Could AI Be Hitting a Wall?

AI development and usage are growing with lightning speed, but the cost of producing AI is beginning to bump up against customers’ willingness to spend. AI leadership has migrated over the last few years as hyperscalers (Microsoft, Google, OpenAI, etc.) have released new AI models. In a nascent but very large market, increasing market share is the key to staying in the race long term. After a period of experimentation without regard to cost, AI customers are beginning to rationalize their use of the technology. That could result in a collision.

Hyperscaler Costs Continue to Rise

The numerous bottlenecks limiting the growth of data center compute capacity are also increasing costs substantially. Frontier models (the most advanced AI models) are expensive (Nvidia’s revenue is a cost to hyperscalers). The rapid introduction of new, ever-more-demanding chips and physical bottlenecks (power, water, memory, etc.) are adding to these cost issues. AI demand has ramped up without any concern for available compute capacity. As a result, compute capacity has remained constrained and, therefore, expensive.

• Physical Limits: To run advanced AI, you need massive amounts of electricity, heavy-duty cooling systems, and specialized computer chips. These resources are scarce and expensive.
• The Price Shock: As a result of these physical bottlenecks, hyperscalers are facing price shocks. The race for capacity and market share does not allow for delays, which has caused project costs to soar and stretched supply chains to their limits.

Pricing Models Change

In the early stages of AI, flat rates were the norm. That encouraged what the industry calls “tokenmaxxing”—using AI for even mundane tasks without regard to cost. (Tokens are the unit of measure for AI usage). This was clearly unsustainable as hyperscaler costs rose, and we are now seeing a shift to usage-based pricing (i.e., per token).

Customers Begin to Rationalize Token Usage

AI users quickly began burning through a full year’s token budget in as little as three months. High costs are forcing companies to stop wasting AI on random experiments. Instead, they are saving the most powerful AI for their most valuable, complex problems, thereby altering token usage to maximize value per token.

The AI Split

The tech world is splitting into two distinct lanes:

• Heavy-Duty AI: Only a few ultra-rich tech companies will use the most powerful, cutting-edge AI because they are the only ones who can afford the electricity and chip bills.
• Everyday AI: Regular businesses will use smaller, cheaper, and “good enough” AI models to handle daily tasks without breaking the bank.

The Good News

The idea of autonomous AI “agents” taking over everyone’s job all at once is unrealistic. Instead, AI works best as a helpful assistant to human workers. It saves money and time when used for specific tasks, for example:

• Programmers: Using AI to find bugs and write basic code faster.
• Customer Support: Using AI to help human agents look up answers and solve customer issues more quickly.
• Office Workers: Using AI to summarize long documents, translate text, and speed up research.

The Bad News

We may be overbuilding AI infrastructure. Everyday AI does not require top-of-the-line data centers. Hyperscalers, in the race to dominate the market, cannot afford to allow market share to migrate to others.

During an interview on the Acquired Podcast, Meta CEO Mark Zuckerberg was quoted as saying, “If we end up misspending a couple of hundred billion dollars, I think that that is going to be very unfortunate, obviously. But what I’d say is I actually think the risk is higher on the other side.”

The industry bias is to overspend now and figure it out later. This overbuild could be quite large if a significant portion of AI users migrate to smaller, “good enough” models. On the other hand, the physical bottlenecks of data center development will serve as a natural barrier to over-expansion. How it all hashes out is not yet clear.

The Bottom Line

AI could be hitting a wall, but it is not necessarily hitting a wall. AI will still change the world and boost productivity, but it won’t happen instantly or frictionlessly. We should expect progress to be bumpy and, at times, disconcerting.

This environment will require investors to repeatedly adjust their expectations and re-assess market positions. As long-term investors, we are highly cognizant of these short-term trends within AI, which will likely cause hesitation and volatility in the equity market. However, it does not change our long-term view of AI’s growth and its ultimate benefits. Given our investment team’s deep expertise and vision regarding the long-term growth of AI, this transition may well create market opportunities for our clients to purchase shares in the long-term beneficiaries of AI at a better value than is currently available.

Have a great week!

What We’re Reading

• U.S. inflation hits 3-year high, highlighting affordability challenge for Americans

• SpaceX Opens At $150—Surging 20% After Largest IPO Ever

• One-of-a-Kind Handbag Is Made From Lab-Grown ‘T-Rex Leather.

Palumbo Wealth Management (PWM) is a registered investment advisor. Advisory services are only offered to clients or prospective clients where PWM and its representatives are properly licensed or exempt from licensure. For additional information, please visit our website at www.palumbowm.com.

The information provided is for educational and informational purposes only and does not constitute investment advice and it should not be relied on as such. It should not be considered a solicitation to buy or an offer to sell a security. It does not take into account any investor’s particular investment objectives, strategies, tax status, or investment horizon. You should consult your attorney or tax advisor.

The views expressed in this commentary are subject to change based on market and other conditions. These documents may contain certain statements that may be deemed forward‐looking statements. Please note that any such statements are not guarantees of any future performance and actual results or developments may differ materially from those projected. Any projections, market outlooks, or estimates are based upon certain assumptions and should not be construed as indicative of actual events that will occur.

All information has been obtained from sources believed to be dependable, but its accuracy is not guaranteed. There is no representation or warranty as to the current accuracy, reliability, or completeness of, nor liability for, decisions based on such information and it should not be relied on as such.

All information has been obtained from sources believed to be dependable, but its accuracy is not guaranteed. There is no representation or warranty as to the current accuracy, reliability, or completeness of, nor liability for, decisions based on such information, and it should not be relied on as such.

The views expressed in this commentary are subject to change based on the market and other conditions. These documents may contain certain statements that may be deemed forward‐looking statements. Please note that no such statements are guarantees of any future performance, and actual results or developments may differ materially from those projected. Any projections, market outlooks, or estimates are based upon certain assumptions and should not be construed as indicative of actual events that will occur.

Past performance is no guarantee of future returns.

General News

By: Adam