Investing in AI — hype and hope

view original post

I have just had a break to walk the UK’s 100-mile South Downs Way from Winchester to Beachy Head on the south coast. The plan was to chalk off 20 miles a day for the first four days and then ease up. Dream on!

Day three was the killer. My blistering pace resulted in . . . well, blisters. Able only to hobble, I had to abandon the final legs of the trek. I will finish it one day — but at a more prudent pace.

Perhaps not surprisingly, I have been reflecting on how we sometimes let our enthusiasm blind us to the practical realities of life. In the investment world the most recent example of this is artificial intelligence (AI).

So hot has AI become that any chief executive who mentions it in their plans is likely to see a decent pop in share price. It is even hotter than mentioning the metaverse was a couple of years ago — but maybe that is best forgotten.

Is the investment excitement around AI all hype and hope? It is exciting, but it is not new to many investors. Specialist funds and ETFs in this space go back several years — long before the launch of ChatGPT triggered much of the recent hype. Those of us interested in big data have understood for several years that the technology would progress and require processing power. We recognise its potential — and some of its limitations.

Valuation discipline

Even if we agree that AI is a big thing and lots of people are going to do an awful lot of it, investors still have to try to work out which companies might benefit. Perhaps more importantly, are their shares worth buying? There is a difference. 

It is essential to justify the price you pay for a share in terms of the cash that will eventually support it. This requires assumptions about costs, growth, tax and, with technology stocks, dilution from shares issued to “pay” management and staff. Analysts valuing tech stocks often do so based on sales revenue. All these assumptions become strained, however exciting the growth story, when a share trades roughly above five times sales.

During the 2000 tech bubble Sun Microsystems saw its shares run to $64 before crashing to $10. Co-founder Scott McNealy later challenged the valuation discipline of investors who had bought or held his shares at the peak. 

“At 10 times revenues, to give you a 10-year payback, I have to pay you 100 per cent of revenues for 10 straight years in dividends,” he said. “That assumes I have zero cost of goods sold, which is very hard for a computer company. That assumes zero expenses, which is really hard for a company with 39,000 employees . . . assumes I pay no taxes . . . [and] assumes that with zero R&D for the next 10 years I can maintain the current revenue run rate . . . What were you thinking?”

Investors should be wary of “What were you thinking?” stocks in the AI universe. Some might argue that chip designer Nvidia has reached that status. Processing power appears to be the bottleneck in the AI boom, so Nvidia looks well placed to benefit. But the shares have trebled since last September on the AI story.

Today, Nvidia is valued at $773bn and has sales of around $27bn. In other words, the company is priced at nearly 29 times sales. Those sales could grow rapidly with AI, but even if the company maintains its current margins forever, by my maths it would need to increase sales by around eightfold to justify the current share price.

Are there alternatives? Nvidia is now “worth” nearly double the value of the world’s largest chipmaker. TSMC, which we own, builds Nvidia’s chips (among others) and has $2.3tn of sales, so is valued at a price/sales ratio of six. 

And then there is Intel. Those with long memories might remember the early 1990s, when processing power was the limiting factor on what your personal computer could do. So important was your processor that PCs were labelled “Intel inside”. The company’s shares peaked at just below $74 in 2000; they are worth a little less than $30 today. 

Intel makes central processing units (CPUs) — not the graphics processing units that Nvidia sells, which were originally designed for gamers and which, for speed, split tasks into many parallel smaller tasks. CPUs tend to do things in sequence. Speed is sometimes mission-critical (for instance, if you want to mine a bitcoin there is no prize for coming second). But for many tasks, speed and cost of processing will see a trade off. Intel shares trade on 2.2 times sales. I have bought some at that price.

We should consider Google and Microsoft too. Both have been working on their own chip designs for some years. Google is expected to use its own chips (called tensor processing units — TPUs) in its cloud servers from 2025. Microsoft’s Rapid Assured Microelectronics Processor design is already being used in the company’s Azure cloud by the US defence department.

These two companies (we hold both) have another advantage. At its core, what AI does is process data very quickly. But it needs that data, which is good for people who have lots of it. Google, for instance, has all the searches we have ever typed into it. Microsoft has its Bing search engine and is busy collecting data of its own.

We hold another company that might benefit — Accenture, the world’s biggest IT consultant. Its clients are mainly companies whose in-house IT will need support in addressing the opportunities and challenges of AI.

Will AI transform the fortunes of any of these businesses? It should provide a useful tailwind.

Invest with caution

But this is a complex area. In the coming months lots of different visions will be offered by people who are much more capable than me with a screwdriver and soldering kit. The technology will advance — how soon before next-generation graphene chips hit the market, offering 10 times the speed of today’s chips? There will be several ways to play the theme. And it will be just as important to consider which of your shares could be undermined by the new tech — AI may destroy as many businesses as it creates.

It is good to have AI beneficiaries in your portfolio. But you should be careful with valuations — if you are relying on growth for a return, set realistic expectations. Hope is good, but will not overcome all. 

Last week reminded me of that. My sore feet recovered in a couple of days; it can take a lot longer to recover investment losses.

Simon Edelsten is co-manager of the Mid Wynd International Investment Trust and the Artemis Global Select Fund