Investing in the Bottlenecks

June 2026 Client Letter

Understanding the infrastructure beneath AI

It’s not often that my wife and I disagree. After 25 years of marriage, we tend to be on the same page about most things.

A few months ago, that changed. Over a driverless car.

We visited the University of Miami with our son, Jack. Before his campus tour, we spent part of the morning walking through nearby Coconut Grove. As we stood at a crosswalk, a car pulled up beside us with no one behind the wheel.

It was a Waymo.

Driverless cars had always felt somewhat futuristic to me, something discussed on earnings calls, technology podcasts, and the news. Seeing one in person made the future feel less distant.

At lunch I downloaded the Waymo app and suggested we take one back to the hotel. Jack and I liked the idea. Allison did not. In fact, she suggested that Jack and I take the Waymo while she take an Uber.

That reaction caught me slightly off guard. Between the two of us, Allison is usually the more technology-oriented. But the idea of sitting in a car with no driver was a step too far.

Unfortunately, the service was still expanding in the area, and no cars were available nearby.

For me, the appeal was not simply transportation. It was the experience of seeing a major technology move from concept toward everyday life.

Over the past several years, much of the investment conversation has centered on data centers, semiconductors, and power demand. Yet these subjects are often discussed in abstract terms that are measured in capital spending, chips shipped, or electricity consumed.

A driverless car is different.

It is an example of what AI infrastructure looks like when it leaves the data center and enters the real world.

Before artificial intelligence can answer a question, drive a car, write code, analyze medical images, or help a business make a decision, the physical foundation has to be in place. The data centers must be built. The chips must be manufactured. The power must be delivered. The cooling systems must work. The fiber networks must connect everything together.

In other words, before a single “bit” of AI output reaches the user, an enormous amount of infrastructure has already been put to work behind the scenes.

Like every major technological shift before it, artificial intelligence is beginning to encounter bottlenecks. Railroads required steel. Electricity required transmission networks. Cloud computing required data centers. Artificial intelligence appears to be in a similar phase.

Increasingly, the story is becoming less about software alone and more about the infrastructure required to support it.

Waymo itself is part of Alphabet, Google’s parent company. Understanding Alphabet helps reveal how much infrastructure sits beneath the surface of artificial intelligence.

Google and the AI Infrastructure Stack

As AI moves beyond software and into the physical world, Alphabet increasingly appears less like a traditional internet company and more like a diversified infrastructure platform.

Many still primarily associate Google with internet search and digital advertising. Yet, over time the company has expanded far beyond those beginnings.

In 2015 Google reorganized under the Alphabet structure, separating the business into three primary segments: Google Services, Google Cloud, and Other Bets.

Google Services includes platforms that billions of people use every day, including Search, YouTube, Maps, Android, Chrome, and Google Play. These products are also becoming distribution channels for artificial intelligence tools.

Google Cloud represents another important layer. Cloud computing today is no longer simply about digital storage or hosting websites. Cloud platforms function as the computing backbone of the modern economy, supporting AI models, enterprise software, and enormous amounts of data processing.

Then there is Other Bets, Alphabet’s collection of longer-range initiatives, including Waymo. While these projects may still appear experimental, they offer a glimpse into how artificial intelligence is moving beyond software and into transportation, robotics, healthcare, infrastructure, and other real-world systems.

Taken together, the structure highlights how Alphabet participates across multiple layers of the AI ecosystem simultaneously.

The company develops artificial intelligence models through Google DeepMind, designs custom AI semiconductors, operates one of the world’s largest cloud-computing platforms, builds large-scale data centers, and invests heavily in networking and computing infrastructure.

Modern artificial intelligence may look like software on the surface, but it depends on a very physical foundation. Behind each AI search, model, or autonomous driving decision are data centers filled with servers, advanced cooling systems, fiber networks, substations, and large amounts of electricity running continuously in the background.

Google’s DeepMind division has explored how artificial intelligence may help utilities and grid operators manage increasingly complex electrical networks. In other words, the same technology that is creating new demands on the power grid may also become part of the solution for managing that grid more efficiently.

There is an interesting irony in that. Artificial intelligence is placing new strain on the infrastructure beneath the digital economy, but, over time, AI may also help direct how that infrastructure is planned, monitored, and operated.

Micron and the Memory Bottleneck

One of the emerging lessons from artificial intelligence is that computing power alone is not enough.

Much of the attention surrounding AI has focused on processors and the enormous computational power required to train and operate modern models. Yet as these systems become larger and more complex, another constraint has begun attracting attention: memory.

Memory is becoming one of the critical bottlenecks within artificial intelligence infrastructure.

That backdrop helps explain the growing importance of companies such as Micron Technology.

Artificial intelligence models continuously move, store, retrieve, and process enormous amounts of information. The faster that information can be accessed and transferred, the more efficiently AI systems can operate.

Micron is one of the world’s largest producers of memory and storage semiconductors, including DRAM and NAND flash memory. While these technologies have existed for decades inside computers, smartphones, and servers, artificial intelligence is creating a new level of demand for high-performance memory systems capable of supporting increasingly data-intensive workloads.

A useful analogy is to think of modern AI processors as high-performance engines. Even the most powerful engine cannot perform at its full potential if fuel cannot reach it quickly enough. In a similar way, today’s advanced AI processors depend on enormous amounts of memory capable of delivering data at extremely high speeds.

Without sufficient memory capacity and bandwidth, processors spend valuable time waiting for information rather than processing it.

Micron has indicated that demand for advanced memory products used in artificial intelligence servers continues to exceed available supply in certain segments of the market. Meeting that demand requires years of planning and billions of dollars of investment in fabrication facilities and manufacturing capacity.

Historically, memory companies have been viewed as highly cyclical businesses, experiencing periods of oversupply followed by shortages and pricing swings. That cyclicality may not disappear. However, artificial intelligence appears to be changing both the scale and complexity of memory demand in ways that investors are trying to better understand.

The story of artificial intelligence may not simply be about who builds the fastest processor. It may also be about who helps solve the bottlenecks limiting the performance of the entire system.

Companies like Micron represent another example of the unseen infrastructure operating beneath the surface of the modern digital economy. If artificial intelligence is entering an era defined by bottlenecks, memory may prove to be one of the most important.

Key Takeaway

Artificial intelligence may be the headline, but many of the investment opportunities may lie in the infrastructure, memory, power systems, and other bottlenecks required to support it.

Market Backdrop: Two Sets of Economic Signals

The market continues to send a message that can feel contradictory.

Clients turn on the television and see headlines about conflict in the Middle East, inflation, higher gasoline prices, and concerns about consumer spending. Then they view their investment accounts and see portfolios that in many cases have continued to advance.

One possible explanation comes from the Conference Board’s May Consumer Confidence Survey. The report showed consumers growing somewhat more cautious about business conditions and employment prospects.

Yet buried inside the same survey was a surprising finding: 54.8% of respondents said they expect stock prices to be higher a year from now, one of the highest readings on record.

That contrast highlights an important point: the economy and the stock market are not always driven by the same forces.

Consumers tend to focus on mortgage rates, food prices, insurance costs, gasoline prices, and job security. Markets, meanwhile, often focus on earnings, capital investment, productivity, and long-term growth opportunities.

One example is the enormous investment currently occurring around artificial intelligence infrastructure. Hundreds of billions of dollars are being committed toward data centers, networking equipment, memory chips, power systems, cooling, and other technologies supporting growing computing demand.

This may help explain why markets can sometimes advance even when headlines feel negative. Markets are forward-looking and often focus on where earnings, investment, and productivity may be headed rather than where they stand today.

That does not mean risks have disappeared. Geopolitical tensions, inflation concerns, interest-rate policy, and economic slowdowns still matter. However, markets also appear influenced by long-term investment trends that may not always align with consumer sentiment or daily headlines.

At the moment, we appear to be living through one of those periods where economic concerns and investment opportunities are telling different stories.

Bringing It Together

Technology continues to move from concept to reality at a quick pace. Yet as technology becomes more powerful, it also creates new challenges that require vigilance.

One area where we continue to see this is fraud.

Fraud attempts have become more sophisticated. Emails can appear to come from a spouse or friend, a bank, a brokerage firm, a government agency, or a familiar company. Websites can closely resemble legitimate login pages. Phone calls may appear to originate from local organizations, law enforcement agencies, or financial institutions.

Recently, I received an email that appeared to be from my wife asking me to click a link to view photos. At first glance the message looked legitimate; however, the wording felt unusual. When I clicked on the sender’s name, the actual email address was clearly not hers.

In another case a client received a phone call from individuals claiming to represent the Collier County Sheriff’s Office. The callers used the name of an actual deputy and sounded highly professional. They alleged that unpaid fines existed and attempted to obtain payment information over the phone.

These situations serve as an important reminder that appearances alone can no longer be trusted.

A few simple habits may help reduce risk:

• Be cautious about clicking links contained in emails or text messages.
• When possible, visit websites directly rather than using embedded links.
• Hover over to verify the sender’s actual email address, not just the displayed name.
• Be skeptical of unexpected requests involving passwords, account access, wire transfers, gift cards, or urgent payments.
• Consider enabling multi-factor authentication whenever available.
• Most importantly, if you receive a communication involving your Fidelity account, passwords, account security, money movement, or any request that seems unusual, consider contacting our office before taking action. We would much rather spend a few minutes reviewing a suspicious email or phone call than have a client become the victim of a scam. A Fidelity representative will never ask you to provide them with your password nor ask you to share an account access code.

Fraud attempts can fail when people slow down, ask questions, and independently verify information before responding.

As always, we welcome the opportunity to discuss your specific situation. If you have questions or if anything in your financial life has changed, please call us at (800) 843-7273. We are always glad to have that conversation with you.

Warm Regards,

 

 

 

Matthew A. Young
President and Chief Financial Officer