December 28, 2025
NOTE: THIS BLOG IS NOT FINANCIAL ADVICE NOR INVESTMENT ADVICE AND DOES NOT REFLECT THE VIEW OR OPINIONS OF MY EMPLOYER
"Technological revolutions are always capital-hungry at the beginning and productivity-rich at the end."
The S&P 500 index finished up ~16% in 2025, with much of this growth driven by the technology sector. The Magnificent 7 (“Mag 7”) -Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta Platforms, and Tesla - represent ~35% of the S&P 500’s total market capitalization. As shown below, on an equal-weighted basis, the index underperformed the cap-weighted S&P 500, reflecting the outsized influence of the technology sector and the Mag 7.
A key driver of this share appreciation has been optimism surrounding the future of artificial intelligence (“AI”). On November 30, 2022, ChatGPT was officially launched to the public. Since then, ~1.8 billion people (~20% of the global population) have used various AI products, including ChatGPT, Canva, Gemini, DeepSeek, Perplexity AI, Grok, Claude, and others.
Since the public launch of these AI tools, users have quickly gravitated toward a few leaders, with the top seven AI tools accounting for ~85% of total market share.
Corporations have also embraced AI, with some already achieving meaningful results across their organizations. As shown below, AI has helped a diverse set of organizations save time, reduce costs, and deliver higher-quality services.
As highlighted above, AI applications are broad in scope. As AI continues to advance, more industries are expected to develop niche AI and automation applications to streamline operations. Time will tell which sectors ultimately benefit most from this paradigm shift.
As AI adoption expands across industries, productivity gains are expected to support global GDP growth. Over the next decade, AI could add ~1.5 - 3.0% to annual productivity growth. Capital expenditure (“CapEx”) spending related to AI has created ripple effects across the economy, including supply chains, productivity enhancements, job creation, and more. Some believe this substantial AI-driven CapEx spending is masking underlying economic weakness and the potential impacts of U.S. tariffs. However, as shown below, AI spending is not an outlier relative to prior investment cycles when measured as a percentage of U.S. GDP.
This impressive innovation and rapid mass adoption have been accompanied by substantial spending on AI. A significant portion of this CapEx has come from large technology companies building out compute capacity through data center expansions (“Hyperscalers”). These hyperscalers—including Microsoft, Google (Alphabet), Meta Platforms, and Amazon—are expected to spend a combined ~US$370 billion on CapEx in 2025.
CapEx spending has increased significantly over the past few years, largely driven by Microsoft, Google (Alphabet), Meta Platforms, and Amazon, which together represent ~30% of the S&P 500’s total CapEx. This spending is expected to continue growing at a rapid pace as advances in AI drive greater usage and, in turn, additional investment. As NVIDIA CEO Jensen Huang noted: “The AIs get better. More people use it. More people use it, it makes more profit, creates more factories, which allows us to create even better AIs, which allows more people to use it. The virtuous cycle of AI has been designed, and this is… the reason why you’re seeing the world’s capex going so fast.”
A typical Google Gemini search (an AI model) uses ~8x more power than a traditional Google search. At scale, this results in significantly increased demand for graphics processing units (“GPUs”), land, and power. Historically, GPUs have been the primary bottleneck for AI growth; however, land and power are expected to become the binding constraints going forward. As Microsoft CEO Satya Nadella stated: “The biggest issue we are now having is not a compute glut, but power—it’s the ability to get the builds done fast enough close to power. If you can’t do that, you may actually have a bunch of chips sitting in inventory that I can’t plug in.”
We remain at the early stages of a multi-year infrastructure build-out, with each new model cycle driving incremental demand for megawatt-scale power. The winners in this race for land and power will ultimately shape the AI ecosystem, as those able to support the infrastructure demands of advanced AI models are likely to attract the largest consumer and B2B customer bases.
The core question that remains is whether AI providers can support this level of spending in a profitable manner. A related question is how long financial markets will be willing to underwrite elevated CapEx to support expanding compute capacity. When assessing the potential returns from AI, investors must consider not only the magnitude of the opportunity but also the timeline over which that potential can reasonably be realized.
The AI industry is progressing toward its long-term objective of Artificial General Intelligence (“AGI”), which would be capable of learning, reasoning, and performing any intellectual task a human can across multiple domains—not just a single, narrow function. Some industry experts expect AGI to be achieved by 2030, although this timeline is highly dependent on the availability of GPUs, land, and power. Even if AGI is achieved, it may not translate into financial success in the near to medium term. For example, HSBC forecasts that OpenAI will generate ~$213 billion in revenue by 2030 but incur nearly ~$500 billion in operating losses. Through 2030, OpenAI is also expected to generate cumulative negative cash flow of ~$282 billion, resulting in an estimated ~$207 billion funding shortfall even after accounting for expected investments.
On October 2, 2025, OpenAI completed a $6.6 billion secondary share sale at a $500 billion valuation. HSBC expects 2026 revenue of ~$35 billion, implying a forward-looking price-to-sales (P/S) ratio of ~14.3x. With rising infrastructure costs, R&D spending, and compute expenses, OpenAI and other AI leaders will need to achieve AGI within the forecasted timeframe to justify further investment. Anything short of AGI would likely result in amplified losses and reduce the capital available to support ongoing CapEx needs.
This also raises questions about how investors will underwrite this risk and how CapEx should be financed. Many would prefer hyperscalers to fund CapEx from operating cash flows rather than through equity dilution or debt issuance. For example, Oracle (NYSE: ORCL) raised US$18 billion through a bond sale in September; since then, Oracle’s share price has declined by ~37% as of December 31, 2025. Additionally, a Financial Times report indicated that private credit lender Blue Owl will not support a US$10 billion investment for Oracle’s next data center. These developments demonstrate that investors are scrutinizing the use of debt to finance AI compute CapEx more closely.
By contrast, other hyperscalers are funding CapEx from operating cash flows. Apollo’s Torsten Slok notes that “hyperscalers are currently spending a record high 60% of their operating cash flow on CapEx.”
In summary, even without considering the potential societal risks, the development of AI represents a paradigm shift with the potential to fundamentally change the world. However, AI growth requires substantial amounts of GPUs, land, power, and CapEx. The combination of high spending and rapidly rising valuations has led many investors to worry that the market may be entering a bubble.
From an investor’s perspective, justifying the associated risks and capital requirements depends on the industry achieving AGI and consolidating market share. The companies best positioned to succeed will be those able to attract investment and secure sufficient land and power to support the necessary build-out. AI is a transformative innovation with inevitable impact; however, it also introduces significant uncertainty and a range of questions that must be analyzed and managed prudently.
NOTE: THIS BLOG IS NOT FINANCIAL ADVICE NOR INVESTMENT ADVICE AND DOES NOT REFLECT THE VIEW OR OPINIONS OF MY EMPLOYER
References
1. Exploding Topics. 65 Most Popular AI Tools Ranked. Anthony Cardillo. Source. December 16, 2025. Accessed December 29, 2025.
2. Exploding Topics. How Many People Use AI?. Anthony Cardillo. Source. October 27, 2025. Accessed December 29, 2025.
3. talentsprint (Accenture). How AI is Cutting Costs & Boosting Business Efficiency?. Source. July 3, 2025. Accessed December 30, 2025.
4. Financial Times. UK civil servants who used AI saved two weeks a year, government study finds. Melissa Heikkilä. Source. June 2, 2025. Accessed December 30, 2025.
5. BMW Group. Artificial intelligence as a quality booster. Source. April, 28, 2025. Accessed December 30, 2025.
6. J.P. Morgan. Is the AI Capex Boom signaling “buy” or “diversify”?. Stephanie Aliaga. Source. February 2, 2025. Accessed December 30, 2025.
7. CNBC. Nvidia CEO Jensen Huang says AI is in a ‘virtuous cycle.’ Here’s what he means. Tasmin Lockwood. Source. October 31, 2025. Accessed December 30, 2025.
8. Google. Measuring the environmental impact of delivering AI at Google Scale. Cooper Elsworth, Keguo Huang, David Patterson, Ian Schneider, Robert Sedivy, Savannah Goodman, Ben Townsend, Parthasarathy Ranganathan, Jeff Dean, Amin Vahdat, Ben Gomes, and James Manyika.
Source.
Accessed December 31, 2025.
9. DataCentre Magazine. Microsoft's Power Problem: AI Chips Are Sitting in Inventory. Megan Baggiony-Taylor. Source. November 07, 2025. Accessed December 31, 2025.
10. TheEconomicTimes. Half-trillion loss? HSBC says OpenAI facing titanic cash burn through 2030. Piyush Shukla. Source. December, 4, 2025. Accessed December 31, 2025.
Comments
Post a Comment