The Sale We Have Been Preparing For
This past week I received calls from three clients—not panicking, but close—because of the sharp decline in technology stocks. What surprised me was how quickly they seemed to forget what I’d been writing and discussing for months: A real correction was coming, and we needed to prepare before it arrived. To me, this July plunge in technology stocks is the selloff I have been waiting for. I don’t know the exact bottom but I still believe this decline could be swift and steep before it runs its course similar to what we experienced last year.
A sharp correction does not end the long-term opportunity. During my 37 years managing stocks, I have seen this pattern repeatedly. A powerful long-term advance is interrupted by a sharp decline, fear makes the drop feel permanent, and the recovery begins before most investors feel comfortable enough to believe it. Most recently, technology stocks experienced a steep, roughly seven-week decline between February and April 2025. The market was reacting to the Chinese DeepSeek news, tariffs, recession fears and rapidly changing headlines. Many of the hardest-hit technology leaders then rebounded sharply during the following weeks. While we cannot know the exact timetable of this current decline, history does provide some perspective. Markets can fall much faster than seems reasonable—and recover before the headlines become reassuring again.
The Boy Who Cried Wolf
Cash may feel frustrating while stocks are rising, but it provides choices when other investors are frightened into selling. In my September 2025 update, The Boy Who Cried Wolf, I wrote: “I think there could be a swift sell-off like March (2025)…. So, it may be prudent to take some gains from the recent strong rally. This does not mean sell your winners blindly, rather like cutting off unfruitful branches on an apple tree and pruning those that are fruitful to make the whole (portfolio) more fruitful.”
We raise cash to go shopping when the next sale arrives. For the past few months, I’ve advocated trimming and pruning so cash could accumulate. In Miles and Miles and Miles, I urged readers to remember the enormous scale of the AI revolution, using the proposed data-center buildout as just one example: “When the next big sell-off comes—a sale—remember this idea of so many proposed data-center projects.” I wrote those words while markets were strong because that is when investors must prepare financially and emotionally for the day when markets are not strong. That day may now be here. The September update is particularly relevant because it also discussed the panic created by DeepSeek in early 2025. Investors suddenly feared that a lower-cost Chinese AI model could undermine America’s AI leaders and invalidate the need for enormous spending on chips and data centers. The recent Kimi K3 announcement has revived many of those concerns. Markets seemed to have moved from: “A formidable competitor has arrived” to: “The entire AI investment cycle is over.”
Those are not the same conclusions. Better and less expensive models may accelerate AI adoption by making it available to more businesses and people. That will create different winners and losers, but doesn’t eliminate the need for computing, memory, networking, electricity, cooling, cybersecurity and data infrastructure.
Phase One to Phase Two
Phase One of the AI revolution focused largely on building and training AI models and delivering answers directly to users through systems such as ChatGPT or Claude. We are now transitioning into Phase Two: reasoning, inference and AI agents.
Instead of merely answering a question, an AI system may now reason through a complicated problem, review multiple sources, operate software and complete an entire project. That requires far more computing than producing a simple chatbot response. This is why the advertised price per AI “token” can be misleading. One model may charge less per token but require far more tokens to produce the correct answer—like a car using cheaper fuel but burning twice as much to travel the same distance. What matters is the total cost of producing a correct answer or completing a project, not simply the price of each token. That cost depends on how efficiently a model uses processors, memory, networking and electricity—and how many steps it needs to produce the desired result. As useful intelligence becomes less expensive, businesses are likely to discover many more ways to use it. Therefore, falling AI costs do not necessarily mean less infrastructure spending. They may lead to an enormous increase in the amount of AI activity taking place.
Phase One built the intelligence. Phase Two puts that intelligence to work.
Investing Where the Puck Is Going
Periods of transition are rarely calm or orderly; they are often tumultuous. That is why our portfolios have been purposefully constructed to target Phase Two rather than merely owning companies that benefited from training the original models. We are attempting to invest where the puck is going—not simply where it has already been. That includes companies positioned to benefit from inference, networking, memory, data infrastructure, electrical power, cooling, cybersecurity and software capable of turning AI into productive business outcomes. We are building portfolios from complementary parts rather than owning several companies exposed to the same narrow segment of technology.
Do Not Mistake the Trees for the Forest
The scale of capital spending helps place this opportunity into perspective. Amazon, Microsoft, Alphabet and Meta spent approximately $217 billion on capital expenditures in 2024 and roughly $358 billion in 2025. Their current plans call for approximately $725 billion in 2026 alone—roughly equal to everything those four companies spent cumulatively during the entire nine-year period from 2015 through 2023. Current analyst estimates place their combined 2027 spending near $950 billion, although that figure will change as the companies update their plans.
Capital expenditures are companywide and are not exclusively AI spending. Historical figures are based on reported cash capital expenditures; current guidance and analyst estimates may use somewhat different definitions. Sources: Company filings and current guidance as of July 26, 2026; 2027 represents analyst estimates.
Not every dollar will produce an adequate return, and the benefits will not be distributed equally. But these companies’ own spending plans make clear that the infrastructure buildout is not disappearing because technology stocks experienced a difficult month. The trees are what we see directly in front of us: Falling prices. Red account statements. Negative headlines. New Chinese models. Questions about excessive spending and whether yesterday’s market leaders can continue growing. Those concerns are real. Some companies will disappoint. Some valuations became excessive. Some businesses that led Phase One may not benefit as much from Phase Two. But the forest is the much larger transformation taking place. Businesses are not going to stop pursuing automation, productivity and lower operating costs because technology stocks declined. Demand for useful intelligence does not disappear because investors temporarily become frightened.
Individual trees may become damaged. Some may even fall. That does not mean the forest has disappeared.
What We Are Doing Now
I’m not going to rush all available cash into the market on one day, nor will we pretend that we can identify the exact bottom. I will deploy capital deliberately and selectively, continuing to evaluate earnings forecasts, valuations, competitive positions, technical trends and how each investment complements the rest of the portfolio. Being prepared for a correction does not make experiencing one pleasant. But preparation helps prevent fear from controlling the decision-making process. I believe this technology selloff is the kind of opportunity I have been preparing for and writing about for months. It may become more uncomfortable before it ends, and no one should assume a recovery must occur immediately. But my long-term view has not changed. I continue to believe we are in the early years of an AI revolution that could create more wealth for investors than the internet revolution did.
This is the moment to remember the forest—and why we prepared before the trees began shaking.
Image created using AI-assisted digital artwork directed by Clint Gharib. Opinions expressed are that of the author and are not endorsed by the named broker/dealer or its affiliates. All information herein has been prepared solely for informational purposes and should not be considered legal or tax advice. It is not an offer to buy, sell, or a solicitation of an offer to buy or sell any security or instrument to participate in any particular trading strategy and is not intended to provide, & should not be relied on for, tax, legal or accounting advice. You should consult your own tax professional regarding your specific situation. Past performance does not guarantee future results. Certain statements contained within are forward-looking statements including, but not limited to, statements that are predictions of future events, trends, plans or objectives. Undue reliance should not be placed on such statements because, by their nature, they are subject to known and unknown risks and uncertainties. Oxford Retirement Advisors is an independent firm with Securities and Advisory services offered through Madison Avenue Securities, LLC (“MAS”), member FINRA/SIPC and a Registered Investment Advisor. Oxford Retirement Advisors and MAS are not affiliated entities