Dear investors and well-wishers,The fund advanced 13% in February and is up >5% in March so far, taking us to 30% for the calendar year-to-date. The Nasdaq 100 is up +8%, ARKK is down -3%, cloud computing is flat, and small caps are up +3% (21 March).
Clarity Pharmaceuticals
Clarity released data from both diagnostics and therapeutics, and continues to look market-leading (deep dive here). Two days ago AstraZeneca bid US$2.2 billion for a Phase 2 prostate cancer therapeutic at a 126% premium to the prior close. As a sense check Clarity’s market cap is ~US$500m, which is why we are targeting an share price of $9.
The company expects 30-50% of heavily-pretreated patients to see PSA drops of over 50%. As of today, Clarity has observed 80% drops in 50% of patients, with no dose-limiting toxicities.
Clarity also has superior IP over the core parts of their copper isotope technology, whereas Fusion targets the same targeting ligand as Point Biopharma and uses the non-patentable element Actinium.
As a reminder, Point released poor Phase 3 data targeting the same ligand as Fusion shortly after Bristol Myers Squibb paid US$1.4 billion for it.
Clarity's data continues to outperform competitors, and this week brought more evidence of a lively M&A market in the space, where acquisitions are priced several times higher than where Clarity trades today.
Semiconductors
In late 2022 I wrote that semiconductors was our single highest conviction position on the thesis that demand for AI would lead to a step-change increase in demand for compute.
Over a year in, and it’s clear this is the regime we are in, and increasingly clear who the market leaders will be.
For most of my adult life a very small number of US tech companies have created enormous wealth by billing almost every company and consumer on the planet, at exceptionally high margins.
Every year you worked someone was likely paying a seat fee for access to Microsoft Office, and every time you used the internet you were likely mined for advertising by Google, Facebook or both, often on an Apple or Google phone.
These companies used (and sometimes abused) their immense power to cut swathes through the rest of corporate life, at best taking the first part of revenue from every industry, at worst destroying those industries.
As well as generating enormous profits, this incurred significant costs for basically everyone else, leading to a stock market most of the gains were captured by these companies. This made investing easy, or hard, depending how you looked at it. We now know for example that buying deep value in media dramatically underperformed the entire time.
But now these companies are facing their own new tax - they are in desperate need of computational capacity to serve new applications in AI.
It’s easy to forget Google’s technological achievement in indexing the web and making it available to billions of people concurrently, or Facebook connecting us with billions of others at speeds so fast it feels instant.
But these herculean efforts are dwarfed by the compute required for the roll-out of large language models.
The tech giants need GPUs that are only designed by NVDIA, which in turn are only made by Taiwan Semiconductor, who can only get the necessary chip printing machines from ASML.
There are substitutes, but AMD lacks Nvidia’s CUDA software, Samsung is a distant second to Taiwan Semiconductor in manufacturing and has an unwelcome habit of competing with customers, and ASML, well, they really are they only option.
And woe betide any major tech company that falls a year behind here. This investment is both essential, and profitable (as they rent this capacity out at high returns anyway).
New developments
And every week there are new developments which add to future demand for compute.
Three come to mind:
Last week Cognition released Devin, a text-to-engineer application which turns text prompts software. Devin acts as a software developer with access to a terminal, GitHub and the internet. Microsoft is launching a product a similar product and there will be others.
So instead of a series of discrete requests to ChatGPT or equivalent, Devin works away constantly in the background, querying the internet, while GPUs around the world fire up and whizz away, with multiple large language models speaking to each other, with each token calculated from the entire project history.
And OpenAI released SORA, the most advanced text-to-video application, each use seemingly recreating the physical world with all its laws and relationships.
Most of us aren’t video creators so text-to-video is a niche application. But the computational demand is substantial.
GPUs generate each frame pixel-by-pixel, taking into account not only every other pixel like image generators, but every pixel in every prior frame, as well as the text history and training data.
These are not even publicly available, but we already have widespread throttling of language models to account for heavy use.
Thirdly, GPT-5 is rumoured to be excellent, and has also not yet been released. The demand shock has barely begun. Perhaps the best way to play is a mixture of clear market leaders and the semiconductor index itself, which has weightings to a recovery in smartphones and will capture up-and-comers.
Stocks
Some parts of the opportunity are easier to analyse than others.
I mentioned the market leaders above (Nvidia, TSM, and ASML).
There are other candidates: KLA and LAM come to mind, as does SK Hynix which provides the memory for advanced Nvidia chips.
But the story becomes more complicated as you move further down the chain, as most of these companies have only one major customer (the market leaders) and so the ultimate commercial position is difficult to assess and changes rapidly.
The larger companies have the power.
I haven’t mentioned AMD which for a time was one of our largest positions, and was trading at a significant discount to Nvidia only a few months ago. But after blow-out results, Nvidia is now the one at a discount. And their recent conference highlighted new technology which emphasized how far ahead they are than everyone else.
It’s still a safe bet that Microsoft and Meta will continue to support AMD, but second is not quite the same as first.The current regime has been in place for nearly a year and a half, counting from the launch of ChatGPT. Similar regimes where software, e-commerce, and social media held market leadership lasted for the better part of a decade.It’s fair to consider carefully where we are in the cycle. The good news is we are still in the phase were leading companies like Nvidia are posting >200% organic growth rates, trade on GAAP PEs of under 30x, and have product lines sold-out well into next year.
Which all suggests there is a long way to go.
