Our view on AI
At top of mind for many investors right now is the enthusiasm and hundreds of billions spent on AI.
- There are elements of speculation today
Some situations in the past where investors eagerly deployed large sums of money into new technologies, such as railroads in the 1800s and fiber in 1999, led to large losses for investors, even when the technologies later proved essential.
Investors are eagerly investing large sums into AI today with the expectation that AI will revolutionize society: from how we learn and write, to the way new drugs will be discovered. Some analysts expect $400 billion in capital spending this year, rising to over $1 trillion by 2030.
Much of this spending is by AI startups funded by venture capital and big tech. Many are training models without clear monetization plans, while rapid advances risk making breakthroughs obsolete within weeks, as seen with DeepSeek in February. OpenAI, for example, generates about $12 billion in revenue but spends more than $100 billion annually on compute, with an additional $350 billion commitment to Oracle. Investors continue to fund such losses in the hope of future payoffs.
- We avoid the most speculative exposures
Given how quickly the situation can change, we avoid companies that are singularly exposed to AI. We never make massive “all-in” bets on a growth trend such as AI, crypto, cannabis, or gold.
As such, we don’t own NVIDIA (to our detriment so far), neoclouds (CoreWeave, Lambda, Nebius), data center equipment providers (Vertiv) or model providers (OpenAI, Anthropic), which either have speculative business models or have revenues that are highly sensitive to drop in AI capex.
- However, we are open-minded
Nonetheless, technological change can be transformative. You would not want to be an investor in Research in Motion or camera makers after the iPhone was released. We want to participate in major trends, but in a prudent way.
Our investment approach is consistent: focus on quality companies from the bottom-up. We look for strong management, durable business models and attractive growth opportunities, and carefully consider the opportunities and risks of AI for each holding.
- Our AI-sensitive names are well-protected
Within our portfolio, 8 of our names are investing in, or are directly impacted by AI. Our goal is to construct a diversified portfolio with industry leaders in different sectors, and we feel comfortable with this level of exposure.
Taiwan Semiconductor Manufacturing Company is the primary manufacturer of AI-related chips and would be harmed by a pullback in AI-related capital spending. Nonetheless, TSM has diverse end markets and has shown an ability to manage capacity prudently to ensure good returns in weak environments.
We own all the hyperscalers (Amazon, Google, Meta, Microsoft), which are investing billions in data centers and AI. Unlike past speculative buildouts, today’s spending is driven by sold-out capacity and strong demand. Margins are rising despite higher depreciation, showing solid returns, and management has signaled capex will be cut if demand weakens.
Furthermore, these companies have multiple paths to generate returns from their AI investment, which reduces the risk of overexpansion including:
- Selling AI capacity to third parties,
- Rolling out AI features within their own business (i.e. Google AI Overviews)
- Using the excess computing capacity for their core business
- Using AI to reduce costs.
Each sits on powerful franchises (i.e. Facebook, YouTube, Office 365) that generate substantial free cash flow even after billions spent on AI. For example, since 2023 Meta has returned nearly as much capital to shareholders as it has invested in data centers.
Finally, valuations remain reasonable given the growth, suggesting the market already discounts some capex risk. Any pullback in spending would only lift free cash flow and capital returns.
Our other tech holdings—Apple, Adobe, and Constellation Software—are exploring ways to integrate AI into their business without heavy capital outlays. While often labeled “AI losers” due to disruption risks, we’ve seen no meaningful business impact yet. Market sentiment could improve if AI enthusiasm cools.