September 2026

Hi {{first_name}}

The fund rose 4.2% over the September quarter, against 0.8% for the MSCI ACWI (AUD), despite a 2.7% fall in September (MSCI +2%).

Macro events drove the month rather than company results. Brent crude rallied from US$91 to a peak of US$109, the US 10 year yield rose from 4.75% to 5.29%, and the Federal Reserve raised rates by 25bp.

Over the past 12 months the fund has returned 12.1% (MSCI: 9.4%), and 17.0% p.a. since inception (MSCI: 13.4% p.a.).

Our up/down capture ratio is 1.4x. We have captured 109% of benchmark gains in positive months and only 78% of benchmark declines in negative months. The Sortino ratio is 2.7.

Investment cases for our themes and top holdings are at laserbeamcapital.com/thematic.

Most of the decline came from software, our Australian holdings and the index hedge.

Early in the month we took losses, cut positions and added to the index hedge. We always put capital preservation first in volatile markets.

On 8 September Meta launched Muse, an AI agent that gives every user their own computer in the cloud and does the work itself. As agents do more, we believe value moves from software front ends and middlemen to the agents and the computing power they run on.

We exited the rest of ServiceNow, having taken profits after a rally of around 25% from our first purchase at US$116 in March.

Halfway through the month we moved the capital into AI hardware, after share prices had pulled back from their June highs. Our view on AI computing demand is unchanged, and the autonomous agents are a clear catalyst.

Most of the buying was done straight after the Fed raised rates. We could move quickly because we know these companies well, in many cases from owning them before. Those new positions almost halved the loss for the month.

We closed September 76% invested, with 11% in short positions and index protection. Cash is 24%, against 7% at the end of August. Close to 60% of the fund is in US dollar assets.

Each line nets the stock with its options hedges.

The market is moving quickly. The breakthrough of the past few months is a simple consumer interface for autonomous agents that run on their own computers.

xAI's Grok Bot came first in August. Meta's Muse went viral in September, and OpenAI launched Dots on 29 September. We expect Anthropic to follow. These agents work around the clock, so they need significant computing power.

We own the companies selling the servers, chips and cloud capacity that every AI workload runs on, whichever model or platform wins: Hewlett Packard Enterprise (6.2%), DigitalOcean (6.2%), Advanced Micro Devices (5.9%) and NVIDIA (3.4%). They sell to every large AI buyer, so our case does not rest on any single customer.

DigitalOcean is a neocloud. It rents GPU capacity and AI tools to businesses that want to run AI applications, including open source models, without the cost and complexity of the largest providers. As open models get cheaper and better, more companies can run their own AI. Consensus has revenue rising more than 50% next year, with gross margin holding in the high 50s.

The opportunity came after the sector had traded sideways for six months. By month end Advanced Micro Devices was 15% above our entry, Hewlett Packard Enterprise 11% and Meta Platforms 8%.

Muse shows the scale of demand. It passed around 3 million US downloads in about two weeks, faster than ChatGPT, Meta AI or Gemini at launch.

Each user effectively rents a small computer in a Meta data centre. One estimate puts the bill at 100 million users at around 1.6 million server processors, about 4% of what the industry makes in a year, plus memory that is already sold out for 2026. Meta plans to double its AI computing capacity from 7 GW this year to 14 GW in 2027.

As frontier models converge and open models get cheaper, we believe the platforms with the users and the distribution will win.

Open models are closing the gap with the frontier at a tenth to a fiftieth of the cost, and they are taking share quickly. When models are cheap and interchangeable, value moves to whoever owns the customer and the computing power.

We own three platforms: Meta Platforms (6.1%), whose edge with Muse is distribution, Amazon (3.4%) and Alphabet (2.6%). We expect Alphabet to follow with a simple, integrated AI agent. It already has the full stack: its own chips, its own Gemini models, and Search, Gmail and Android, which billions of people use every day.

Everyone is racing to vertically integrate. Players across the stack are moving up and down the value chain to build stronger moats and defensibility.

SK Square [402340 KS] (4.2%) replaced Micron as our memory exposure. SK Square is a Korean listed holding company for SK hynix, a leader in the high bandwidth memory used in AI servers. At month end SK Square traded about 45% below the value of its SK hynix shares. SK hynix's Korean shares in turn traded about 40% below its Nasdaq listed shares, which are equity in the same business. Both discounts have reasons behind them and we are not counting on either to close, but they give us a margin of safety. We bought after SK Square's shares halved.

Around 42% of the fund is now in AI hardware and the large platforms. There are two main risks. The first is that AI spending or agent adoption slows. The second is the one in the headlines: rising bond yields. The US 10 year yield rose from 4.75% to 5.29% in September. Higher yields lower what investors will pay for future earnings, which hits growth companies hardest. Current yields are priced in. Higher yields are not.

For our holdings, we believe earnings growth is offsetting the higher discount rate. At current levels we are comfortable with the valuations. If the 10 year keeps rising after oil falls, or earnings growth slows, we would cut exposure further and increase hedges.

Also, adoption is still early. Only around 2% of US households paid for an AI subscription in April 2026, up from almost none at the start of 2023. We believe it will keep accelerating, which underpins our positioning.

The ASX 200 fell 3.2% in September and is down 0.7% over the past 12 months (excluding dividends). The Small Ordinaries Accumulation index fell 3.0% and is down 9.3% over the past 12 months.

The Australian market is becoming more challenging. Our Australian industrial holdings, Praemium and Zip, fell with the weak market and were among our largest detractors.

Innovaero, an Australian sovereign armed drone manufacturer, listed on the ASX on 22 September and closed the month at A$0.43, below the A$0.50 offer price but about 190% above our original entry. The next step is qualification of its OWL B system under Mission Talon Strike and whether that converts into production orders. We remain confident.

Australia remains a good hunting ground for commodities and resources, which are around 12% of the fund. We plan to exit our other Australian industrial holdings by year end. The exception is Innovaero, our defence holding, which we will keep while qualification and orders play out.

Beyond resources, the property and consumer cycle is getting tougher and we believe the opportunities are on the short side. Our short in Commonwealth Bank of Australia [CBA] added to returns as the stock fell 6% in the month, and we are looking for more ways to position for a weaker domestic economy.

Eventually this weak sentiment will create a great buying opportunity. We are not there yet.

Late in September the US share market sat within 1% of its record high, yet more than 70% of its stocks were at least 10% below their own highs. AI spending is holding the index up. Outside AI, the market is acting fairly rationally. It’s already in the drawdown you would expect from higher rates, inflation and war.

We believe pressure builds on the US administration to end the war and bring down oil and bond yields before the midterms on 3 November. If it does, the rest of the market has room to catch up. If it does not, we will become more defensive, and our hedging strategy is designed to reduce large drawdowns.

In July the hedge was our largest contributor. In September it cost around 0.7% as markets recovered in the second half of the month. We expect to pay for every hedge, like insurance. Early in the month it cut net exposure from 93% to 34%, which let us reposition rather than sell into a falling market.

We enter October with 24% cash and net exposure of 66%.

The portfolio is built around five areas: AI Hardware and Cloud (around 30%), AI Platforms (12%), Resources and Critical Minerals (12%), Defence (8%) and Healthcare through Eli Lilly (7%). We also hold a small short against the Australian property and consumer cycle through CBA.

We continue to prefer offshore over domestic exposure, for both our investments and our cash.

We will add as opportunities meet our criteria and concentrate the fund’s capital in the best ideas globally. We will reduce exposure when stops are hit, the facts change or conditions call for the hedge.

Live performance and holdings are always updated at laserbeamcapital.com/performance.

Please call any time if you’d like more detail.

Regards

Hedge Partners Pty Ltd ACN 685 627 954, trading as Laser Beam Capital (Hedge Partners) is a Corporate Authorised Representative (CAR No. 1314946) of Non Correlated Advisors Pty Ltd ACN 158 314 982 (AFSL No. 430126). Hayden Beamish is an Authorised Representative (AR No. 1314950) of the same AFSL holder. Hedge Partners and Hayden Beamish are authorised to provide general advice only to wholesale investors. Nothing in this communication constitutes an investment offering unless expressly stated. Past performance is not a reliable indicator of future performance. Metrics are subject to change. References to holdings, top contributors, top detractors or example position may not represent the full portfolio. Certain positions may be withheld from disclosure where we are actively managing position size, have not reached full weight, or cannot disclose for any other reason. This email is for information only and is not investment or financial advice. Before acting on any information, obtain independent taxation, financial and legal advice and consider it carefully. This email and any attachments are confidential and intended only for the named recipient. If you are not that person, please delete it and notify the sender. Email transmission cannot be guaranteed to be secure or error free. The sender accepts no liability for any viruses, errors or omissions arising from email transmissions. Important information: This document contains forward-looking statements which are identified by words such as ‘will’, ‘may’, ‘could’, ‘believes’, ‘estimates’, ‘targets’, ‘expects’, or ‘intends’ and other similar words that involve risks and uncertainties. These statements involve assumptions, known and unknown risks, uncertainties and other factors that may cause actual events, results, performance or achievements to be materially different from any future events, results, performance or achievements expressed or implied by such forward-looking statements in this document. Consequently, undue reliance should not be placed on these statements. The author does not warrant or represent that the actual events, results, performance or achievements will be as discussed in those statements.