Puttachat Kumkrong/iStock via Getty Images
Dear Clients and Friends,
Market Update
Despite the highest tariffs to US consumers since 1933 and a 22% correction in the S&P 500 (SP500), (SPX) earlier this year, we enter Q4 with equities near all-time highs and monetary policy easing. On Sept 17, the Fed cut short-term interest rates to 4.00%–4.25%, citing progress on inflation and softer labor. This supports growth without declaring victory. Headline consumer price index (CPI) is up 2.9% year-over-year; helpful but not a green light. Markets are pricing two more 25 basis point cuts by year-end to 3.50%–3.75%, which is contingent on cooling of core service and wage inflation. If inflation stalls, the Fed can pause which sets up a measured, not rapid cutting cycle.
Tariffs, having dominated headlines this year, have not spiked everyday prices as they’ve hit small-weighted categories of CPI (e.g., lumber, furnishings) but were offset elsewhere. Goods deflation and cheaper traded input helped, with import prices flat-to-down through mid-2025. Post-COVID, goods prices normalized as prices of used cars and durables rolled over and shipping got cheaper. A slight drop in housing rental prices also helped. Core PCE sits in the high-2s, with stickiness in services, not tariff-exposed goods. Relief in energy commodity prices also helped, even as electricity and natural gas prices rose. On a net basis: the major swing factors preventing regular Americans from feeling noticeable tariff pain were deflation across goods and import/shipping.
That said, low-income consumers are strained, while high-income consumers remain resilient. As short-term interest rates drift down and real incomes stabilize, 2026 could see a modest recovery in overall consumer spending. We prefer owning businesses with a higher-income customer base, durable pricing power, and clear value propositions over broad rate-sensitive cyclicals. Regarding confidence across the economy, we are watching for whether inflation re-accelerates (energy, tariffs) and the pace and for loosening of mortgage/credit.
On the corporate side, boardrooms are leaning into mergers & acquisitions. Valuations have re-accelerated even with mixed deal volumes as we see larger and strategic deals. This is due to funding costs easing and companies pursuing scale. Regulatory timing can stretch closings, but momentum is positive where synergies are real. We expect continued M&A across AI-adjacent tech, infrastructure, and select industrials with the middle-market improving as rate volatility fades.
Regarding earnings and valuation, profits must carry returns. With the S&P 500 near all-time highs and forward 12-mo price-to-earnings 22–22.5x (above 5- and 10-yr averages), expansion of stock valuation multiples is limited. The next leg up depends on continued profit growth and free cash flow durability. We continue to favor businesses with pricing power, high gross margins, recurring revenue, disciplined capital allocation, and improving efficiency.
Are we in a bubble?
Capex defined as Capital Expenditure
Corporate investment in Artificial Intelligence has been the primary driver of the market, with capital expenditure extending beyond GPUs to the entire infrastructure stack: power, networking, real estate, and software. However, the nature of this investment is undergoing a profound shift, and to understand it, we must look at the company at the center of it all: OpenAI.
To understand OpenAI’s current position, the best historical analogy is not a recent mobile platform like iOS, but rather Microsoft (MSFT) Windows in the PC era. Just as Microsoft leveraged DOS, and the applications built for it to become the default operating system for the enterprise and then consumers, OpenAI has leveraged ChatGPT to aggregate hundreds of millions of users. This massive user base is the ultimate source of power, allowing OpenAI to exert gravitational pull on the entire AI value chain, from the chipmakers up to the cloud providers.
This leverage is now being used to orchestrate the buildout of the AI ecosystem on OpenAI’s terms. Consider the chip layer:
-
Nvidia (NVDA): After receiving a $100 billion investment from Nvidia, OpenAI is deploying much of that capital to purchase Nvidia’s chips.
-
AMD (AMD): To create a second source and reduce reliance on a single supplier, a playbook move straight from IBM (IBM)’s handling of Intel (INTC) in the 1980s, OpenAI struck a deal to purchase 6 gigawatts worth of AMD’s chips. In return, AMD granted OpenAI warrants for up to 160 million shares, or roughly 10% of the company.
What’s notable is that these aren’t straightforward transactions. They represent a form of vendor and circular financing, where capital and equity are exchanged in a closed loop. Skeptics point to this as being eerily reminiscent of Cisco (CSCO)’s vendor financing during the dot-com bubble, where Cisco would fund its own customers’ purchases. This is the first clear sign that the market is entering a new, more speculative phase.
Until recently, the AI infrastructure boom could be characterized as a disciplined, although expensive, race funded almost entirely by the massive free cash flows of a few hyperscalers: Microsoft, Google (GOOG, GOOGL), Amazon (AMZN), and Meta (META). That has fundamentally changed, and OpenAI is again the catalyst. To secure the staggering amount of computing power it needs, OpenAI has entered into massive, long-term contracts that its partners cannot fully fund from their own cash flow. The most prominent example is the reported ~$300 billion deal over five years with Oracle (ORCL) for compute capacity, linked to the ambitious “Stargate” data center project. To fulfill a contract of this magnitude—requiring an estimated 4.5 GW of power, Oracle has no choice but to tap the debt markets.
This is the inflection point. As Doug O’Laughlin from Fabricated Knowledge notes, the entrance of a debt-fueled player breaks the stable oligopoly.
“What we have lacked in this cycle so far is almost no debt-led funding. The Oracle RPO shows what’s possible if you’re willing to flex your balance sheet, and this will kick off an uneasy race… What had been a disciplined, cash-flow-funded race may now turn into a debt-fueled arms race.”
This shift into leverage and circular financing models feels like a classic late-cycle marker. It echoes the telecom bubble, where strategic equity spending gave way to massive debt issuance to fund fiber buildouts, leading to overcapacity and spectacular bankruptcies. Today, the profits in the AI industry are flowing primarily to the suppliers, Nvidia, data centers, and power companies, while the “true AI companies” face uncertain revenues and intense competition, another concerning sign.
This dynamic fits perfectly within the framework described by Byrne Hobart in his book Boom, which argues that capitalism occasionally disregards normal return-on-capital constraints to single-mindedly fund a new technology. Debt, as O’Laughlin states, is the “real accelerant”. This FOMO-driven parallel investment can be incredibly productive, solving many interlocking problems at once.
The critical question for investors, however, is what durable infrastructure will remain if/when the bubble eventually bursts? The telecom boom left behind swathes of dark fiber that power our internet today. An AI bubble spent on GPUs with a three-year useful life offers a far less durable foundation. A bubble that spurred a massive buildout in power generation, on the other hand, would benefit the economy for decades to come.
Given these classic bubble markers—parabolic capex, a shift to debt and circular financing, and narrative-driven valuations—our approach remains cautious and selective.
-
Our Preference: We favor platforms with fortress balance sheets and repeatable free cash flow (namely hyperscalers like MSFT, AMZN, GOOGL, META) that can fund their AI ambitions internally and pace spending as monetization clarifies. Their diversified profit pools (cloud, ads, software) and control over distribution provide a cushion. We also favor the essential “picks and shovels” suppliers who own critical bottlenecks in the tech stack and get paid on the buildout, thus avoiding the downstream utilization risk.
-
Our Concern: The real risk lies with new entrants and those in the middle of the stack who lack diversified businesses, rely on speculative financing, and are entirely dependent on the hope that AI adoption and monetization will arrive before the funding dries up.
In short, while we see signs of a capital expenditure bubble, we believe this momentum-backed phase can persist. We will continue to participate selectively, focusing on companies where the moats and cash flows are real, while actively avoiding potential balance-sheet accidents.
Ferguson Plc (FERG)
Ferguson is the largest scaled specialty distributor for North American plumbing/HVAC/waterworks. Its revenue is split ~51% residential, 49% non-residential; ~60% repair & replace (R&R) and 40% new housing builds. About 85% of revenue is finished goods; plumbing is ~50% of mix. ~95% of revenue is U.S., the rest Canada. Their customer base comprises of 22% waterworks, 17% residential trade plumbing, 14% residential build/remodel, 14% commercial/mechanical, 14% fire/fabrication/facility/industrial, 12% HVAC, 7% residential digital.
When we began due diligence for Equity Income early in the year, sentiment was washed out: FERG shares traded down 16% post-earnings on fears of commodity deflation and weak outlook. Yet revenue held despite housing weakness. The market priced in zero new-home construction growth plus low-single-digit commodity price deflation. In our opinion, the stock selloff was overdone: (1) housing demand had been weak for over 2 years, (2) commodity products are only 25% of FERG’s mix, (3) specialty distributors had consistently shown pricing resilience. With residential (40% of revenue) in the 5th–6th inning of weakness and non-residential (waterworks/civil/infra) strong, the worst was past us.
The overhang was uncertainty on timing: new-home recovery might lag while non-residential could slow first. FERG’s share price implied only low-single-digit earnings growth over the next two years, but our math suggested an easy path to beating expectations. So far, pricing power has held, and non-residential demand has been solid.
Consistent with our playbook, we expect continued compounding organically and via accretive M&A. Management has consistently added 2 pts of growth from small bolt-on acquisitions at attractive valuations, with immediate profit uplift when plugged into Ferguson’s procurement, tech, and logistics rails. Scale drives a structural edge—better sourcing, working-capital efficiency, tech—and pricing has proven sticky through deflation scares.
Diversification has helped Ferguson outperform smaller, concentrated peers while housing construction remains stalled. Multi-year tailwinds persist in waterworks, commercial, civil infrastructure, and industrial; data-center buildouts also help pull through demand for commercial plumbing/HVAC/pipes/valves/fittings.
Broadcom (AVGO)
We added Broadcom to Equity Income in April; it has been our best year-to-date performer. A mega-cap semiconductor conglomerate, Broadcom sells primarily design and critical components for leading-edge chips as well as data-center infrastructure software. We believe the real fight in AI is between Nvidia and Broadcom. With a fresh inflection underway, Broadcom is positioned to be the #2—if not #1—winner over the next decade.
Led by arguably one of the greatest CEOs of all-time in Hock Tan, Broadcom excels at game-changing M&A and has consistently executed for shareholders producing one of the most successful stocks in the past 2 decades. It competes with NVIDIA in AI and networking chips, while its sticky, non-cyclical software (92% gross margin) buffers cyclicality in semiconductors. Whereas NVIDIA delivers general-purpose “glove-fits-all” compute stacks, Broadcom dominates custom AI chips (ASICs) with Google as its anchor customer, designing application-specific computing chips for Google’s (GOOG),(GOOGL) many applications (Search, YouTube, Cloud). Meta and OpenAI have also signed multi-billion-dollar deals to make their own chips. Custom-AI revenue rose 225% in 2024 to $8.6B and is expected to grow ~170%+ this year to $24B.
In networking, what allows these chips to talk to one another and work in conjunction and an adjacent and hypergrowth category in AI compute, Broadcom leads with Ethernet—the standard technology across non-AI clouds/enterprise data centers—pressuring proprietary approaches; NVIDIA in response even announced its own Ethernet solution.
Its data-center software (VMware) is the 800-lb gorilla in server virtualization, helping manage computing costs (including AI). This high-margin segment should grow teens this year as customers shift from licenses to subscriptions, then at least mid-single digits. Broadcom also holds a near-monopoly in iPhone connectivity (chips for cellular/Bluetooth/Wi-Fi/GPS).
We bought our position after a >30% pullback in share prices on tariff fears that implied a 30% EPS cut over 2–3 years. We saw as severely oversold given (1) Broadcom’s lower China-tariff exposure, (2) Google’s rapid AI-chip progress driving demand for more chips from Broadcom and interest from other hyperscalers, (3) continued share gains/innovation in networking, and (4) upcoming recovery across Broadcom’s non-AI businesses.
Broadcom sits in a duopoly behind NVIDIA across the AI stack—owning the Ethernet scale-out, custom AI chips for the biggest buyers, a high-margin software cash engine, and iPhone connectivity. 2026 should be another step-up year in AI and networking, with subscriptions fueling steady free cash flow.
As always, we are grateful for your continued trust and partnership, and we are always happy to answer any questions you may have so please feel free to email, call, or come into the office to see us.
Sincerely,
Ed Zhang
Please note that we will be hosting our annual dinner for our clients and friends at the Briar Club in Houston on Thursday, January 29th. We will have time for the usual fellowship and our outlook for the markets. We will also return to our format of a keynote speech with Dan Clifton of Strategas Research Partners returning as our speaker. Dan is an expert on politics and its impact on the financial markets and spoke at our dinner back in 2023. You should receive a formal invitation in mid-December. We hope you can make the event, and we will be live streaming it for those who cannot attend.
Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.
Source link
