April 2026 Investment Update

Dear investors and well-wishers,

The fund declined 2.9% in March and was up 29% over the twelve months to end of march.

After a strong rally over the last two weeks, the fund has recovered the drawdown and is now positive calendar year-to-date. Our risk system moved us to over 50% cash at the lows, and then gave a broad set of buy signals, similar to the setups that preceded triple digit gains from the lows in April last year, and November 2023.

But the first quarter was tough:

Anything could happen this time around, but it’s encouraging that the quants finished selling and are now significant buyers, with some way to go:

These charts have been surprisingly effective over the last few years, and it’s interesting to see it work again.

Our best explanation for this is that most shares don’t trade hands in sell-offs, as the substantial holders are the same at the beginning and the end. But systematic funds do change positioning significantly, and are one of the major swing factors in the market.

Since they use price and volatility to systematically change exposures, it’s possible to estimate what they will be required to do given a set of public market data, and post hoc trading desks which see most of the flows can confirm.

It has been a winning strategy to buy when these funds are heavily short, knowing that at some point that exposure will flip back.

Semiconductors

Our largest exposure is in semiconductors, heavily weighted towards megacaps like Nvidia, AMD, and Broadcom, with smaller positions in companies like IQE and AAOI.

AMD and Broadcom are looking positive as custom chips and CPUs respectively become more relevant. Agents use GPUs for the language modelling steps, but all the retrieval, tool calls and coordination is on a CPU.

The rolling shortages and focus of speculative attention moved from memory to optical companies, as there is a widely flagged shift away from copper connectivity to faster optical solutions.

AAOI had a huge rip as the market priced in a multibillion dollar near term revenue forecast (the market cap was only a few billion). We bought at ~$27 and sold 83% of the position at an average of $88 for a ~3.3x return, sticking to our strategy of realising profits around 2.5x - 4x. We still have a 3.6% position.

There has also been notable relative weakness within the sector from Nvidia. This is likely as the threat of custom chips is materialising. Not necessarily in a thesis-busting way, but certainly Marvell, Broadcom and enablers of Google’s TPUs and Amazon’s chips which run much of Claude has taken some of the spice out of Nvidia’s market.

Multiples were low - Nvidia’s FY2025 multiple nearly reached 15x in the drawdown, which is not expensive for their dominant high margin position.

Neoclouds

I get asked for our view almost every day. This year we are in a compute shortage, so while that remains the case, the suppliers of excess capacity should do well.

I would caution that these are commodity companies, look how many there are here, and some Australian names don’t even make the list:

I would also note that hyperscalers, who know best, can build datacenters the cheapest, have the best access to Nvidia and other GPU suppliers, will only rent capacity when it’s in their commercial interest.

So it’s no surprise that neoclouds focus on forecast revenues, when fully-costed margins are negative or low for even the most mature companies. Coreweave posted a -23% GAAP profit margin on US$5 billion, and that’s before capex.

Beware high EBITDA margins that strip out the capital cost of GPUs, the datacenter, and the financing costs. They are the main costs!

I’ve also seen references to recurring revenue, when these are all fixed term contracts, often with exit clauses, and the key reason Meta or Microsoft would lease rather than build themselves is so that they can access better GPUs in the future.

There are plenty of scenarios where suppliers from Nvidia down will do well, even where token generation is done in-house by the hyperscalers. I’ve seen sectors of excess commodity capacity wiped out before, notably in mining services, and also more recently some of the very same companies in their prior incarnations as bitcoin miners.

But does the likelihood of 2030+ contract renewals matter for an IPO today? Probably not.

Other than GPU pricing, which can be tracked, Coreweave’s and other bonds have traded well lately, so the window looks open.

Jim Chanos noted an interesting dynamic - when Coreweave and the 100+ competitors buy chips this is booked as revenue and profit this year for their suppliers, but the cost is spread out as depreciation over a number of years. Neoclouds go to extraordinary lengths to adjust this cost out.

Best to make sure you’re properly accounting for all costs, including quite high interest rates, when modelling these commodity leasing companies.

Small and mid-cap growth

We also cautiously and carefully entered beaten down consumer stocks like Reddit, and Robinhood, focusing on those growing well over 25%. If conditions deteriorate I suspect we will build up a large cash balance again, but the last two weeks is a reminder that beaten down sectors can recover very quickly.

I truly believe this is the best way to harvest cash from these kinds of volatile growth stocks, which seem to have >50% sell-offs every year.

Outlook

I wrote in my last letter, Wartime Markets, that conflicts are generally good for indices, so the current strength shouldn’t be a surprise, though somehow, rallies like this always are.

Military spending combined with a generational AI capex buildout is more than enough to support markets, even while consumers are squeezed.

So it seems we have a strong market led by defence, industrials, and semiconductors, while consumers and related stocks are squeezed. I don’t think this is a regime worth fighting.

And of course, the bears can’t get too excited as even the worst consumer scenarios are shielded by Government spending. Australia has one of the largest bureaucracies per capita of any developed nation, so this softens the pain in the private sector.

We’ve now had a steep sell-off in small and midcap growth - which is still continuing for the obvious AI losers - a sharp but modest sell-off in broader indices with a rapid recovery and breadth thrust, all while the market absorbed higher costs (negative) and spending (positive) that come with war. And meanwhile token consumption is exploding upwards, and this is all but guaranteeing aggressive semiconductor capex at market-moving scale.

This is what a wall-of-worry looks like, and we will stay aggressively invested in the winners while our risk management system flashes green, and move defensive when it flashes red, regardless of the reason.

In recent weeks we’ve seen our systematic approach act decisively in both directions, which is encouraging, as far as I can tell unique in the market.

Best
Michael