Ivana Delevska

Doesn't see any sort of slowdown in development, especially in hardware investments. Over the past year, dollars have shifted from the big data centres and more toward distributive architecture. We've gone from a model-training era to an agentic era. The big dollars are being spent on inference and using these AI systems. That will continue.

Even if there are regulatory changes on the model side, the application and adoption sides are still in very early innings. That's where she's looking for opportunities.

Important for people to understand that the AI trade is not trading as one thing anymore. You really need to pick and choose the areas you want to be invested in. The fundamentals remain very strong and we're still in the very early innings of adoption.

Doesn't own any at the moment. Reason is because her firm tries to generate alpha, and they usually find it in companies that are between $10-100B in market cap. But her team follows them closely because they set the tone for the rest of the infrastructure spend.

If you look at the 3 biggest hyperscalers today, the best position is probably in GOOG. Doing lots of internal development and investment in its AI models. Gemini is lagging Anthropic and OpenAI, but it's a close third. GOOG is really at the forefront of innovation, especially compared to the other 2 hyperscalers.

AMZN is well-positioned because of its partnership with Anthropic. In third place is MSFT, which really hasn't come up with a differentiated strategy.

Returns for hyperscalers over next cycle will not be as good as prior cycle. Returns to date have been exceptional.

If you look at the 3 biggest hyperscalers today, the best position is probably in GOOG. Doing lots of internal development and investment in its AI models. Gemini is lagging Anthropic and OpenAI, but it's a close third. GOOG is really at the forefront of innovation, especially compared to the other 2 hyperscalers.

She owns no hyperscalers right now.

AMZN is well-positioned because of its partnership with Anthropic. In third place is MSFT, which really hasn't come up with a differentiated strategy. With the selloff in hardware any of those three is at an attractive entry point, with GOOG definitely first, followed by the other two.

She owns no hyperscalers at the moment.

An interesting name to put on your radar, as energy is one of the biggest bottlenecks for data centres. Capacity sold out to 2030, so any earnings surprises will come more from margin improvement. (Doesn't expect a beat on topline.) Definitely a critical provider to the buildout.

Her team is less positive on consumer spending, so they're not in this name. Key for long-term success will be how they embed AI into the iPhone. Not an impressive job thus far. Management changes could result in better innovation on AI -- has potential, but needs to show it.

For her, it's less about the foldable phone and more about a unique AI platform.

Interesting, but risky. If you look at the AI value chain, the hardware and infrastructure layers are where a lot of the value has been accruing. Has shown innovation. Have to see how spending translates into growth. If you own it, hold.

Biggest competitors are SK Hynix and Samsung. Extremely cyclical, and that's the risk. You need to buy at the bottom of the cycle; that's actually when the valuation screens high, because it's related to very low earnings.

Key is to look not at the multiple, but at the earnings trajectory and how many years that will last. She believes we're past the mid-cycle, so it's riskier. She wouldn't oppose using it as a trade.

Cybersecurity leader in the identity space. The whole space looks interesting -- used to target human users, but now broadening to AI agents. Most players will benefit in the short term, and the key will be to find those that benefit for the long run.

Leader in liquid cooling. AI trade has shifted from big data centre spending to more distributive architecture that accommodates AI agents. So the need is different, with servers looking more like traditional ones. Will do well, but better returns elsewhere.

Her firm holds only a few very select companies. While they closely follow about 100 companies, only a few make the cut to be included in the portfolio. These are disclosed on their website.

She does invest from time to time, but not now. Level of improvement seen to date is not meeting the expectations that people had 2 years ago. The downside surprise was that the working quantum computer of today doesn't outperform accelerated computing. Commercial adoption is not there yet. Need to have a really long-term horizon, about 10 years out for broad deployment.

As soon as we have some sort of indication that a computer performs on par, or better than, accelerated computing, that's when you want to invest. You don't necessarily have to wait 10 years.

Lots of risk, as many of the publicly traded stocks may not be the ones that end up winning. One to put on your radar.

(Note the short timeframe.)  Really pulled back over the summer. Attractive entry point. Needs to execute on the optical transceiver side.