After a strong 2024, the public investing narrative around AI has turned negative.

2024 was a fruitful year for the AI trade. Deepwater’s concentrated basket of AI companies (see below) returned an average of 70%. For comparison, the Nasdaq—driven in large part by the Mag 7 and their AI exposure—finished the year up 32%.

In 2025, the momentum has reversed. It began with Deepseek’s sharp selloff in late January, followed by rising concerns that tariffs could push the economy into a recession. Historically, recessions have been the primary trigger for transformative tech market sell offs, dating back to the “Tronics Boom” of 1959–1962. So far this year, our AI basket is down 25%, compared to a 16% decline in the Nasdaq.

​Recent AI stock performances may suggest concerns about the industry’s long-term fundamentals. However, recent updates from leading AI companies indicate that demand for AI technologies is surpassing industry expectations.​

Recent Developments:

Microsoft (April 9):

Noelle Walsh, President of Microsoft’s Cloud Operations, stated in a LinkedIn post that:​

“This year, Microsoft is on track to spend more than $80B to continue building out our datacenter infrastructure and these investments are informed by near-term and long-term demand signals.” ​

This $80B investment aligns with prior expectations, indicating that Microsoft is reiterating their capital expenditure plans for the calendar year 2025. One comment that had a cautious tone was Walsh noted that Microsoft would be “slowing or pausing some early-stage projects,” a move anticipated to impact capital expenditures in CY26. While directionally this is negative, it’s not new. On the company’s December earnings call, the company noted that CapEx growth rates would slow next year.​

Google (April 8):

At Google Cloud Next 2025, CEO Sundar Pichai reaffirmed the company’s plan to invest approximately $75B in capital expenditures for the year, consistent with expectations set three months prior. In late 2024, investors might have expected this number to increase. Given the current more cautious sentiment on AI, we view this reiteration as a positive. Overall, Google is expected to increase spending by about 40% in CY25 over CY24. Additionally, Google Cloud CEO Thomas Kurian highlighted that there are now 4 million developers using Gemini, and there has been a 20-fold increase in the use of Vertex AI, Google’s machine learning development platform, over the past year. ​

Shopify (April 8):

In an employee memo, CEO Tobi Lutke informed employees that they must demonstrate that tasks cannot be accomplished by AI before requesting additional headcount and resources. Lutke added that there is a “fundamental expectation” for employees to incorporate AI into their daily work. ​

OpenAI (April 1):

CEO Sam Altman posted on X:​

“we are getting things under control, but you should expect new releases from openai to be delayed, stuff to break, and for service to sometimes be slow as we deal with capacity challenges.” and “working as fast we can to really get stuff humming; if anyone has GPU capacity in 100k chunks we can get asap please call!” ​

Altman’s comments are consistent with recent outages experienced with GPT. Separately, six weeks ago, the company announced they now have 400m weekly users, up 33% from early December when they announced that figure had reached 300m. This data point is positive for two reasons:​

1. It demonstrates the rapid increase in AI usage for a relatively simple chatbot. Imagine the potential usage when AI achieves general intelligence.​

2. While the 400m figure is impressive, it is still small compared to its potential. It is estimated that Google is used by about 2.5B people daily, and the use case for AI is believed to be greater than search. If we assume AI could reach 3.5B daily users, this implies the market can grow by more than eightfold in the coming years.​

Additionally, this month, OpenAI closed a $40B investment round with a post-money valuation of $300 billion, up from the previous October 2024 round, which valued the company at $157B post-money. ​

2025 has begun with challenges to our prediction of 2–3 more years of a bull market driven by AI innovation. The Nasdaq is down 16% year-to-date. To achieve an above-average return for 2025—considering the Nasdaq average annual return of approximately 10.9% over the past 20 years—the index would need to gain about 35% for the remainder of the year. While this is unlikely, I anticipate the Nasdaq will make a run from here and outperform its 10.9% average for the remaining months of 2025.​

My optimism stems from a strong belief in AI’s potential and historical market patterns. Notably, I expect AI-first companies to experience higher growth rates than currently anticipated, as infrastructure development extends into late 2026. Beyond 2026, we may witness a wave of AI-first companies going public, alongside an acceleration of AI software firms powered by AI agents. This trend should benefit the broader market, as companies begin to realize cost savings and margin improvements with the ramp of these tools starting in 2027.​

Historically, market pullbacks like we’ve seen this year are both expected and healthy during a bull market. For instance, from 1994 to 2000, the Nasdaq experienced multiple drawdowns of 10% or more as it climbed to the peak of the dot-com bubble. So far, the AI trade has seen two such pullbacks. These periods of skepticism allow for market appreciation of outperformance. In other words, if everyone were a believer in the power of AI, it would be difficult for these companies to exceed expectations—a dynamic essential to supporting a bull market.​

Disclaimer