Ivana Delevska

Right now, seeing a lot of investment going to capital equipment. So this is the most attractive area in the data centre value chain. As you need more memory and chips, it comes with increasing capex requirements. Attention has shifted from tracking the capex of the hyperscalers (who are investing in the chips) to the chip companies (who are investing in incremental capacity).

Absolutely. She likes memory and other components as well, still very attractive. 

However, as we get further into the cycle, we're going to start talking more about capacity additions on the memory side. At some point, once supply seems to be exceeding demand, these stocks will flatten out. 

Not just yet, but it means that the better risk/reward right now is to invest on the capital equipment side -- companies that provide the machinery to build more memory capacity. That cycle typically lasts longer than the memory cycle.

Uses for space such as broadband connectivity and defense have been around for several years. But these markets are pretty small compared to the data centre trade. From a historical perspective, the space race looks a bit too risky to get involved in.

Now we're talking about data centres in space. The technology isn't very different from what satellites are doing today -- using solar power and having compute on board. The idea is much closer to reality than people think. Launch costs need to decline significantly for it to be economical.

The main company building these centres is SpaceX, the leader in space launches as well as operating satellites. Elon Musk has a lot of know-how in AI and chips. It will clearly dominate the space. But don't underestimate other players such as GOOG. We're going to see a lot of activity here. It's not that in 2 years all the data centres will move from Earth to space. With limited supply and limited launch capacity, there's going to be a lot of potential for returns here.

We still haven't seen an AI plan, and so she's pretty negative on the stock. As we go further into the cycle and the data centre drives prices for things like memory, component costs for AAPL will start to increase significantly. That's a risk. Pretty high growth margins means it would still be profitable.

Expectations are very low, so earnings could surprise to the upside. But to get excited long term, need to see progress on positioning in AI.

Attractively positioned in the value chain, as it's on the foundry side producing chips for designers. Benefits no matter who gains traction. Valuation a bit ahead of itself. Better value in other spots across the value chain. Still upside to earnings, but not enough for her to be interested.

Her firm discloses all of its holdings on its website, spear-funds.com.

One of the leaders in quantum computing. Well positioned, as it has an open platform that other companies can build on. The quantum computing trade surprised a bit to the downside last year on commercial contracts. She trimmed. Looking for a better entry point once she sees momentum in earnings from those contracts.
 
Much riskier than other names she talks about.

Key will be how well it executes on new chip that it's introducing. Stock hasn't done that well over past year, so this would be a good entry point, especially if the execution goes well. Doesn't own, as small-cap plays should have more upside in next 2-3 years.

TSM has by far the largest market share, so it's a much safer play. MU is one of 3 leading players in memory, which is very cyclical. Right now, it's closer to the bottom than the top, so still further upside. However, these stocks stop working way ahead of the cycle peaking.

Not an expensive stock, but risky because of the cycle. So you need to watch it closely.

You need to closely monitor all the capacity announcements coming from MU, SK Hynix, and Samsung. Usually after 1-3 years of capex deployment in equipment, that's when you see the market start to stabilize.

Historically a bitcoin miner. Over time, has really built expertise in how to build compute very similar to what data centres need on the AI side. Very volatile, as business is misunderstood by market. Securing long-term contracts with hyperscalers (or CoreWeave), power capacity, and land value. Pretty solid business model. Should still do well as a standalone if CRWV deal falls through. It'll come down to execution. 

She owns APLD instead.

Tricky. A buy, but price is a bit ahead of fundamentals. Driven by news around US government supporting it. Still a ways to go on its turnaround, but encouraging that US government and NVDA are involved. Still trying to optimize manufacturing capacity, while everyone else is expanding. Lots of wood still to chop. 

For her, it always falls short on upside. But she wouldn't be opposed to an investor buying here, especially on the pullback.

In optics, which she's very focused on. Secondary offering hit the stock pretty hard. Pretty well positioned in its market, so it's a Buy here.

Usually her firm goes for a slightly higher market cap. So she owns COHR and CIEN.

The software sector has been treated as a whole, driven by companies with a resilient SaaS business model trading at very high multiples. What's happened over the past 3 years is that people are realizing that a lot of these business models can be replicated with very little effort by using AI. So the sector's come under attack.

The area for investors to avoid (or at least do more thorough due diligence) is that of software applications. Think CRM or HUBS. If the AI solution is equally good, these companies will face competitive threats.

On the flipside, some companies are providing the infrastructure for new entrants to build new tools and applications. This is the part of software that she does like. These tools include cybersecurity, which she finds pretty interesting right now.

There's a real bifurcation between applications and infrastructure. Sticking with infrastructure is the better way to go.

The risk to software applications from AI is why the stock has been underperforming. If you've been paying a lot for subscriptions, you may be able to replace that with a new AI tool and a small team of engineers. That's what has investors concerned.