As earnings season continues through the month of August 2023, many investors are searching for signals as to whether the second half of 2023 will continue to be led by growth-oriented tech companies or if equities are poised for a pullback. It was the perfect time to speak with an active manager for a take on the global technology space.
Framework for company evaluation
Possibly the single most important part of the discussion was how Dom described his team’s framework for analysing companies, delineated in four areas:
- Lynchpin technology: the company provides something essential to users.
- Innovating in secular growth markets: Dom described this area with the phrase, ‘you cannot sell lifeboats to sinking ships.’ Selling a great product into a market that is doomed to ultimately fail is not the best route to success.
- Improving fundamentals: the specific fundamentals that Dom called out were a) organic revenue acceleration, b) operating margin expansion, and c) free cash flow conversion improvements.
- Reasonable valuations: Dom’s take on valuation was that it always matters but, if you are in the ‘range of reasonable-ness’, the other three areas in the framework could dominate. If valuations are extreme, then its ability to have a more immediate impact could be higher.
When investors think of artificial intelligence (AI) in August 2023, they are usually thinking about Nvidia and semiconductors very soon afterwards. We note here that many are waiting for Nvidia’s upcoming earnings report, scheduled for 23 August 2023 (and not yet released at the time of writing).
Dom set the foundation for the discussion, indicating that there is an expectation that the market for specific semiconductors designed to accelerate AI processing could be roughly $150 billion, annually, by 2027. It is about $30 billion at present which would mean, if this forecast ends up true, the compound annual growth rate implied is in the range of 50%.
Dom referred to Nvidia’s May 2023 earnings report as the ‘print heard around the world.’ He cited that Nvidia’s earnings per share for the fiscal 2026 year went from $6 to $12, which implies roughly $30 billion in incremental net income and free cash flow—a massive figure. Yes, the valuation is high, but Nvidia’s actual results are among the best ever seen in the space.
One of the primary questions we focused on is the potential for a second supplier of graphics processing units (GPUs) for AI acceleration because, looking at the space today, it is not clear that there will be a second supplier. We talked about AMD’s upcoming MI300 chip, and Dom was quite bullish on the capability, but it’s important to remember that Nvidia’s dominance is not simply in hardware—their CUDA software is also important, and it will take time for the industry to create frameworks for easier switch-ability across GPU chipsets.
In the memory space, at least up until very recently, it has been a lot easier for customers to switch between chips provided by SK Hynix, Samsung and Micron, to name a few of the big players. Today, there is a big focus on high bandwidth memory (HBM) and SK Hynix has jumped out to an early lead, but Dom’s view was that Micron and Samsung would ultimately catch up.
We had this conversation on the 11 August 2023, so the discussion of China vs US in the geopolitical sphere had certainly ratcheted up recently with further restrictions. Dom believes the primary consequence is that chip manufacturing will begin in different regions, not just Taiwan.
Will the forces of ‘inflation’ or ‘deflation’ hold the most sway? We asked this because, if the world is making chips in places (like the US and Europe) where labour costs are higher, then it stands to reason that the prices of those chips should go up. Dom’s take was that, historically, Moore’s Law (that is, the principle that every two years you could roughly double the transistors per unit area on a chip at the same cost) was very deflationary. At present, the capital costs have risen significantly. Wafer fabrication equipment intensity, a figure that Dom tracks, was around $35 billion in 2015 and $95 billion in 2022 (even if this year it is looking more like $75 billion). Dom’s view is that it is trending toward $120 billion, which is a big increase relative to 2015.
On the ‘deflation’ side of the ledger, Dom cited AI as a possible productivity enhancer, making the big statement that the impact could look similar to that of electricity.
When we got into the specific AI discussion, Microsoft was one of the top-performers. The company has already set a price point with its CoPilot software package, $30 per user per month, and the primary question to ask today is whether Microsoft will be able to sustain this type of premium pricing in light of the increased capital expenditure requirements, thereby keeping their returns on investment high.
Even if Nvidia’s valuation is often cited as being quite high amongst market participants, Dom was quick to remind people that there are many opportunities that are not quite so expensive. Yes, technology has appreciated in value in 2023, but Dom noted that valuations are not yet approaching extreme levels. There are also opportunities to look at companies outside of the US—for example, Dom cited the very reasonable valuation of Taiwan Semiconductor Manufacturing Co. (TSMC).
There was an enormous amount of detail in this discussion, which can be listened to in full here.