There is one number making the rounds in financial circles that deserves more attention. OpenAI and Anthropic, the two companies most closely associated with the current artificial intelligence boom, are projected to spend roughly $65 billion this year alone to train and run their AI models. That figure exceeds their combined revenue. It is not a rounding error, and it does not look like a one-time expense. It appears to be the price of staying competitive in one of the more capital-intensive technology developments in recent history.
For investors and financial advisors, that number raises a question that can no longer be ignored: What exactly are markets pricing when they assign near-trillion-dollar valuations to companies that are consuming capital at this pace? Just as important, what does the infrastructure behind those spending levels suggest about the broader opportunity set across the market?
“Everyone’s talking about oil, but I think what the world is mainly short of is tokens.” — Ben Pouladian, engineer and tech investor
A Classic Constraint, Playing Out in Real Time
A recent Wall Street Journal report described a market in which AI demand is outpacing the computing capacity needed to support it. This is not a warning about a distant problem. It is a description of current conditions.
Rental prices for Nvidia’s advanced Blackwell chips have climbed sharply in a short period.
CoreWeave, one of the largest publicly traded AI cloud companies, has also raised prices and now requires longer customer commitments than it did only months ago.
Bank of America analysts, in resuming coverage of CoreWeave with a buy rating, said demand for AI compute services is likely to exceed supply through at least 2029. If that view proves correct, the constraint is not cyclical. It is structural.
That matters for investors. Capacity shortages can affect more than AI software providers. They can shape outcomes across semiconductors, data center infrastructure, electrical equipment, power generation, cooling systems, and the application layer built on top of that stack.
“There’s a massive capacity crunch that’s unlike anything I’ve seen.” — J.J. Kardwell, CEO, Vultr
The strain is already affecting competition. Anthropic, despite rapid revenue growth, has faced enough outages to frustrate some enterprise users. OpenAI, meanwhile, reportedly pulled back from its Sora video product in order to redirect scarce compute resources toward higher-priority enterprise and coding applications. These are not minor setbacks. They are strategic decisions forced by scarcity.
The Economics of Inference
At the center of this story is a concept investors should understand: inference cost.
Training an AI model is expensive. But inference, the ongoing cost of running that model every time a user submits a prompt, may be the more important economic test. It is the recurring expense that will help determine which companies can scale profitably and which remain trapped by their own operating demands.
Today, inference costs account for a large share of revenue at leading AI firms. That means a meaningful portion of every dollar collected must be spent simply to deliver the service. Free users, while valuable for adoption and distribution, still create real costs. Margin profiles may improve over time, but they remain below the levels public software investors have historically rewarded with premium valuations.
OpenAI CFO Sarah Friar described the strain plainly in a recent interview: the company is still searching for available compute and making difficult decisions about what not to pursue because capacity is limited. For investors used to traditional software businesses, that is an important distinction. In this case, the core input behind revenue growth is itself scarce and getting more expensive.
Historical Parallels
The current AI build-out invites comparisons to earlier technology booms, and both are useful.
The 19th-century railroad expansion transformed commerce and created enormous wealth. But many of the best long-term outcomes accrued not to the rail operators themselves, but to the businesses around them: steel producers, landowners, freight operators, and suppliers. The railroads often struggled under the weight of their own capital needs.
The dot-com era offers a different lesson. Demand for internet infrastructure surged, valuations detached from fundamentals, and investors poured money into the promise of future scale.
The build-out created lasting value, but many of the companies most celebrated at the time did not survive long enough to realize it.
The investors who fared best were often the ones who separated durable infrastructure opportunities from speculative application-layer stories.
AI in 2026 is not a replay of either period. The use cases are real. The revenue is real. Enterprise adoption has been unusually fast. But the capital demands are also real, competition is fierce, and the infrastructure bottleneck is not something enthusiasm alone can solve.
What This May Mean for Portfolio Positioning
For investors with exposure to AI-related stocks, whether through direct holdings, sector funds, or broad index allocations, the compute story raises several useful questions.
Infrastructure may present different characteristics and risks compared to the application layer. Companies that supply the essential tools of the AI build-out, such as chip designers, data center landlords, power providers, grid equipment companies, and cooling specialists, may face fewer of the same unit-economics pressures as companies that rely on rented compute they do not control.
Enterprise demand appears more economically important than consumer adoption. Consumer AI products attract attention, but enterprise contracts are often where revenue durability and pricing discipline begin to matter more.
Pricing trends may be worth monitoring. In supply-constrained environments, increases in supplier pricing can reflect underlying scarcity. These dynamics may influence margins for both suppliers and customers, although the impact can vary depending on market conditions and individual company factors.
Finally, investors should pay attention to valuation risk as large private AI companies move toward public listings. Offerings of that size can absorb significant capital and force public markets to confront the gap between compelling growth narratives and operating economics.
Where the Opportunity and Risk Sit
Sixty-five billion dollars is a staggering figure, but in the context of artificial intelligence, it may be closer to a starting point than a ceiling. The scale of infrastructure required to support AI, from data centers to chips to energy, continues to expand rapidly. And if current trends persist, the real constraint isn’t demand. It’s supply.
At West Advisory Group, we’ve seen investment themes like this before, where excitement builds quickly, capital flows aggressively, and the narrative becomes widely accepted. While AI is clearly transformative, we also recognize that not every company participating in that story will benefit equally. That’s why our PEAK FORMula is built around helping clients stay grounded in what they can control, rather than getting swept up in what they can’t.
Focusing on What Matters Most
For investors, the more important question isn’t whether AI will reshape the future; it’s where the economic value will ultimately land. Some companies may thrive, while others may struggle under the weight of rising costs, competition, and shifting market expectations.
This is especially important for those within five years of retirement, transitioning from the accumulation phase into the distribution phase. At this stage, the focus shifts from chasing growth stories to building a strategy that balances opportunity with preservation, income, and long-term sustainability.
Through our PEAK FORMula, we help clients evaluate both direct and indirect exposure to major investment themes like AI, ensuring their portfolios reflect not just the potential upside, but also the risks tied to valuations and concentration.
Our team of experts works to align each portfolio with the client’s broader financial plan, helping them navigate innovation without losing sight of their long-term goals.
Because in the end, successful investing isn’t about predicting the next big thing, it’s about positioning yourself to benefit from opportunity while staying protected from unnecessary risk.
