Market Volatility Cools as Amazon Earnings Support the AI Trade

Happy Friday,

Yesterday’s short squeeze, which I was expecting, removed a large part of the fear priced into the market. That does not mean risk has disappeared.

Volatility expectations are cooling, positioning across the major indices is improving and Amazon’s earnings support the broader AI trade. However, I am still trimming positions and becoming more defensive after the rally.

Let’s begin with what my volatility model is currently showing and what it means for SPY.

What My Volatility Model Is Showing

One of the methods I use—and advise people to use—for volatility forecasting is Generalised Autoregressive Conditional Heteroskedasticity modelling, specifically the Glosten-Jagannathan-Runkle version.

The name sounds complicated, but its purpose is kinda simple: it estimates how volatile the market is likely to be based on recent price moves. It also accounts for the fact that volatility tends to cluster and that sell-offs usually increase future volatility more than rallies of the same or a similar size.

The model currently forecasts annualised volatility of 15.00% over one day, 15.07% over one week and 15.33% over one month, compared with the VIX at 17.09%.

This leaves a gap of 1.76 points, meaning the options market is still pricing slightly more movement than the model currently expects, and that additional premium will likely subside.

However, that gap was above five points only a day earlier, so a large part of the additional fear priced during Wednesday’s sell-off was removed during Thursday’s recovery.

For newer traders, this does not mean SPY is guaranteed to move higher. GARCH is not a directional model. It estimates the likely size of future market moves, not whether those moves will be up or down.

A volatility forecast of around 15% translates into a normal daily SPY move of roughly 0.9% to 1.0%.

The main takeaway is that volatility pressure is cooling and the market is no longer pricing the same level of stress. Direction, however, still needs to be assessed separately.

And yes, I use asymmetric variants for confirmation, but I do not want to make this write-up even more complicated.

The Situational Awareness Liquidation

On to the Situational Awareness liquidation everyone is talking about.

Taking out someone who does not have enough liquidity to cover their leverage and manage a downturn is not unusual. It was also widely understood that Leopold was using significant leverage to generate these returns, which would naturally attract attention from sophisticated market participants.

Hedge funds use leverage, but it is generally much more carefully managed. When an inexperienced player appears with excessive exposure, people take notice.

The hike call from Citadel, which I told you I was sceptical about, may have been one of the factors that pushed Situational Awareness over the line. However, getting the fund anywhere near that line is ultimately Leopold’s fault.

Anyway, after the short squeeze I was expecting, I am trimming and trailing positions. I will likely wait for some good news from the administration on the Middle East while keeping an eye on oil.

If Trump wants to ensure the market continues higher, he will need to provide some positive headlines so momentum piles back in and we close higher. Now is generally a very good time to do that, and I will increase exposure if it happens.

Amazon Earnings Support the AI Trade

On to Amazon, one of my core positions—and yesterday it delivered big.

A couple of things I want to share relate to the expected returns on AI infrastructure:

“On average, it takes a little less than three years to break even on that investment.”

“The servers currently have a useful life of at least five to six years, and most of our AI capacity these days is being contracted for at least five-year terms.”

“Our data centres have 30-plus-year useful lives and should support at least five to six generations of server economics.”

“I think we’re still in the relatively early stages of how much demand there’s going to be for AI.”

This is the same message we heard from GOOGL, MSFT and META, with demand exceeding supply through at least 2027.

The binding constraint remains capacity.

The $496 billion of contracted backlog, combined with capacity being sold out two years forward, confirms that demand continues to exceed supply.

Very good earnings overall, in my opinion, and a positive for the broader AI trade—especially with Amazon raising cash capital expenditure. Higher memory costs are expected to add around $20 billion, while available capacity is still falling short of demand.

I also like that the market did not treat the negative free cash flow as a major problem. Amazon is investing to build and supply against a contracted backlog of $496 billion, so the spending has a clear purpose.

Apple, unfortunately, disappointed. However, it did provide another positive signal for memory by warning that supply constraints and memory cost pressures will continue and intensify.

SPY, QQQ and SOXX Levels

SPY is looking more neutral, and we will likely bounce between 740 and 750.