In the Loop – October 2, 2026
“In The Loop” is designed to give you a short update reflecting major developments, earnings, and investment trends across some core Equity Income and Growth holdings. All clients should be aware that individual buy/sell recommendations will be conveyed directly to you on an individual basis. Have a great weekend.
September lived up to its reputation as a difficult month, but the headline indexes did not tell the whole story. The S&P 500 declined just 0.5%, while the Nasdaq gained 1.9%. The Dow, however, fell 4.3%. More revealing was the equal-weighted S&P 500, which declined 4.4%, its worst month since March. In other words, owning the index looked considerably better than owning the “average stock.” A relatively small group of large technology and AI-related companies continued to support the capitalization-weighted indexes while much of the rest of the market corrected.
The Iran conflict remained one of the biggest macro variables throughout September. Oil prices swung sharply as markets attempted to price disruptions around the Strait of Hormuz, attacks on regional energy infrastructure and, later in the month, the possibility of a negotiated path toward de-escalation. Brent spent portions of September above $100 per barrel, while WTI also briefly traded above $100 before retreating as diplomatic hopes improved. The result was an unusual combination of higher oil prices, inflation concerns and upward pressure on long-term interest rates.
That relationship is important. Higher energy prices do not simply affect what consumers pay at the pump. They influence inflation expectations, transportation costs, corporate margins and ultimately the bond market. Treasury Secretary Scott Bessent specifically highlighted the unusually strong relationship between oil, crack spreads and longer-term Treasury yields, arguing that rates could decline once the market gets to the other side of the Middle East conflict.
September may have provided the clearest test yet of what we have been calling the Warsh-Bessent Framework. The Federal Reserve is focused on controlling inflation, while Treasury is increasingly focused on the structure and functioning of the long end of the yield curve. Despite Treasury efforts—including bond buybacks—the bond market continued demanding higher yields. By September 29, the 10-year Treasury reached 5.26% and the 30-year reached 5.59%.
This is why we continue to emphasize the old market saying with a slight modification: “Don’t fight the Treasury.” But September also reminded us that Treasury does not completely control the bond market. Oil, inflation expectations, government borrowing and economic growth ultimately matter. The implication for investors is straightforward: companies dependent upon cheap refinancing and distant future earnings become increasingly vulnerable as long-term rates rise. Companies generating cash today, possessing strong balance sheets and investing behind identifiable demand should be better positioned.
Perhaps the most important chart for September was the gap between the S&P 500 and its equal-weighted counterpart. The regular S&P 500 fell only 0.5%, compared with a 4.4% decline for the equal-weight index. For the third quarter, the capitalization-weighted S&P 500 gained 2.82%, while the equal-weight index declined 1.67%.
That divergence tells us market leadership has become extremely narrow. By late September, financials and utilities were among the weaker sectors, both down roughly 5%, while technology remained the standout area of strength. Semiconductor and AI-related shares continued to attract capital even as higher rates pressured much of the broader market. Nine of the S&P 500’s eleven sectors ultimately finished September lower.
September also provided an interesting stress test for our Two-Speed Economy framework. Speed 1 represents the physical build-out taking place around AI infrastructure, power generation, the electrical grid, data centers, defense, domestic manufacturing and other strategic infrastructure. This has been our team’s preferred investable theme for the last 18 months. Speed 2 represents the more traditional consumer- and service-oriented economy, where spending and financing costs can have a much greater influence.
The results were more nuanced than simply “Speed 1 won.” The strongest portion of Speed 1 remained the AI infrastructure ecosystem—semiconductors, memory, networking, compute and selected power-related companies. Traditional rate-sensitive infrastructure, utilities and leveraged businesses struggled as long-term yields moved higher. Meanwhile, portions of Speed 2, including consumer-sensitive businesses, financials and traditional IT services came under pressure. That distinction matters. Our thesis is increasingly evolving from simply owning infrastructure to owning infrastructure companies with strong balance sheets, free cash flow, pricing power and limited refinancing needs.
That is why we recently added what we call High-Rate Survivability to our investment framework: net debt-to-EBITDA, interest coverage, free-cash-flow conversion and debt maturities over the next three years. September demonstrated why those characteristics may become increasingly important if the 30-year Treasury remains above 5%.
September also produced another wave of dramatic warnings about artificial intelligence. Researchers, technology executives and policymakers publicly debated whether increasingly capable AI systems could eventually escape human control or pose catastrophic risks. Those concerns should not simply be dismissed; AI safety, cybersecurity and governance will become increasingly important as the technology advances.
But markets have been hearing versions of the “AI is moving too fast” argument throughout this cycle while corporations continue writing increasingly large checks for compute, networking, memory, power and data-center infrastructure. That creates an unusual investment environment: the technology can simultaneously create legitimate societal concerns and an enormous capital-investment cycle. Our job as investors is not to decide the philosophical debate. It is to follow the capital.
And right now, the checks continue to be written.
September’s message was not that the bull market ended. It was that the market became considerably more selective. The S&P 500’s modest decline concealed significant weakness underneath the surface, long-term Treasury yields moved to levels not seen in decades, oil remained a geopolitical wildcard, and AI continued to dominate both investment spending and investor psychology.
Our response remains consistent: emphasize balance-sheet strength, free cash flow, pricing power and identifiable capital spending, while becoming increasingly skeptical of companies whose valuations require cheap money to return.
The next phase of this market may be less about simply owning growth and more about identifying who is receiving the checks, who is writing the checks—and who can continue doing so with the 30-year Treasury above 5%.
That is the market we are watching as we enter the fourth quarter.
Individual Company Updates
One of September’s strongest themes remained the race to solve the AI power bottleneck, and Bloom Energy continues to move closer to the center of that story. Bloom’s fuel cells can provide onsite electricity without waiting years for new transmission lines or grid interconnections. Its expanded agreement with Oracle calls for as much as 2.8 gigawatts of capacity, with an initial 1.2 GW already contracted. In September, Bloom took the story another step forward by highlighting its ability to produce 800-volt DC power natively, potentially eliminating several power-conversion steps between generation and next-generation AI computing equipment. Bloom estimates that for a 1-GW AI data center, its architecture could reduce non-compute capital expenditures by $3.6 billion, or 27%, and five-year total cost of ownership by approximately $5.5 billion compared with traditional AC architecture. Those are company estimates, but they illustrate why onsite power is becoming such an important part of the AI infrastructure conversation. In our Speed 1 framework, Bloom is increasingly less a traditional clean-energy story and more an AI infrastructure company whose product happens to be electricity.
The next stage of the AI buildout isn’t simply about adding more GPUs. As AI clusters become larger, moving information between processors, memory, storage and entire data-center campuses becomes a bottleneck of its own. That’s where Marvell fits. Its most recent quarter produced record revenue of $2.74 billion, up 37% year over year, while data-center revenue increased 46% and represented an extraordinary 79% of total sales. Management said AI bookings remain exceptionally strong and raised its revenue outlook for both fiscal 2027 and 2028. September reinforced the thesis as Marvell showcased next-generation Ethernet switching, PCIe 6.0, CXL memory, optical interconnects and its new 2-nanometer optical technology. Think of Nvidia and other accelerators as the engines of the AI factory; Marvell is increasingly supplying the high-speed roads connecting those engines. That’s why MRVL remains one of our preferred ways of participating in the networking and connectivity bottleneck created by increasingly large AI clusters.
Micron provided perhaps the clearest evidence this week that the AI capital-spending cycle is spreading well beyond GPUs. Fiscal fourth-quarter revenue reached a record $54.2 billion, versus $11.3 billion a year earlier, while operating cash flow surged to approximately $44 billion. For the full fiscal year, Micron generated nearly $90 billion of operating cash flow. The reason is increasingly straightforward: artificial intelligence requires enormous amounts of high-performance memory, and memory is becoming another critical constraint on how quickly AI infrastructure can scale. Management expects an even stronger fiscal 2027 and says longer-term strategic customer agreements provide greater visibility into future demand. The important takeaway isn’t simply Micron’s extraordinary earnings growth—it is the emerging scarcity value of memory. AI infrastructure increasingly requires Compute + Networking + Memory + Power, and Micron occupies one of those four critical bottlenecks.
ERock is a newer public company that deserves attention because it sits directly at the intersection of two themes we have discussed repeatedly: AI infrastructure and power scarcity. ERock provides modular, natural-gas-fired onsite generation that can be deployed considerably faster than conventional utility infrastructure. Its contracted power-system backlog reached approximately $1.7 billion in the second quarter, roughly 10 times the prior-year level, driven primarily by AI data-center customers. A 470-MW Anthropic order extended production commitments into 2028, while the company has begun construction of a 366-MW generation facility supporting Meta’s data-center campus. Following its June IPO, ERock ended the quarter with approximately $627 million of unrestricted cash, no debt and an undrawn $250 million credit facility.
New, Small-cap companies are harder to value because they do not have a long history of financials to analyze. In this circumstance, our team does a deep dive into the management team.
What particularly caught our attention about Erock is the board of directors. Former U.S. Energy Secretary Dan Brouillette brings an unusually deep understanding of energy policy, utilities and infrastructure. Tony Satterthwaite, former President and COO of Cummins, brings decades of industrial manufacturing and power-generation experience. Charles Boynton, CFO of Nextpower and former CFO of SunPower, contributes public-company finance and power-industry expertise. Mark Patterson, a former senior investment banker at Merrill Lynch and Citigroup who also serves on Digital Realty’s board, adds direct exposure to data-center real estate and capital markets. And Hans Kobler, founder of Energy Impact Partners, brings decades of experience investing in power and energy technology. CEO John Carrington previously led Stem and held senior commercial leadership roles at First Solar. This is an unusually relevant collection of power generation + utilities + manufacturing + data centers + capital markets + government policy experience for a company trying to scale into one of the largest electricity-demand expansions in decades.
ERock therefore represents something slightly different from Bloom. Bloom is using fuel-cell technology to rethink how AI facilities receive power; ERock is essentially saying the grid can’t arrive quickly enough, so bring utility-grade generation directly to the customer. With data-center construction timelines increasingly running ahead of traditional grid expansion, that “speed-to-power” proposition could become one of the more important pieces of the Speed 1 infrastructure buildout.
These four companies illustrate why we continue to focus on who is receiving the checks from the AI capital-spending boom. Bloom and ERock address the power constraint. Marvell addresses the networking and connectivity constraint. Micron addresses the memory constraint.
As AI models become larger and inference becomes more widespread, removing one bottleneck simply exposes the next one. That is why we continue to believe the better way to think about AI isn’t simply as a technology trade. It is becoming one of the largest physical infrastructure buildouts in decades, and that remains firmly in our Speed 1 economy.
Have a great weekend.
Formidable Asset Management (“Massey Romans Capital”) is an investment adviser registered under the Investment Advisers Act of 1940. The information presented in the material is general in nature and is not designed to address your investment objectives, financial situation or particular needs. Prior to making any investment decision, you should assess, or seek advice from a professional regarding whether any particular transaction is relevant or appropriate to your individual circumstances. Although taken from reliable sources, the Firm cannot guarantee the accuracy of the information received from third parties.
The opinions expressed herein are those of the Firm and may not actually come to pass.
Related posts
In the Loop – April 17, 2026
Markets hit all-time highs as consumer sentiment nears 80-year lows. Explore the
In the Loop – September 18, 2026
Investors are confronting what may be the tallest “Wall of Worry” in years.
In the Loop – March 20, 2026
Before I get into the weeds, I want to remind all long-term investors that since