World wild markets this week. Initiating coverage on Balwin and Raubex — both offering value. Meta looks cheap, but maybe we’ve all moved on from Facebook. Boxer results really strong — and does that play into the Pick & Pay valuation? Gold miners’ costs to watch. Energy. xAI miles behind the other LLMs.

—

I’m Simon Brown. This is WorldWideMarkets, episode 677 for 13 May, recorded Tuesday mid-afternoon.

WorldWideMarkets, powered by Standard Bank Global Markets Retail, and Shyft — the global money app that puts travel, shopping, payments and investments in the palm of your hand. Enjoy the cheapest forex rates anytime, anywhere. Shyft, powered by Standard Bank.

Thanks to Shyft. Thanks to Standard Bank.

Before we go any further: I had a really great webcast Tuesday morning with 1Invest. We were talking offshore ETFs — there’s a bunch of those — but what we were really digging into was their market maker. He was talking around market making, ETF pricing, liquidity, when is the best time to buy ETFs, and so on. I learned a ton. It was hugely worthwhile. You’ll find it at [justonelap.com/etf-blog](https://www.justonelap.com/etf-blog) — it’s on the home page. Well worth a watch.

—

## Boxer Results

Let’s kick off with numbers. I want to touch on the Boxer results. They were frankly not a bad set of numbers at all. The key thing for me: food inflation was trading at minus 1.2%. That is a giant negative number. They had about 4-and-something percent like-for-like, which gives them roughly 5% real growth. The other point is that their operating profit margin was 5.7%. That is ShopRite-type numbers. The market liked it. The stock was a roughly R55 issue at IPO and has shot up to around R85 — that’s based on Monday’s close.

I don’t think it is necessarily massively expensive. It’s not cheap. Forward PE of 21, which puts it broadly alongside ShopRite — although probably a touch more expensive. Price-to-book is a little chunky. Dividend yield is not much. But those numbers were really, really good.

What are we getting from consensus? A strong sell, a hold, and a buy — with an average target price of R81.46, a low of R65 and a high of R100. That’s certainly saying it has run hard. But here’s the thing: Boxer’s market cap — these are Monday’s close numbers — is R38 billion. That’s a good size.

—

## Pick & Pay: The Valuation That Makes No Sense

Pick & Pay, which owns 65% of Boxer, has a market cap of R15 billion. The short answer is that Pick & Pay should have a market cap of roughly R25 billion based on the Boxer stake alone — which means you’re buying the rest of Pick & Pay for minus R10 billion.

Now, Pick & Pay has had a rough time recently. The chart is decidedly unpleasant. Trading back around the R20s. But it is minus R10 billion implied.

This gets me thinking — should we be looking at Pick & Pay? Average analyst target is R26, highest is R37. Two strong sells, five holds, and only one buy. It is a difficult situation.

I don’t love buying something because I get something else as part of the deal. In essence, you’re buying Pick & Pay and you’re getting Boxer at roughly R55 a share instead of R85. But then, why not just buy Boxer? Well, because you pay R85.

People like John Picard at the Investec Value Fund are all over Pick & Pay. His view is: this just doesn’t make sense. Pick & Pay is not a minus R10 billion business. Even if it goes bust, it is not a minus R10 billion business. It just makes no sense.

I’ll leave it to you. I have no skin in this game — I’m a ShopRite holder and a happy one. But the valuation mismatch is real and worth considering.

—

## Balwin & Raubex: AI-Powered Initiations of Coverage

Two sets of results came out: Balwin and Raubex. You may have seen me writing about using AI for valuations. I’m currently going through the whole process of doing initiations of coverage — which is something that would typically happen at a brokerage. They’d take a new stock, one they’ve never covered before, and produce a 20-, 30-, or 50-page document on it. I’ve done the same on Balwin and on Raubex.

I want to walk through the process — but first, I like the Balwin numbers. I think they were great. I interviewed Steve Brooks on my show on Tuesday morning. My concern is inflation, interest rates, and the petrol price. If it weren’t for those three, the stock is a no-brainer. We’ve got a 12-month target of R455; the stock is around R360. That’s a nice uplift. We’ll probably get our dividend reinstated at some point too. There is definitely something here — I just worry about the broader picture for the consumer.

I’ll put all of these documents on the show notes. Go to [justonelap.com/wwm](https://www.justonelap.com/wwm) and you’ll find them there.

Now, the process: I use Claude AI and the initiation of coverage skill, give it a bunch of instructions, and it goes off and crunches the data and comes back with a report. I then take that report and give it to a second LLM — in this case I used Claude again, but also ChatGPT — and instruct it to take the report, use the source documents (annual financial statements, annual reports, etc.) and whatever it can find on the web, and fact-check the document. I then merge those two fact-checks and go back to Claude and say: here is your document, and here is what my fact-checkers say.

It gets quite interesting. On the initial Balwin document, it got some things wrong. It said Stephen Brooks owned 23% of the company. He owns 33%. It got some numbers wrong. When I challenged it, in some cases it responded: “You’re right, it’s 33%, not 23% — correcting.” In other cases, around valuation, the debate was more nuanced.

There were errors in the hard numbers — we picked those up. All of them? I don’t know. But what I did get, interestingly, was that ChatGPT was impressed with the report. It said: “This is not AI schlock. This is a well-written, well-founded, well-thought-out report.” I’ll take that — it’s not schlock, but it certainly is AI-generated.

What you’re seeing is inconsistency in formatting between the Balwin and Raubex reports — they look quite different from each other. We need to get that right.

A fun observation on Raubex: the fact-checker found some discrepancies, but when we traced them back, an IOL article and a Business Day article were wrong. The data as we had it was correct. So even the fact-checking has to be fact-checked.

For Balwin, I did the full 51-page initiation. For Raubex, I also did a shorter two-pager — again, different formatting. Both are bullish. The Raubex 12-month target is R65.90. The key question for Raubex is how long the war in Ukraine goes on and what that means for their business. But this is not a consumer-facing stress story. The results were good, and there’s an interesting shift towards more private-sector work as well as concrete roads. I didn’t do an interview with Raubex management, so I relied entirely on the results and the report — but the level of depth the 51-page document gave me was genuinely useful.

I have an idea of what I’m going to do with these reports in time — probably put them on a website. There are issues around FSCA licensing and all of that, which we’ll get to. But good reports. Two stocks I like: Raubex and Balwin. Both have something there. Balwin’s challenge is the tough consumer environment, which will hurt.

—

## Gold Miners

We’ve been getting results from the gold miners. Goldfields on Thursday, AngloGold Ashanti on Friday. AngloGold Ashanti paying out tons of dividends and a $2 billion share buyback to return capital. I hold AngloGold Ashanti.

Goldfields saw costs edging higher. There are two components to that. One is energy costs — you’re using diesel and electricity on a mine. Now, to be clear: if you saw energy cost increases in February and you’re reporting to end of February, I get it. If your reporting period runs to end of March, then yes, talk to me about March energy price increases — I’m expecting those. We keep an eye on oil; we’re still looking at all-in sustaining costs of $2,000 and lower, with a gold price of $4,500. So if all-in sustaining costs go up 10% from $1,800 to $1,980 — yes, it takes some shine off, but you’re still making over $2,500 an ounce in clear profit. And remember, the gold price is well up on where it was a year ago.

What I’m also watching is other cost increases. I remember chatting with Chantal Marks from FNB last year and asking her: what worries you about gold miners? She said: lots of acquisitions, and fatalities moving higher. Fatalities rising shows carelessness, and carelessness costs money. Those are the things to watch.

On the gold price: $4,500 is your big number. A break below that takes us down to maybe $4,200 or $4,000. That’s your gold price for now — and it’s holding steady.

—

## Brent, Platinum & Palladium

Brent Crude is looking ugly. Around $104 at time of recording, with a live price tracking lower. Trump can’t get his peace deal. That is just the problem — no peace deal is hurting everyone and everything.

Platinum and palladium: consolidating between R22,000 and R22,000 on platinum, and palladium making a recovery. Both actually doing okay — not shooting the lights out, but R1,500 is an important level for palladium, with R1,400 as the next key support. Sibanye-Stillwater remains my preferred play here.

—

## Meta: The Cheapest It’s Been Since 2022

I want to look at Meta. The market didn’t much like the results from a week or two ago, sold it off quite aggressively. It’s trading at just under $600.

Let’s look at what we’ve got for price targets: seven holds, 48 buys, eight strong buys, no sells whatsoever. Low analyst target is $615, average $825, high is $1,015. The market says it’s cheap. And if we take an overview: forward PE is 18-and-something. Over 10 years, the one standard deviation below the mean is 19.8. This stock has not been this cheap since the tech sell-off in 2022.

So what is Meta? They spent a fortune on Reality Labs — their VR business. That’s now being folded into AI. Llama, their open-source LLM, was at one point really competitive. It’s now fallen behind. Zuckerberg has reportedly spent hundreds of millions, if not billions, acquiring AI talent, and by reports their latest model is fairly good. I haven’t tested it, so I can’t comment directly.

They’ve also got WhatsApp, Instagram, and Facebook. And yes, you’ll tell me that Facebook is dead, that only boomers use it, that kids are fleeing Instagram for TikTok. Both of those statements are true. Except they’re also not — and that’s just it.

Meta is trading below the PE of the broader market. That just makes no sense. Facebook may be mature and skewing older, but the boomers are using it and they’re clicking on the adverts. Yes, I appreciate that when Meta discovers scammers using their advertising platform, they don’t block them — they identify them as scammers and charge them a premium, because scammers will pay a premium. I appreciate that their glasses tech, which records ambient surroundings, raised ethical concerns. I appreciate the Instagram/TikTok pressure. Yes, yes and yes.

Instagram and WhatsApp are absolutely huge. This is just a cheap stock. If you’ve ever wanted Meta, now is your chance. I think there’s significant upside from where it’s trading — it’s cheaper than the broader market, it’s growing faster, it has better tech, and the valuation leaves meaningful room to the upside.

—

## xAI: 100 Miles Behind

I want to touch on xAI. A really interesting data point came across my desk — courtesy of Keith McLachlan, so thanks to Keith. It’s the Open Router token usage data from MacroMicro, showing the distribution and total volume of token usage by model provider on the Open Router platform.

Here are the Monday numbers, in billions of tokens: Anthropic is the biggest at 615 billion. Google at 593 billion. OpenAI at 373 billion. And then xAI at 95 billion.

In other words, xAI is running at roughly a quarter of OpenAI’s volume — and OpenAI is the smallest of the top three.

There were rumours swirling around the SpaceX IPO that Elon Musk was telling investment banks: if you want to work on this deal, you need to have an xAI account. And the banks were reportedly saying: sure, $20 or $200 — whatever tier you want — we’ll make tens or hundreds of millions in fees, we’ll take your $20. But he’s just not getting organic traction.

There’s another data point: SpaceX is selling data centre capacity to Anthropic. Now, xAI sits within the broader SpaceX/X ecosystem. If they were selling spare capacity to a competitor, it means they have more capacity than demand. Which means xAI is simply not getting the usage.

I’m on X. I was given a blue tick, so I have access to Grok. I have used it two or three times and it was nothing special. I also have an objection to its ability to generate non-consensual imagery. So I don’t use it. The few times I did, it was unremarkable.

The three that matter are Anthropic, Google, and OpenAI — which is effectively Microsoft. xAI is 100 miles behind. And the data on Open Router says the same thing.

—

## Upcoming Event: How to Spot the Next Capitec

Remember, end of the month we’ve got our Power Your Money event. We’re looking at Capitec — specifically: how do you spot the next Capitec? I suspect the answer is mostly about process and disruption. I’ve read a couple of books and I’m putting together the presentation. I think it’s 28 or 29 May — webcast or in person. Go to [justonelap.com/events](https://www.justonelap.com/events) for more information and to book. Come along, it’s going to be fun.

The key point, of course, is that even if we find the next Capitec, it takes time. The Capitec story is well over 20 years old.

—

WorldWideMarkets, powered by Standard Bank Global Markets Retail and Shyft — the global money app that puts travel, shopping, payments and investments in the palm of your hand. Enjoy the cheapest forex rates anytime, anywhere. Shyft, powered by Standard Bank.

Thanks to Shyft. Thanks to Standard Bank. Remember: Shyft is doing the migration. If you’re still on WebTrader, get onto Shyft — you should have received the email. The last cohorts, of which I’ll be part, are going across soon. Ultimately WebTrader shuts down and everyone moves to Shyft.

We’ll leave this week there. My name is Simon. We’ll be back again next week. Until then, look after yourself — and if you can, look after somebody else as well. Cheers, all.