Chip stocks have been falling in recent days and the market has begun pricing in a new risk around a possible slowdown in AI development. However, no real limits have arrived yet, hyperscalers are not cutting investments, and fundamentally nothing major is changing. Most of the movement so far is mainly fear and uncertainty.

So let's look at it practically: which companies would such a scenario hit the hardest, who could conversely benefit, and what makes sense to balance a portfolio with today.

1. I see the biggest risk in hardware, but not equal for all companies. If training of the most advanced models slows down, there will be no need to expand computing infrastructure so aggressively. Therefore, manufacturers of AI accelerators and memory are most at risk, namely Nvidia, AMD, Broadcom, Micron, SK Hynix and the like. The next layer is Vertiv and data infrastructure, because less computing power means less need for electricity, cooling and other infrastructure. And then also companies like Lam Research, Applied Materials, KLA, ASML...The market would probably punish them immediately, but the fundamental impact comes with a delay - first the hyperscaler has to cut capex, then the chip maker reduces planned capacity, and only then do orders for new manufacturing equipment decline.

But again, no capex decline has come yet nor any definitive decisions, so all the drops are still just from fear.

2. On the opposite side, I see software and especially cybersecurity. These companies paradoxically have a very good position in both scenarios. If AI continues extremely fast, the amount of automated attacks grows, the need for identity security, endpoint protection, clouds and AI agents. If development slows down precisely because of security problems, companies will also be forced to spend more on governance and security. That's why CrowdStrike jumped about 14%, Palo Alto about 13% and Rubrik also about 15%.

But it's good to mention that all of them are quite expensive today with high multiples and I definitely wouldn't chase them personally right now.

With classic software, the point is a bit different. Slower progress buys them time. They can embed AI into existing workflows, monetize it on their customer base, and they don't have to fight every few months with another huge leap of frontier models. Even Reuters already warned before this sell-off that sentiment around software is starting to shift from "AI will destroy them" to "they can monetize AI".

3. And then there are companies that are almost completely unrelated to AI. Those can make sense today as a counterweight. If I wanted to really stabilize a portfolio, from what I'm watching today, I like McDonald's the most. The stock is about 25% below its 52-week high and forward P/E has fallen to around 19×, compared to more than 23–25x where the company has traded in recent years. I get a very strong brand, franchising model, relatively predictable cash flow and a business that basically doesn't care how many GPUs Anthropic needs. VICI is the second interesting counterweight. At a price around $25, it is about 26% below its 52-week high, dividend yield is above 7% and P/FFO about 10x. Moreover, it has long-term lease contracts and a completely different revenue driver than tech companies. On the other hand, higher interest rates are a headwind for it.

There are also Booking or Uber, which are at interesting levels and I see them as cheaper growth diversification outside AI infrastructure. They are not outright defensive companies like McDonald's or VICI, so in a recession they would probably also take a hit. But their results stand on completely different drivers than the whole AI cycle.

So personally I'm definitely not selling anything, but I will watch it, especially the capex outlooks of Microsoft, Amazon, Meta and Google. If they started cutting it, that would be a fundamental change in the investment thesis.