Like I said on a recent thread concerning Saul’s board itself, I am keen to continue posting my monthly updates here out of respect to Saul and the other board members, and out of gratitude for all that I have learnt here over the years. It does take me a long time to write these reviews - but it is a valuable process for me at least.
September has been a good month for me up 17% in this month alone. Year to date I am up 67%. My ATH was at the end of June when I was up 101% for 2026, so I have come down since then, but I am still very pleased with 67% so far this year - especially after the stellar growths of the last 3 years. And I think Q3 will be even better for all of my companies - which is why I continue to hold them
Here are my results since 2024.
-
2024 + 70%
-
2025 + 117%
-
2026 (monthly) (YTD)
-
January -5.1% -5.1%
-
February -4.1% -9%
-
March -1.3% -10.2%
-
April +33% +19.5%
-
May +51% +80.3%
-
June +12% +101.2%
-
July -28.7% +43.6%
-
Aug -1.3% +41.7%
-
Sept +16.9% +67%
That is 6.15 times what I had in my portfolio at the start of Jan 2024! So this clearly shows the power of compounding and running a concentrated portfolio of the best growth names. It has done wonders for my portfolio.
I’ve been investing and keeping track of my returns since 2005 , and I’ve been a member of TMF for most of this time. But it wasn’t until I found Saul’s board in 2020 that my returns really took off (apart from my huge loss in 2022 of course that I have written about elsewhere on this board). But you have to be prepared to stomach big losses with this kind of investing in order to get the big gains.
My gains would have been even higher this year - but like I said in one of my previous reviews a couple of months ago - I did take a big chunk of money out of my portfolio back in June. And I am pleased I timed that withdrawal well - because that turned out to be the high point so far this year.
As of today, at the end of September, my portfolio looks like this.
-
Nebius (NBIS): 24%
-
Astera Labs (ALAB): 22%
-
Micron (MU):19%
-
AppLovin (APP): 12%
-
Celestica (CLS): 12%
-
IREN (IREN): 9%
-
Electro Optic Systems (EOS): 1%
For those who have been following my monthly reviews over the last few years now you will see once again that I have made virtually no changes yet again. The only changes I did make in September were to buy a few more AppLovin and IREN after both of those had fallen in price. I still like them both a lot.
I’m still very concentrated in companies that I believe are benefiting from the buildout of AI infrastructure, with APP and EOS being the exception. I remain very bullish on all seven. That doesn’t mean I expect them to rise in a straight line. July was a painful reminder of how quickly this sort of portfolio can fall (down 29% for the month). But when I go back to the earnings reports and the announcements since then, I see more reasons to be excited about the businesses than I did at the start of the summer.
Here are the YTD gains on each of these stocks and also the gains from when I first bought each one.
Stock YTD % Gain % Gain since first bought
NBIS +177% +570% (Jan 2025)
ALAB +111% +206% (Dec 2024)
MU +269% +269% (Jan 2026)
APP -54% +306% (April 2024)
CLS +21% +862% (Feb 2024)
IREN +11% -30% (Oct 2025)
EOS +18% +285% (July 2025)
So you will see that Celestica is my longest held stock at over 2.5 years now, with AppLovin just behind it. But it is because I made Nebius such a large holding for me initially that I have continued to do really well in this name - and it is Nebius that still remains my number one holding and highest confidence position.
Nebius (NBIS)
The Q2 figures were quite extraordinary. Group revenue reached $582.3 million, up 454% YoY. The AI cloud business generated $575 million of that revenue, up 514% YoY, and reached $3 billion in annualised run-rate revenue at the end of June. Its adjusted EBITDA margin was 50%. Those numbers tell me that this is becoming a very substantial business, even though the company is still in the middle of a huge expansion.
What interested me even more was the discussion of new contracts. Nebius closed four large AI cloud deals in Q2 with an average total contract value of more than $1 billion each. Management said that the deals were being signed at better prices, and that about 70% included customer prepayments. The expected payback period on the Q2 deals came down to one year and ten months. Much of that new capacity is due to arrive late this year, so I think we will see the real effect of those contracts in 2027.
Nebius exited the second quarter with an ARR of $3.0 billion. That means June 2026 has a $250M/month revenue.
2026 Exit ARR $7-9B means Dec-2026 will be $666M per month~! (from $8B/12) about x3 revenue growth in 6 months visible now.
I also like that Nebius is continuing to build beyond GPU rental. Token Factory’s production inference workloads more than tripled in Q2, and the Eigen AI and Clarifai teams have joined the company. Then in September Nebius announced a partnership with Palantir to provide a sovereign AI stack to Palantir customers. I see these as useful steps towards the full AI cloud platform Arkady has been describing.
But perhaps the most bullish announcement came a couple of weeks ago when Nebius announced they were putting up their pricing for all of their GPU’s including Hopper. That shows that there is still huge demand out there and shortage of supply.
The combination of signed demand, improving deal economics and a growing software offering is why I continue to hold it with such confidence.
Astera Labs (ALAB)
Astera’s Q2 was another excellent report. Revenue was $392.4 million, up 104% YoY and 27% from Q1. GAAP gross margin was 73.3%, and the company remained profitable. The growth came from more than one product line, with record revenue for Aries as well as strength across its AI fabric products.
The big thing I am watching now is Scorpio. Management expects the Scorpio X-Series switches to become Astera’s largest product family in Q3, one quarter earlier than previously expected, and guided to Q3 revenue of $540 million to $560 million. At the midpoint that would be a very large sequential increase. We will soon find out how the production ramp is progressing, but I thought that guide showed considerable confidence.
What appeals to me about Astera is that each new generation of AI hardware needs more ways to move data reliably between chips, servers and racks. Astera already has products in signal conditioning and switching, and it keeps finding more places where connectivity or memory is becoming a bottleneck. Its September announcement expanded the Leo memory controller family, including a product designed to put memory directly on the AI fabric for demanding inference workloads. The company says it is seeing new design wins across AI labs, hyperscalers and neoclouds.
I have held ALAB through some very big swings. As long as Scorpio keeps ramping and the other product families keep contributing, I think there is a lot more growth ahead.
Micron (MU)
I sold my Nvidia holding in January and put the proceeds into Micron. It was a bold switch, but I’m really glad I did. I thought memory would become far more valuable as AI systems expanded, and Micron’s most recent results gave me little reason to change that view. They report their Q4 tomorrow.
Its fiscal Q3 revenue was $41.46 billion, compared with $9.30 billion in the same quarter a year earlier. The company reported $25.39 billion of operating cash flow and guided to approximately $50 billion of revenue and an 86% gross margin for fiscal Q4. Those are remarkable figures for a company many investors have been used to treating as an ordinary cyclical memory stock.
Its next earnings call is on 30 September, the day after I am writing this, so I will be reading that one especially closely. But I still see them as massively undervalued with an incredibly low PEG ratio. In fact you almost need a magnifying glass to see it!
APPLOVIN (APP)
I first bought it in April 2024 at around $75, and I have held through some enormous rises and falls since then. The share price can be unsettling, but I keep coming back to what the business is actually producing. I actually bought some more again this past month.
Q2 revenue was $1.924 billion, up 53% YoY. Net income was $1.267 billion, adjusted EBITDA was $1.614 billion, and free cash flow was $863 million. That is an astonishing amount of profit and cash from a company still growing revenue at more than 50%. For Q3, management guided to revenue of 2.055–2.085 billion and an adjusted EBITDA margin of 83%. I was very pleased with those numbers.
The development I have been waiting for is the opening of AppLovin’s advertising platform to everyone. It happened on 22 June. For years an advertiser needed a relationship with the company, and more recently a referral code. Now any business can sign up for AppLovin Ads, powered by the Axon recommendation system. Adam says the platform reaches more than a billion daily mobile game users. If AppLovin can show online shops and other businesses a good return on their advertising spend, I think the opportunity is far bigger than its original market of gaming advertisers.
I like that the company is widening the ways advertisers can use it. An online shop can set a return-on-ad-spend target; other businesses can pay towards a target cost per purchaser; and AppLovin has added lead generation for areas such as insurance and home services. The public launch is still very recent, so I am not assuming that every advertiser who signs up will immediately become a big spender. What I want to see over the next few quarters is those new customers getting results and then increasing their budgets.
APP has also been buying back shares: in Q2 it spent $551 million on repurchases and share buybacks. I would still own it for the growth in Axon and the advertising business even without the buybacks, but the cash generation gives management room to do both. I remain very bullish, and I think the opening of the platform could prove to be a significant moment in the company’s history. And I think their next Q will satisfy many anxious investors in APP.
IREN (IREN)
IREN is perhaps the clearest example in my portfolio of a business changing before our eyes. I originally followed it with Bitcoin mining still an important part of the story. Now I am holding it for the AI cloud business it is building on its power and data-centre footprint.
The numbers need reading carefully because the AI revenue already earned is much smaller than the contracted opportunity. AI Cloud Services revenue rose to $70.5 million in the June quarter from $33.6 million in the March quarter. For the full fiscal year it was $128.8 million, compared with $16.4 million the year before. At the August results, management said it had $4 billion of contracted annualised run-rate revenue for 2026 capacity, while $1 billion was operating at that point. Those are different measures, and getting the contracted capacity installed and accepted by customers is the next test.
There has been some very encouraging progress. IREN delivered Horizon 1, the first of four 50 MW liquid-cooled deployments for Microsoft, and said Horizon 2 was being commissioned. It also announced a new multi-year agreement with a leading frontier AI lab and named several other recent customers. Its acquisitions of Mirantis and Nostrum add software and services to a business that I had once thought of mainly in terms of power, buildings and GPUs.
The financing arrangements matter a great deal here. IREN said the financing and prepayments for the Microsoft contract cover 96% of the associated GPU capital expenditure, while its newer GPU financings support other deployments. That does not remove the cost or the execution risk, but it helps explain how management plans to fund such a rapid buildout.
I think IREN could look like a very different company a year from now.
Celestica (CLS)
Celestica is my longest-held stock, and it continues to impress me. Its Q2 revenue was $4.70 billion, with adjusted EPS of $2.54 and an adjusted operating margin of 8.2%, another company high. Rob Mionis raised the full-year outlook again: $20.5 billion of revenue, $11.30 of adjusted EPS and $600 million of free cash flow. That revenue outlook represents 65% growth for the year.
What really caught my attention was the outlook beyond this year. Management said it now expects 2027 revenue growth to accelerate beyond the 65% anticipated for 2026, with adjusted EPS growing faster than revenue. That is an unusually strong statement for a business already growing at this rate. Of course, it is an expectation rather than an achieved result, and customer plans can change. But it comes alongside stronger customer forecasts, better component supply and new programme wins.
I continue to like Celestica because it is supplying the physical systems needed for data-centre growth and has been executing quarter after quarter.
Electro Optic Systems (EOS)
EOS is different from the other stocks I hold, but I think it deserves its place in my portfolio. Modern warfare has made the threat from cheap, numerous drones impossible to ignore. EOS has remote weapon systems, high-energy laser technology and, following the MARSS acquisition, AI-enabled counter-drone command-and-control capability. I like the prospect of those pieces being offered together rather than as isolated products.
The first half results showed how quickly this business can grow when contracts move into delivery. Revenue was almost 4 times the previous years revenue. EOS has also spoken about positive first-half EBITDA. Defence revenue can arrive unevenly, so I am particularly interested in how much of that order book converts into revenue and cash over the next few reporting periods.
There have been two interesting announcements just this past week. EOS signed a letter of intent with the Netherlands Ministry of Defence to explore making its Apollo laser an operational weapon system, with possible production in the Netherlands. That is promising, although it is not yet a purchase contract. Its R400 remote weapon system was also listed on a US counter-drone procurement marketplace, giving eligible US government and allied customers another route to buy it. Again, a route to orders is not the same as orders already won, but I like seeing these doors open.
EOS has been a smaller holding for me, and I know defence contracts can be slow and unpredictable. Even so, I believe its combination of products is becoming increasingly relevant, and I am keen to see what they can do with the stronger order book and MARSS now part of the business.
Thank you to everyone on this board who contributes to make this board what it is. And of course thanks to Saul. We hope you are well and enjoying retirement.
Best wishes to everyone for October,
Jonathan