CIO remains bullish on the prospects for AI, and suggest investors continue to focus on industry fundamentals. (UBS)
AI chipmaker NVIDIA fell sharply after a lukewarm reception to its 2025 product unveiling and keynote presentation at the Consumer Electronics Show (CES) in Las Vegas. Samsung Electronics rallied despite a big miss on its fourth-quarter profit estimates and continued delays to launching its AI-friendly memory. Chat AI startup Anthropic is reportedly raising a new round of private funding at a significantly higher valuation.
Alongside all this, strong US data and worries over the incoming Trump administration’s plans have pushed US yields higher, capping rate-sensitive tech stocks. Concerns over valuations have also come back to the fore, after a powerful tech rally that has delivered annual gains for the S&P 500 of over 20% for two straight years.
Without taking any single-name views, we see several takeaways for investors:
Focus on capital spending, not keynotes. NVIDIA CEO Jensen Huang at CES 2025 revealed new plans to use AI for training robots and autonomous cars, next generation gaming chips, and its first desktop AI computer. Investors responded with a 6% sell-off, the company’s worst single-day decline in four months. Samsung’s rally by contrast came on both a new focus on profitability and after the NVIDIA CEO endorsed its prospects of supplying AI-friendly memory chips. We think it is still the early stages of the AI tech cycle, with strong evidence of continued robust demand for the technology and its supply chain. We see no end in sight to heavy spending on advanced AI chips, and recently raised our estimates for the Big 4's combined capex growth to USD 224bn in 2024 (+51% y/y) and USD 280bn in 2025 (+25%y/y). The prospect of wider AI adoption into other major industries like auto or robotics suggests further upside to capex as more companies outside of large-cap tech build capabilities.
Geopolitics will continue to inject volatility into tech. The Biden administration’s new rules will reportedly seek to control the sale of AI chips “on both a country and company basis,” with a tiered approach that rewards compliance with American standards and security policies. The first tier of the US and 18 allies will have fairly free access but must limit computing power outside tier 1 countries. Tier 2 nations (most other countries) will face limits on overall AI computing power, according to the report, while tier 3 nations (such as China) will face broad prohibitions. So far, President-elect Trump’s geopolitical rhetoric has focused on tariffs and trade instead of sanctions or tech export restrictions. Investors should expect heightened sensitivity to any early Trump policy moves around tech export curbs on advanced chips and chip-making equipment.
AI remains the hottest ticket in venture capital. Anthropic’s latest funding round reportedly raised its valuation from USD 8bn to USD 60bn, against revenue of around USD 875mn, according to media reports. Large funding rounds at high valuations are the new normal for AI startups. AI accounted for nearly half of all US venture capital flow in 2024, according to Pitchbook, and almost 36% of global venture deals. On average, 11% of initial AI companies raising a first round of funding make it to the seventh round, versus only 5% for VC overall. This funding longevity is reflected in their scale, with AI now accounting for roughly 15% of all unicorns, according to CB Insights data. While questions over high valuations for AI startups may be justified, we expect more AI unicorns to be created in the near to medium term.
So we remain bullish on the prospects for AI, and suggest investors continue to focus on industry fundamentals. We think any pockets of near-term volatility, whether on geopolitics or rising yields, may offer a chance to build exposure to quality AI stocks via buy-the-dip and structured strategies. Without taking any single-name views, we continue to like semiconductors and software segments within the thematic AI universe. Memory makers rank third in our pecking order, with signs of a potential bottom in the cycle, followed by semi equipment manufacturers, which have more exposure to tech curbs.
For investors considering VC exposure, we do think early-stage AI may deliver attractive absolute returns, with particular opportunity in natural language interfaces, AI agents, AI core platforms for data training, development and management, and in media content creation. However, venture investors must also be willing to accept a long-term time horizon and have sufficient risk appetite to stomach sizable failure rates. Strong investor demand for early-stage AI exposure also means the terms on offer tend to be less investor-friendly in aggregate.
Main contributors: Solita Marcelli, Mark Haefele, Sundeep Gantori, Delwin Kurnia Limas, Jon Gordon, Matthew Carter, Dean Turner
Original report: How to ride the AI wave in 2025, 9 January 2025.
