Download

OPOs v IPOs
October 2026

I. Old Public Offerings vs Initial Public Offerings
Warren Buffett famously quoted that he was more interested in old public offerings (OPOs) than initial public offerings (IPOs); i.e. that buying existing listed companies was likely to be of more interest and of lower risk, with greater diversity, than getting caught up in the hype of higher-risk new listings.

A classic example of this was witnessed in Australia over the last couple of months with the Firmus IPO. That IPO failed to satisfy the ESG i.e. governance standard. It also failed the Thomas Ross Burrell OBE test: that you value a business at a point in time, taking into account some reasonable growth, but ensuring that there is a good margin left for the retail investor to support the aftermarket.

Firmus failed to have a developed business, the founders failed to support the roadshow of the IPO and the founders were greedy and sought to capitalise on the momentum and sentiment of the AI theme.

It was therefore not a complete surprise to see the float flounder as experienced heads refused to draw a cheque to support the high valuation.

In contrast, investors would be well placed to consider Google, one of the major hyperscalers listed in the USA. Burrell engages daily in direct international purchases and sales of listed securities. The direct international portfolios have performed well over recent years.

Google (Alphabet Inc: GOOG) reported its second-quarter results in July. Their quarterly revenue was US$120B (consensus: US$117B), There was an 82% surge in cloud revenue to US$25B on a record 36% operating margin. Internet search revenue increased 17% to US$63B. YouTube ads contributed US$11B. The stock is currently trading on an undemanding P/E of 17x based on historical earnings, and 22x based on 2027 earnings.

In common with all of the hyperscalers, there are material increases in capital expenditure. In its second-quarter results, Google increased the 2026 capex target to US$205B, with US$45B expended in the quarter.

The major issue for businesses such as Nvidia and the hyperscalers, including Google, Microsoft and Amazon, is whether they can generate the revenue from the increased capex to maintain their margins. This issue may be ameliorated by selling completed data centres to natural infrastructure owners, e.g. superannuation and pension funds.

A key factor to understand is that when an AI search is conducted, it does not go to the computer of the business or person conducting the search. Rather, Copilot will go to Microsoft’s servers, Gemini will go to Google servers, and Anthropic Claude to its servers. The hyperscalers will charge for tokens, which is an amount for each four characters (three quarters of a word). It is no wonder that one can quickly rack up a material number of tokens when only four characters are being counted, particularly if the prompt being undertaken involves large quantities of data.

It is this management of tokens that is the challenge for business. For the hyperscalers, the challenge is trying to estimate what the real demand for AI will be when businesses and consumers have to pay realistic amounts for tokens.

Another feature of the Firmus IPO, which failed the now-known railway lines for data centres, is that the requirements for a data centre are as follows:
•Preferably, equity money to fund the construction of the data centre.
•Power contracts for the long term at reasonable prices.
•Land and building capability is a given for what are specialised buildings.
•Contracts with the hyperscalers to ensure that there will be demand for the data centre output.

The model for Firmus was unclear to your diarist. Electricity historically had base-load, intermediate-load and peak-load, with peak-load (usually gas) demanding higher prices but only being switched on when required because of high costs. In the Firmus case, there would be so much demand for data centres that Firmus would be able to charge peak-load prices and generate monopoly profits.

Your diarist has material doubts that this business model will be successful in the longer term.

If you have been thinking of buying a direct international stock, please contact Burrell to receive some independent research on Google or any other direct international stocks that may be of interest. There is a BWET 28 global model portfolio supported by independent research.

II. Computer Chips
In the 18 months to 31 December 2025, the major gains from AI were in the hyperscalers. Also included was Nvidia, which has the intellectual property protection on the major AI computer chips known as Blackwell and Rubin.

There were some other stocks in the chain of inputs to data centres that also performed well, including ASML. ASML is a Dutch company that manufactures the machines that make the chips. These machines sell for US$200-300m per machine. ASML has been regarded as having a global monopoly on this business, although more recently, there is some suggestion that the Chinese have copied or replicated a form of this technology.

After 31 December 2025, the hyperscalers took a break and the momentum switched to other inputs to data centres, including servers, storage and computer chip manufacturers. There was somewhat of a bubble in these stocks preceding this momentum switch during the period to June 2026. As a result, portfolios that had concentrated on the hyperscalers marked time, while those who held chip producers saw upside.

That all came to a halt in mid October 2026, when Microsoft unveiled its latest MAIA 300 chips, which it says will be its major chip for inference, that is, processing AI models. Microsoft says that MAIA 300 will process 40% more tokens per megawatt of power than graphics chips, such as Nvidia chips.

Amazon has Trainium (Training) and Inferentia (Inference) chips. Google is on its eighth generation of AI chips, which it calls Tensor Processing Units (TPUs).

The hyperscalers that process AI requests are therefore becoming increasingly vertically integrated, using their own chips and reducing their costs and dependence on higher-priced Nvidia chips. The higher-priced chips will still be required to train the models, etc.

This is the latest iteration in markets trying to understand the AI phenomenon. There is real demand for AI across consumers and businesses, but the extent of that demand remains difficult to predict. Various players in the AI universe are positioning themselves to ensure they can earn a reasonable return on the large capital expenditure required.

Happy investing.

Disclaimer & Disclosure: Burrell Stockbroking Pty Ltd and its associate’s state that they and/or their families or companies or trusts may have an interest in the securities mentioned in this report and do receive commissions or fees from the sale or purchase of securities mentioned therein. Burrell Stockbroking and its associates also state that the comments are intended to provide information to our clients exclusively and reflects our view on the securities concerned and does not take account of the appropriateness of the recommendation for any particular client who should obtain specific professional advice from his or her Burrell Stockbroking Pty Ltd advisor on the suitability of the recommendation. Whilst we believe that the statements herein are based on accurate and reliable information, no warranty is given to its accuracy and completeness and Burrell Stockbroking Pty Ltd, its Directors and employees do not accept any liability for any loss arising as a result of a person acting thereon.

This document contains general securities advice only. In accordance with Section 949A of the Corporations Act, in preparing this document, Burrell Stockbroking did not take into account the investment objectives, financial situation and particular needs ('relevant personal circumstances') of any particular person. Accordingly, before acting on any advice contained in this document you should assess whether the advice is appropriate in the light of your own relevant personal circumstances or contact your Burrell Stockbroking advisor. If the advice relates to the acquisition, or possible acquisition, of a particular financial product, you should obtain a Product Disclosure Statement relating to the product and consider the Statement before making any decision about whether to acquire the product.

Burrell Stockbroking Pty Ltd (ABN 82 088 958 481), a Participant of the ASX Group and the NSX.