of capital to it, whatever your bias, the massively expanded VAL makes a given dollar at risk more uncertain, be smaller, or find an option structure that limits your losses in the scenario in which the trade moves against you. And speaking of dislocations, I'm a big fan of the Big Short book and movie, and I'm firmly in the camp of Big Short versus margin call, which I also enjoyed by the way. In the Big Short, Mark Bound, played by Steve Carell and representing Steve Isman, asks the exact two-part question which gets to the heart of how to think about systemic risk. He asks, is there a housing bubble and if there is, how exposed are the banks? You need two ingredients for a real spillover event, one, a large mispricing, and two, leverage. When you get these in combination to a substantial degree, disaster awaits. Ultimately, the market is forced to confront the mispricing, in this case, of mortgage credit risk and correlation. When that process imposes losses on mark-to-market sensitive investors, a reflexive risk on wine can materialize. There are plenty of instances when a repricing does not lead to a wide-scale spillover. The internet bubble comes to mind, although 2002 was quite a credit widening event. The unwind of the Euro Swiss peg in 2015 is another. There were some smaller hedge funds that went under, but it didn't become systemic. What you need is a significant combination of Mark Bound's two-part question, a big mispricing and wide-scale exposure to it through leveraged institutions that are marked to market sensitive. Then to hit the home run that John Paulson did, you have to perfect the structuring and timing of a convex trade. What was so entirely unique about the pre-GFC era was that a centerpiece of the bubble inflating was massively downward pressure on risk premiums like the VIX and credit spreads. While these measures will start 2026 at pretty low levels, they ended 2006 much lower in a system in which a tidal wave of leverage was set to come undone. Two-year implied vol on the S&P 500 hit 13 in late 2006, it's now 19. The straddle cost 50% more using 19 vol versus 13 vol. With that little detour, let's return to our main themes on risk. As discussed, stocks are behaving like options and the market is reacting to the consistency of stock up vol up. These aren't just meme or dgen stocks. These are market behemoths like Google and Nvidia. Over time as the tech trade has gotten larger and larger, so too has its weight in the S&P 500. It's no secret that the S&P is typically concentrated with high-volved tech names. This ain't your father's index and you gotta know what you own. Passive investing can lead to some strange outcomes. In 2000, depending on how one measures it, the PE of the S&P reached between 30 and 40. Today it's quite elevated but not at that extreme. The peak of the tech bubble will forever be a very tough valuation come. What is extreme today is the concentration of the S&P with very volatile stocks. The index that attracts so much capital and is a benchmark that no one could ignore is top heavy like never before with stocks all pursuing the same AI riches. Here are some stats. First, the top four stocks are 27% of the index. The top eight have combined market cap of 22.4 trillion and are 40% of the index. These top eight have a two year weighted average implied vol of 37%. The next eight have a combined market cap of just 6 trillion that's 10% of the S&P and get you to half of the overall market cap of the index. 16 stocks are half the S&P. These next eight stocks however have much lower implied vol than the first eight. The weighted two year vol for the second eight is just 27. In words, we can describe the S&P as quote in index that is highly tracked, highly concentrated with highly volatile, highly valued tech stocks that have proven remarkably uncorrelated to each other. And that's the second theme I want to highlight as I have all year. The low level of correlation among stocks in the risk implications of this new phenomenon. First, let's establish that markets generally price what they see and experience. A scatter plot of 30 stocks will show a very consistent cross-sectional relationship between realized and implied vol. The same goes for correlation. As 2025 ends, one year implied correlation on the S&P is basically a match for one year realized correlation. Just as the marginal price setters for vol are beholden to the feedback between realized and implied, so too is the Matthew dispersion crowd reliant on how correlation carries. Low realized correlation justifies low implied correlation. But to be clear, one year implied correlation on the S&P at 21% is really, really, really low. There's no equivalent except last month, last quarter and last year. This isn't entirely new and that is part of what I think makes it risky. When a clearing price endures no matter how high or lower it appears to be, it makes its way into how we consume risk. Because the dispersion trade buying single stock vol and financing most of the premium by selling index vol is working, even at low levels of implied correlation, more of it will be done. The profits it generates gets recycled back into the same trade that spit them out in the first place. There are a couple of things to think about here. First, consider the relationship between realized correlation and realized vol for the S&P. A chart I posted on Twitter shows that for a given level of realized vol, realized correlation used to be considerably higher than it is today. There are two ways to interpret this. First, single stock vol is doing more of the heavy lifting today to generate the overall index vol level. The second way to look at this is that given these very high single name vols that come from a top heavy tech concentrated S&P, a tremendous amount of diversification is occurring to keep the index vol where it is. The incredibly low level of realized correlation is a significant vol suppressant. Will it continue? I'm not so sure. The second chart I posted on Twitter illustrates a similar point, but does it through implied vol. I created an index of the simple average one year implied vol for Nvidia, Google, Microsoft, Apple, Amazon, Meta, Broadcom and Tesla. One of the time series shown is the ratio of that to one year S&P implied vol. The second series is one year S&P implied correlation inverted, not surprisingly these move closely in tandem. So the question might logically be, is single stock vol too high or is index vol too low? That's actually not the question. It doesn't really matter. It's the relative price that matters and I strongly believe it's too low. That is to say that single stock vol is too high relative to index vol. Or as I prefer to say it, index vol is too low relative to single stock vol. Starting at this way is consistent with the view that the repricing higher of implied correlation is more likely to occur in tandem with a higher overall implied vol environment. If the global economy slows for example, commitment to the capex cycle could get tested, causing a broad and correlated retreat in share prices. Let's explore how the relationship between single stock and index vol reprices. First, a shorthand for implied correlation. What we do is we take index vol, we divide it by the weighted average level of single stock vol, and we square that ratio using 37.4 for the average of the big eight in the index I created and 17.4 for one year S&P implied vol. That squared ratio is 21.5%. That's right where Bloomberg has one year implied correlation on the S&P. Let's flip the formula around and ask what happens to index vol as we keep single stock vol the same, but move implied correlation to 35. The 13.3 point bump in correlation adds 4.7 vols to S&P vol. That is a very large move in one year implied vol. And to be clear, 35 is still low for implied correlation historically. There are two primary channels for this repricing. First a macro shock like the April tariff tantrum. As I shared in a chart on Twitter, implied correlation spiked during that episode. It was of course self imposed by Trump and thus relatively easy to undo via a just kidding on April 9. But there are many channels for macro shocks, monetary policy, geopolitics, a slowing economy, a glitch in the shadow banking system to name a few. The second channel is less about global macro and more about the AI ecosystem and how intertwined these companies are. They are all chasing the same trade in AI, spending fabulous sums making lofty assumptions and increasingly raising lots of debt to do so. The CapEx spending itself is keeping this going. As long as this CapEx cycle is robust, the market caps are supported which in turn supports the CapEx. It does remind me of how both the mortgage credit and LBO funding machinery kept the leverage cycle going 20 years ago. And there's a more technical vantage point from which to contemplate the repricing of single stock vol to index vol. This concerns the prominence of stock up vol up in today's market even in the mega caps. It's the feature of this market that is most like the internet bubble. As mentioned, Google is up 65% year-to-date and its two-year implied vol was recently as much as 12 higher from the start of the year. As suggested earlier, the stocks themselves are options as they rise the market pays more and more for the lottery ticket. Here's the analogy back to the.com era. From the peak in March 2000 to the end of 2004, the triple queue fell by 65%. The Nasdaq VIX, however, fell from 50 to 19 in the process. In the current market, if the AI trade loses some of its shine, you could see the stocks driving it, not just fall and price but fall and implied vol as well. What goes up must come down kind of thing as the optionality of the trade declines. This process would also lead to implied correlation rising, perhaps by a fair amount. All of this is to say that we are at the lowest level of S&P one year implied correlation we've seen, and there are multiple pathways to it rising from here. It's a real vulnerability for the market as it would make the S&P considerably more volatile than it currently is, and we know that the same conditions that make stocks more volatile make them more correlated as well. It's a double whammy. We can look back on 2026 as a year of highs and lows. The S&P is up 17% on the year, even as it experienced a 19% drawdown along the way. Don't call it a comeback as LL Cool J told us. One month realized that the index was as high as 51 and as low as 6. For correlation, the peak was 67 and the low was, wait for a dean warmer, 0.0. Correlation has no grade point average. Note that we are ending the year with one month realized correlation of just 8. You can't blame the market for pricing one month implied at just 11. I argue that the number is eventually going higher because stocks are eventually going to start moving more closely together. The market, the investing public and the economy at large, are overexposed to the AI trade. So too are the AI stocks overexposed. I certainly can't predict when or if something will go wrong, but these ultra low correlations are the equivalent of driving without a spare. As I close this discussion, I want to thank you for being a listener. I was able to drop 26 podcasts this year with extremely high quality guests. These are hedge fund founders and asset management CIOs, Fintech founders, heads of strategy efforts and leaders of independent research firms. The conversations are not about predicting the next move, but in seeking to add value to the process of portfolio construction and risk management. I've also, including this one, dropped 15 podcasts in which I shared my own thoughts on risk. I'm looking forward to a year of expansion for the Alphax change in 2026. I've got some creative new ideas for delivering content and looked forward to bringing them your way. Until next time, have a relaxing holiday and rest up for what promises to be a critical year in markets. Be well. You've been listening to the Alphax change. If you've enjoyed the show, please do tell a friend. And before we leave, I wanted to invite you to drop us some feedback. As we aim to utilize these conversations to contribute to the investment community's understanding of risk, your input is valuable and provides direction on where we should focus. Please email us at
[email protected]. Thanks again and catch you next time.
PodcastTranscript AI