One Knightsbridge Capital Monday Espresso Podcast

Monday Espresso Podcast - 6th July 2026

[00:00:00] Rory: Good morning, today is Monday the 6th July, we're in the back half of the year already. I'm Rory Dowie, Portfolio Manager here at Marlborough. Joining me today once again is Andrew Shaw. As you all know, Andrew's one of the Senior Investment Analysts on the team. So Andrew, good morning

[00:00:14] Andrew: Morning, Rory.

[00:00:15] Rory: Another busy week in markets last week, and at the World Cup for that matter, I need my coffee this morning, that's for sure.

[00:00:21] Rory: Today I'm keen to cover a few things. We had jobs numbers in the US, we saw continued weakness across some of the semiconductor stocks last week. Today, we'll talk a bit about token optimisation, and there was also some news about Meta. But firstly, Andrew, markets last week, the US was closed on Friday. How did we get on?

[00:00:37] Andrew: It was a mixed week across major markets last week, Rory. For us, it was a bit of a story of West versus East. The US, whilst not trading on Friday given the Fourth of July holiday, finished up just under 2%. The UK was up 1.3%, and Europe was also strong, up 2.5%, interestingly, having lagged for most of the year, Europe has now been one of the strongest Western markets over the last month, up over 5%.

[00:01:05] Andrew: If we move to the East, Asian markets lagged. Korea, having been one of the standouts this year, fell 4%, while broader emerging markets fell around 1%.

[00:01:15] Rory: Mixed week then, but solid returns in the Eastern world at least. Let's start with the main economic data last week, Andrew. The jobs numbers in the US, investors keep a close eye here as it signals the health of the economy. What did the print tell us?

[00:01:27] Andrew: The US added just 57,000 jobs in June, which is well below the Dow Jones consensus estimate of 115,000 jobs. Unemployment did dip from 4.3 down to 4.2. It was the softest reading since February's negative print, and prior months were revised lower with May's gain cut to 129,000 jobs down from 172,000 jobs, and Aprils to 148,000 jobs down from 179,000.

[00:01:59] Rory: So weakness across the board there. Interestingly, the market took that print quite well though, didn't it, Andrew? Why was that?

[00:02:05] Andrew: Yeah, normally a softer payrolls print reads as a growth scare and equities fall, but the setup mattered. New Fed chair Kevin Warsh has been leaning towards hiking because inflation was running hot on energy.

[00:02:18] Andrew: So the market's biggest overhang wasn't when do they cut, it was do they hike again?

[00:02:24] Andrew: A weak jobs number in that context doesn't mean rate cuts. It takes the threat of a near-term hike off the table, which is ultimately good for equities, and this was a classic case of bad news is good news.

[00:02:36] Rory: Very clear, Andrew. Thank you. Keen to change tact now. We've spoken a lot about the strong performance in semiconductors this year, ultimately those companies that benefit from the large capital expenditures in the AI build-out.

[00:02:48] Rory: There are a few moving parts last week on this front, which led to the semiconductor index falling 4%.

[00:02:53] Rory: You can kind of think of that index like the S&P 500, but rather than the top 500 companies in the US, it tracks semiconductor stocks. First we had news from Meta, the owners of Facebook and WhatsApp. Starting there, Andrew, what happened?

[00:03:06] Andrew: Meta is developing plans for a cloud infrastructure business that will sell access to AI computing power and models, which is setting up competition with Amazon Web Services, Microsoft Azure, and Google Cloud.

[00:03:18] Andrew: There are two implications of this. Meta has spare compute capacity, if you recall, they wound down their Reality Labs business, and they're putting that to work elsewhere. As I've often said with Meta AI capabilities, unlike Microsoft, where there are clear use cases such as utilising it in Microsoft Excel, for example, I don't really need AI in my WhatsApp.

[00:03:42] Andrew: It's kind of like having the microwave in the garage. It's just not necessary, and it's fine in the kitchen. The other option is that Meta has seen the success of the cloud business across other providers and realising it's where the value is accruing. Meta was up 9% on the news last week.

[00:04:00] Rory: Some interesting developments there from Meta. Moving on, the second main news story was one of token optimisation. Andrew, firstly, can you break down what a token is and why it matters in the context of AI?

[00:04:13] Andrew: Yeah, of course. An AI model doesn't read words, it reads tokens, and a token is roughly a word or a chunk of a word. Every prompt going in and every answer coming out of the LLM is measured and billed in tokens. So tokens are the fundamental unit of stuff in AI, very much the same way as a barrel is the unit for oil

[00:04:37] Rory: Very clear. So I guess what is token optimisation then?

[00:04:41] Andrew: As the word suggests, it is essentially squeezing more value out of each token. The more advanced these AI models become, the more tokens they take up. And that's fine if you've got a heavy duty workload, but for simple tasks, utilising a highly advanced AI model that consumes a lot of tokens and costs a lot more money doesn't really make sense.

[00:05:01] Andrew: I heard a great analogy the other week, and that was, it's kind of like using a blast furnace to boil a kettle.

[00:05:08] Rory: Good analogy. And why does that matter in the context of the AI trade then?

[00:05:12] Andrew: Every AI provider lives on one equation. What they earn per token minus what it costs them to produce a token. They sell tokens, so API calls on subscriptions, and broadly it costs them GPUs or semiconductors and power to make them.

[00:05:29] Rory: We're talking about them because actually if you look at token costs over the last two months, the price of tokens has actually fallen 20%, i.e., they've gotten cheaper. So whilst the number of tokens being consumed is going up exponentially, the pricing side has fallen off quite a bit. And for us, that's an indication of potential margin pressure for the AI labs that produce these models.

[00:05:50] Rory: Then there are obviously those second order implications then of the demand for semiconductors, and for us, this is by far the end of the AI wave, but it's not helping the sentiment on the AI trade at the moment, particularly given the strong run we've seen on the semiconductors over the last few months.

[00:06:03] Rory: And for us, really, that's the main reason why we've seen some softness in that space over the last couple of weeks, and that's one we'll, continue to monitor closely.

[00:06:12] Rory: Just keeping an eye on the time there, Andrew, finishing off, what have our listeners got to look out for this week?

[00:06:17] Andrew: It's looking like a relatively quiet week this week at the moment. Just a few economic data points from around the world. We do get the minutes from the Fed's June meeting on Wednesday and some other macro indicators.

[00:06:30] Rory: It's a slightly quieter week. Thank you for joining me, Andrew. To our listeners, please reach out if you have any questions. I'm wishing you all a wonderful week ahead.