Posted on 08/28/2026 8:31:11 PM PDT by SeekAndFind
Artificial intelligence spending is racing toward levels once considered unimaginable. Yet 250 years of market history suggest the boom may have considerably further to run before financing pressure, higher rates or an outside shock finally brings it down.
The artificial intelligence boom began as a technology story. It has evolved into one of the largest capital investment cycles in modern history.
Since 2024, companies have poured hundreds of billions of dollars into semiconductors, power infrastructure, data-center construction and networking equipment. Amazon, Alphabet, Meta Platforms and Microsoft are expected to spend roughly $700 billion on capital expenditures during 2026 alone, with much of that money directed toward AI infrastructure.
That figure could approach $1 trillion in 2027.
For perspective, the four companies spent approximately $125 billion on plants and equipment in 2021. Their annual investment has increased more than fivefold in just five years.
Investors tolerated the first stage of the buildout because Big Tech could finance it comfortably with operating cash flow. That calculation is changing. The next phase will require a mixture of corporate debt, private credit, equity issuance, joint ventures, leasing arrangements and government support.
Credit markets have started to notice. The cost of insuring the debt of several hyperscalers against default has risen, even though their balance sheets remain strong by conventional measures.
AI stocks have also experienced a meaningful pullback from their recent highs. That combination has raised an uncomfortable question: Is the AI capital spending bubble already beginning to burst?
History suggests the answer is probably no.
Transformational technologies have repeatedly produced periods of excessive investment.
Railroads generated a capital rush during the 19th century. Electrification drove another in the 1920s. Telecommunications and internet infrastructure fueled the dot-com era. Housing and mortgage finance created the next major credit boom.
Each cycle followed a familiar sequence:
A historical analysis of these cycles reveals a useful, if imperfect, benchmark. The U.S. economy has often absorbed total investment equal to roughly 25% of annual gross domestic product before a transformational capital boom reached its most dangerous stage.
Call it the “Rule of 25.”
The railroad expansion preceding the Panic of 1873 involved investment equal to approximately one-quarter of the economy at the beginning of the cycle. Internet infrastructure spending around the dot-com boom reached a similar proportion. The industrial and electrification surge of the 1920s also approached that range.
With annual U.S. GDP now around $30 trillion, the Rule of 25 would place the theoretical AI danger zone near $7.5 trillion in cumulative domestic investment.
Current spending remains well below that level.
Estimates suggest hyperscalers could spend several trillion dollars globally through 2029. Other companies, including Oracle, OpenAI, Anthropic and specialized cloud providers, are investing alongside them. Even under aggressive projections, total domestic AI spending may take until the early 2030s to approach the historical threshold.
The framework does not guarantee another five years of rising AI stocks. It does suggest that the overall infrastructure cycle may be much younger than the headline spending figures make it appear.
The most important change in the AI trade is occurring away from the stock market.
Big Tech’s earliest data centers were largely funded through internal cash generation. The next several trillion dollars will place greater demands on external capital.
J.P. Morgan Securities analyst Tarek Hamid estimates that more than $5 trillion could be invested in AI data centers between 2026 and 2030. Under that projection, approximately $1.5 trillion would come from cash flow, while credit markets, new equity and alternative financing structures would provide the rest.
That funding mix matters.
Debt introduces interest-rate sensitivity. Equity issuance can dilute existing shareholders. Private-credit structures may obscure leverage. Long-term leases can create financial obligations that resemble debt even when they are presented differently on corporate balance sheets.
Companies are already experimenting with more complicated arrangements. Some projects use special-purpose vehicles funded by private-equity firms, insurers and credit funds. The technology company may hold only a minority stake in the facility while agreeing to rent its computing capacity and cover operating expenses.
These structures can help keep construction moving, but they also distribute risk across parts of the financial system that are harder for public-market investors to monitor.
The AI boom therefore faces two separate tests. Companies must prove that the infrastructure can generate sufficient revenue, and capital markets must remain willing to finance construction until that revenue arrives.
The second test may become decisive first.
Capital spending bubbles rarely collapse simply because investors recognize that spending is excessive. They usually require a catalyst.
Previous booms were interrupted by banking panics, natural disasters, regulatory changes, tighter monetary policy or deteriorating credit conditions. The AI cycle could follow the same pattern.
The clearest threat is a sustained rise in long-term interest rates.
Higher Treasury yields affect the AI trade through several channels:
A multibillion-dollar facility generates returns over many years. As borrowing costs rise, the present value of those future cash flows falls. Projects that appeared highly attractive at a 4% financing rate may look far less compelling at 6% or 7%.
Many AI-linked companies trade at elevated earnings multiples because investors expect years of rapid growth. Higher bond yields reduce the relative value of distant earnings and can trigger sharp valuation compression.
Alphabet, Amazon, Meta and Microsoft have powerful existing businesses. Smaller AI infrastructure companies often rely more heavily on debt, customer commitments and continued access to outside financing.
If credit conditions tighten, the market may divide quickly between companies that can self-fund and those that depend on capital remaining cheap.
Special-purpose vehicles, private-credit arrangements and long-term data-center leases have expanded during favorable market conditions. A recession or refinancing shock would reveal where the leverage ultimately resides.
Investors should therefore watch the 10-year Treasury yield as closely as AI revenue growth. A persistent increase in yields could damage infrastructure stocks even while demand for computing power remains strong.
The case for remaining exposed to the AI cycle rests on more than historical comparisons.
There are early signs that the infrastructure is being used productively.
Cloud demand remains strong across the largest providers. Alphabet, Amazon and Microsoft continue to report expanding AI workloads, large cloud backlogs and demand that exceeds available computing capacity in some areas.
The economics of a successful data center can also be unusually attractive. A facility that costs tens of billions of dollars to build may eventually support a comparable amount of annual revenue if utilization stays high and computing prices remain firm.
That possibility separates AI infrastructure from projects that create capacity without a clear source of demand.
Businesses are paying for AI models, cloud computing, software tools and automated services today. The unresolved issue is whether future revenue will be large enough to justify the speed and scale of construction.
The answer may vary sharply by company.
A hyperscaler can use the same infrastructure across cloud services, advertising, search, productivity software and internal operations. A specialized provider with a concentrated customer base has fewer ways to absorb excess capacity.
The boom can continue while producing both major winners and severe casualties.
Investors naturally focus on which company will build the dominant AI model. That may be the least predictable part of the opportunity.
Models can improve rapidly, pricing can fall and customers can switch providers. Infrastructure suppliers may benefit regardless of which model captures the most users.
The next stage of the buildout is expanding beyond graphics processors. AI factories require memory, networking equipment, fiber, power distribution, cooling systems, connectors, transformers and specialized software.
That broadens the potential beneficiary list.
Companies such as Amphenol and TE Connectivity provide components needed to move enormous quantities of data. Eaton supplies power-management equipment. Vertiv sells cooling and other data-center infrastructure. Utilities and independent power producers may benefit from rising electricity demand, while engineering and construction firms can participate in the physical expansion.
Valuation still matters. Many infrastructure names already trade at premiums reflecting years of expected growth. Even a strong business can deliver disappointing stock returns when investors pay too much for it.
The more durable strategy is to identify companies with approved projects, reliable customers, strong balance sheets and products that remain essential across multiple AI architectures.
Data-center development is facing growing resistance from communities worried about electricity prices, water consumption, land use and pressure on local power grids.
That backlash creates a strange market dynamic.
Slower permitting could restrict the supply of new computing capacity. Companies with operating facilities, secured power agreements or fully approved projects could become more valuable as available capacity grows scarce.
This means political opposition to data centers may hurt the overall pace of construction while improving the economics of certain existing operators.
Investors should distinguish between companies announcing ambitious projects and those that already control power, land, permits and long-term customer contracts. During an infrastructure rush, execution rights can become more valuable than technological promises.
The strongest bearish argument is easy to understand. Spending has exploded, financing is growing more complex and companies are racing to build capacity before anyone knows the ultimate level of demand.
Those are classic bubble conditions.
Yet identifying a bubble does not reveal when it will end.
Capital booms often continue well beyond the point that disciplined investors consider reasonable. New financing extends the cycle, rising share prices attract more capital and infrastructure spending creates revenue for a wide network of suppliers.
That feedback loop can persist for years.
The most likely outcome may be a prolonged expansion punctuated by increasingly violent corrections. The broad buildout could continue even as individual AI stocks fall 30%, 50% or more.
Investors who treat every pullback as proof that the cycle is finished risk leaving too early. Those who assume every AI company will survive risk learning the opposite lesson.
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
Plot twist: this article was written by ChatGPT. /s
Just bet on MSFT and Google.
I would say right-sie. There will be a shaking and the unprepared might be driven out of the market. It just won’t disappear.
GPT summary for Freeper day drinkers:
The article argues that the enormous surge in AI infrastructure spending **probably is not yet a bubble about to burst**, comparing today’s investment cycle with historical booms in railroads, electrification, telecommunications, and housing; it introduces a “Rule of 25,” suggesting serious danger may arise when cumulative investment reaches roughly 25% of GDP, while current AI spending remains well below that level. The author says the bigger risk is **how future AI construction will be financed**, as companies increasingly rely on debt, private credit, equity, leases, and joint ventures, making the sector more vulnerable to rising interest rates and tighter credit. At the same time, strong demand for AI computing and growing revenue could sustain the buildout for years, although weaker companies may fail while major infrastructure providers benefit.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.