Wartime Markets

Dear investors and well-wishers,

We returned -7.9% in February, taking us to -13% calendar year-to-date.

Once again war throws out the most carefully laid investment plans.

At the start of this year market debates were over the second-derivative of AI-related capex, LLM business models, whether SAAS companies would re-accelerate and gain efficiencies or welter under ferocious new competition, and how AI would affect labor markets.

Superintelligence has continued to accelerate. Some comments in my last note were obsolete only days later, when Anthropic came out with major new cybersecurity capabilities and a new chip company posted >10x faster inference than even Groq and Cerberus, albeit with some limitations.

If you want to see the future of inference - and why we are extremely wary of neoclouds building datacenters with the current generation of Nvidia chips, including those trying to list in Australia, try this chatbot.

But instead of these now familiar debates we got a major new conflict.

It’s likely no coincidence that the two most (militarily) successful operations in living memory took place shortly after the 10-100x increase in agentic superintelligence.

And in fact we know this, courtesy of a spat between Anthropic and the Department of War. It’s widely reported that Claude was used both planning and during the raid on Venezuela, and in Iran identified over 1,000 targets in the opening phase.

The traditional war markets playbook has played to the script: energy, commodities, defence and inflation are up. Consumers are getting squeezed as supply is disrupted and capital is redirected, and this coming at a time where interest rates were already rising, and waves of AI ‘efficiencies’ ie job losses are filtering through the market.

We’re now sitting on well over 50% cash. The last time we reached this the fund more than doubled ~6 months following. There is enough of a sell-off here to target the same again, but for now, our quant models are cutting risk further.

This might all resolve quickly with a Trump announcement of victory. Enough military infrastructure has certainly been destroyed, but to for markets to stabilize, Iran needs to stabilize too. This is a country with a population nearly 50x the size of Gaza, and over a thousand times as large. It only takes a handful of radical militants to cause chaos.

The world will adapt. Countries like Australia will reconsider their energy independence, and market prices will at least somewhat limit the damage.

But in the near term this is the major driver of price action. We are oversold now by many metrics, but the question remains whether this triggers a consumer recession (likely) and whether AI-job losses also start showing in the numbers (anecdotal so far, but also likely later this year).

Quants have finally sold. This is the position we were in March/April last year. These are a major swing factor in markets, but are unreliable for timing.

Similar extremes of positioning and sentiment were also reached in January 2022, and that was the beginning, not the end. But when the time is right this is the buying that will reverse and push the market to new highs.

Again, this doesn’t look like a simple a market reset. Rather high interest rates and energy prices are sucking capital out of the economy, so the war likely needs to resolve before the quants flip again.

Winners and losers so far this year

Grail posted negative trial results in their major UK trial. We were optimistic after strong data from >27,000 patients in the United States, but the company fell short in their larger NHS-Galleri trial in the UK with 142,000 participants. This had a much higher bar, targeting a statistical reduction in Stage III-IV cancer diagnoses. They did see some promising signals and will proceed with an FDA filing, but the path to success undeniably narrowed.

And while it’s not over for Grail, it does burn ~$500 million a year, and the clean bull thesis we invested on now has hair all over it. We exited on our risk management framework regardless.

Even after the sell-off, we realised an average +46% profit, but in the midst of a market sell-off a -45% drop in a large position was… unwelcome.

Castle Biosciences also sold off:

And Rhythm Pharmaceuticals had a disappointing trial miss, attempting to expand their Bardet-Biedl syndrome franchise into more common disorders caused by variants of the same gene, where they attempted to expand their drug setmelanotide into more prevalent genetic obesity variants. They missed across all four cohorts.

While disappointing, and certainly detractors for this calendar year, we exited all three positions at a profit. The companies subsequently sold off further, which really demonstrates the power of our quant approach, even when data disappoints and they contributed to our calendar year-to-date drawdown.

These were partially mitigated by a buyout for Day One:

In Australia, Clarity and Syntara were also detractors.

War-induced inflationary spirals are terrible for companies with revenues multiple years out, as we know better than anyone.

The data continued to be strong though, with a decisive win in a head-to-head investigator-led trial of Clarity’s diagnostic.

And Amplia came out with a stunning new update, where 5/64 patients with metastatic pancreatic cancer recorded Complete Responses, with one patient still on trial (so clearly a strong responder). Complete Responses are extremely rare in pancreatic cancer, so this was strong validation of our conviction that there’s a real drug here.

In semis, Applied Optoelectronics was a winner. We reduced our position by 70% at ~3x. The company is guiding to a material uplift in revenues over the next few years, but we’re sticking to our strategy of closing winners at 2.5-4x returns. There’s every chance we will get the chance to go again on this stock later this year.

Marvell is our remaining semiconductor play, posting record revenues and solid 42% year-over-year revenue growth. But this year has felt like death by a thousand cuts as one by one major companies like Nvidia rolled over and we exited with small year-to-date losses.

I expect many of these companies to be amongst the first to recover, especially now Nvidia is trading at a forward PE of 15x. But our strategy doesn’t require us to guess, so we are out until we get one of those time-honoured re-entry signal. Nvidia is one of the examples we tracked back over 25 years, and this risk framework substantially increased overall return and cut the drawdown.

And while we’re as bullish as anyone on demand for compute, there’s every chance capex is flat or negative in 2027. The second derivative has changed, and if this alternate scenario plays out, we’ll be grateful for the sell signals and the cash balance.

For now, we watch and wait.

Big Tech

Big tech joined the software sell-off. Microsoft is now down 35% to a forward PE of 19x, one of its larger drawdowns in recent history:

And Facebook/META dropped nearly 20% this month alone after losing a child exploitation lawsuit in New Mexico.

It will be interesting to see if Zuckerberg scales back spending in Reality Labs, given the weak consumer uptake so far.

I bought their latest device, and while it was fun to play around with in the office and shoot virtual aliens popping out of furniture, I have to admit none of us used it after the first day.

Even Google sold off, now down 13% this year.

On the one hand, EPS growth in these companies is exceptional for their multiple, and big tech companies are the benchmark for any new investment: is this better than a hyperscaler growing EPS over 25%?

But with the exception of Apple, these companies have changed materially, embarking on major capex plans, switching from gushing free cash flow to competing in one of the largest capex builds ever.

So it’s no surprise earnings multiples have contracted. And multiples are particularly misleading when major changes happen on the balance sheet rather than through the income statement.

Nevertheless, for different reasons, they are all some of the more obvious AI beneficiaries, and I’m sure we will be buyers again, perhaps in the very near future.

Outlook

We’re in one of those situations where markets have sold off, but there has been no real panic or capitulation.

Pressure on software continued, not just in the market but from daily updates of new AI capabilities. The impact of Agentic AI systems that came of age at the end of last year has yet to really be seen in backwards looking numbers, but it’s obvious in usage data.

The war is the biggest unknown. As I write it looks like there will be some kind of escalation to gain control of the straits of Hormuz, but my guess is as good as yours.

The impact of all this is an area-under-curve problem. A quick resolution - which leads to a stable Government with rogue militancy under control - would no doubt reverse this sell-off fairly rapidly. Consumers would only take a hit for a few weeks.

But wars always take longer than expected, and the execution of the majority of the country’s leaders may make their successors more rather than less radical, and it only takes a few to cause trouble. And every day it goes on, more cash is sucked out of consumer budgets.

Australia was already suffering from high inflation driven by Government spending, much of it in relatively unproductive areas. And there are already calls for higher wages from the Government sector, which would lock in these cost-of-living price increases.

And we already expected this year to see the first major wave of AI job losses. In our software development agency, we have already cut headcount significantly, and dramatically increased output.

Job losses in a war/energy-driven inflationary spiral is a bleak scenario, which becomes more likely the longer this goes on. And given policy settings, the usual demand release valves are not going to help.

All this speculation is only partially relevant to our strategy, however. Our quant risk models have raised a significant amoutn of cash, which might soon be over 60% of the fund, parked in cash equivalents at >4%, so we have some breathing room while the market churns and all the possible future paths narrow and crystallize into history.

Our quant models were specifically designed for times like 2022 where market sell-offs morph into extended bear markets. In 2022 high Government spending around the world cushioned the blow, and the fact that pain was concentrated mostly on the tech and biotech sector, meant it was more of a market crash rather than a global recession.

It remains to be seen if this time is different, but if we had to guess, holding a lot of interest-paying cash ready to deploy aggressively when markets stabilize is the right way for an equity fund to position.

Michael