October 15, 2025

A Quarter of Change

Exuberant markets and rising speculation defined the third quarter of 2025. Major U.S. equity indices advanced despite historically stretched valuations and pockets of speculative froth. Fervor around artificial intelligence (AI) and cryptocurrencies drove risk assets to new highs even as seasoned observers and some regulators warned of bubble‑like conditions. In this backdrop, we adhered to our discipline: embracing genuine innovation while remaining wary of extreme pricing in pockets of the market. Our portfolio changes reflect a steady commitment to rational, long‑term investing amid the hype, even if today’s valuations remain short of the early‑2000s extremes.

Speculation ran hot in Q3. AI‑linked companies helped push the S&P 500 and Nasdaq to all‑time highs as investors displayed an insatiable appetite for anything AI‑related. Tech giants and AI‑focused startups saw valuations re‑rate rapidly as investors bet on AI’s potential across industries from healthcare to finance. While this enthusiasm has created substantial wealth and bolstered corporate confidence, it also raises concerns about overheating. Regulators have begun to scrutinize frothy valuations, and many AI stocks appear priced for perfection as if success were certain despite the likelihood that adoption could be slower, competition fiercer, and regulation less predictable than markets assume.

History reminds us that “priced for perfection” episodes, most famously the late‑1990s tech boom, warrant caution. By some assessments, AI‑related stocks have entered the euphoric phase of a financial bubble, where prices diverge from fundamentals and the risk of permanent capital loss rises.

Crypto markets also surged. Bitcoin moved through the $100,000 milestone during the quarter and, by late Q3, hovered near $120,000 amid heavy ETF‑linked inflows and optimism for easier policy. The broader crypto market grew to an estimated $3.9 trillion, with Bitcoin comprising more than half. Such dramatic inflows and swift price moves epitomize speculative fervor and amplify volatility across risk assets. We have seen this movie before: flows can reverse without warning, and crypto’s notorious swings remind us how quickly sentiment can change.

How do we navigate this? By staying measured and pragmatic. AI is a genuine productivity engine. We use AI tools daily to enhance our research, and blockchain continues to drive financial innovation. Yet we will not suspend valuation discipline or succumb to FOMO. History shows that even transformative technologies can be overhyped. In an uncertain world, a margin of safety matters. While others chase momentum, we focus on evidence: separating durable economics from narrative‑driven stories. This balanced stance, respecting innovation without paying infinite prices, defines our approach to 2025’s speculative fever.

Corning Then vs. Now: Who Is the “Corning” of 2025+?

Corning (GLW), a portfolio holding, has benefited from renewed enthusiasm for fiber optics. As AI models demand ever‑greater data throughput, fiber lighter, faster, and more efficient than copper should be indispensable to next‑generation infrastructure. GLW is up ~75% year to date. As a nearby company (and one with a museum we recommend), we are pleased to see our local economy benefit as well.

We welcome the market’s recognition of Corning’s role, but it also invites déjà vu. In the late 1990s, fiber optics was similarly heralded as the Internet’s transformative backbone. From 1998–2000, Corning and its peers became emblematic of the dot‑com boom: extraordinary demand projections, soaring valuations, and widespread conviction that a new era had arrived.

The parallels today are striking. Once again, Corning sits at the center of a sweeping technological narrative, this time, AI infrastructure, though the business mix and capital discipline are meaningfully different from 1998–2000.

Corning’s stock rose from roughly $13 on 1/1/1998 to a peak of $113 on 8/28/2000. Operating performance was solid at best and, with ill‑timed acquisitions, deteriorated. As Corning’s 2000 10‑K notes, “In 2000, Corning completed 12 strategic business combinations valued at approximately $10 billion within the Telecommunications segment, including the pooling of interests with Oak Industries, Inc. (Oak) in January 2000.”

(Note the following is a best estimate using Corning’s 1998-2001 filings)

Acquisitions in 2000	Price	Currency
Oak Acquisition	$2,210.0	Stock
Acquisition of Optical Technologies from Pirelli and Cisco	$3,600.0	Cash
Acquisition of Intellisense	$500.0	Stock
Acquisition of NetOptix	$2,100.0	Stock
Siemens Optical Cable & Hardware business	$1,400.0	Cash
Small Deal	$85.0
Small Deal	$117.0
Small Deal	$24.0
Small Deal	$32.0
Small Deal	$63.5
Small Deal	$67.0
Total	$10,198.5

Corning Income Statement (in m except EPS)	1997	1998	1999	2000
Revenue	$3,554.3	$3,831.9	$4,741.0	$7,127.0
Operating Income*	$629.2	$482.3	$737.0	$1,154.0
Net Income*	$439.9	$421.3	$515.8	$884.6
Earnings Per Share	$0.60	$0.54	$0.65	$1.01
Average Stock Price during the Year	$15.7	$12.1	$22.6	$72.6
P/E (Avg Price during the year)	26.3x	22.3x	34.8x	72.2x
Diluted S/O (adj. for stock 3:1 stock split)	736.2	777.6	795	879.3
Cash Flow From Operations	$654.1	$682.2	$866.9	$1,421.2
Capex	($745.6)	($730.4)	($757.1)	($1,524.9)
FCF	($91.5)	($48.2)	$109.8	($103.7)
*2020 excludes $462.6m in acquisition-related charges

2000 Acquisition Financials (in m)	1999	2000
Revenue (Pro-Forma to include acquisitions)	$5,525.8	$7,298.8
Revenue Acquired	$784.8	$854.8
EV/Sales Multiple	13.0x	11.9x

From 1997 to 2000, Corning generated cumulative ($133.6m) of free cash flow (pre‑acquisition), investing roughly $3.76B in capex. Management clearly did not expand capacity, expecting demand to crater four years later. In 2001, Corning launched a restructuring that closed seven major manufacturing facilities and consolidated several smaller ones, noting significant underutilized space and idled fiber capacity as telecom demand slowed. Idling newly built facilities seemed unthinkable until it wasn’t. Business slows gradually, then suddenly.

Layering on ~$10B of acquisitions at a combined ~12x EV/sales proved expensive and nearly existential. The deals added ~$2.7B of debt and increased shares outstanding by ~20%. The stock fell from $113 to $1.50 over the next two years, and Corning issued dilutive equity and preferreds to remain in compliance with debt covenants.

Corning Today

Today looks different. Capital intensity is declining even as revenue and earnings inflect higher. On a TTM basis, Corning trades near 36x earnings, hardly “cheap,” but well below 2000 extremes. Some business segments remain underwhelming, and this is not a peak‑on‑peak setup (peak earnings and peak multiple) given a multi‑year data‑center buildout. Still, AI euphoria and private valuations for concept‑stage startups rhyme with the 2000s. We must balance the appeal of future returns against the possibility that some have been pulled forward. Historical MSD–HSD returns on equity and capital do not warrant 36x indefinitely; compressing toward a mid‑teens multiple over time would be a material headwind.

Corning Income Statement (in m except EPS)	2022	2023	2024
Revenue	$14,189.0	$12,588.0	$13,118.0
Operating Income*	$1,838.0	$1,187.0	$1,387.0
Net Income*	$1,791.00	$1,460.00	$1,703.00
Earnings Per Share	$2.09	$1.70	$1.96
P/E (Avg Price during the year)	15.3x	17.9x	24.2x
Diluted S/O	857	859	869
Cash Flow From Operations	$2,615.0	$2,005.0	$1,939.0
Capex	($1,604.0)	($1,390.0)	($965.0)
FCF	$1,011.0	$615.0	$974.0

The punchline: not everything tied to AI is a bubble, and 2000 should not be the sole template for downside. Corning is one example among many pick‑and‑shovel beneficiaries. If demand for AI components fell materially, revenues would decline across the ecosystem. Looking only at stock prices then or now misses important nuances. If 2022 reflects “normalized” EPS and P/E, Corning might trade closer to ~$30 (vs. ~$80 today), implying ~‑63% downside, but that ignores significant near‑term FCF, which could support a downside nearer ~$40 (~‑50%). A 50% drawdown is still unacceptable, but it is a far cry from the 99% decline from 2000-2002.

What Rhymes with Corning circa 2000?

Meta Platforms	2023	2024	2025	2026e	2027e
Revenue	$134,902.0	$164,501.0	$198,299.0	$228,389.0	$256,938.0
Operating Income	$67,414.0	$87,867.0	$104,190.0	$110,906.0	$121,735.0
Net Income	$439.9	$421.3	$515.8	$884.6	$884.6
Earnings Per Share	$20.98	$27.61	$34.26	$37.03	$40.90
P/E	33.1x	25.2x	20.3x	18.8x	17.0x
Cash Flow From Operations	$71,113.0	$91,328.0	$114,834.0	$134,668.0	$155,402.0
Capex	($27,045.0)	($37,256.0)	($70,616.0)	($105,059.0)	($118,191.0)
FCF	$44,068.0	$54,072.0	$44,218.0	$29,609.0	$37,211.0
Capex % of Sales	20.0%	22.6%	35.6%	46.0%	46.0%

Oracle	2023	2024	2025	2026e	2027e
Revenue	$49,955.0	$52,961.0	$57,399.0	$67,384.0	$83,460.0
Operating Income	$20,904.0	$23,054.0	$25,034.0	$28,400.0	$34,095.0
Net Income	$14,177.0	$15,709.0	$17,283.0	$19,932.0	$24,149.0
Earnings Per Share	$5.12	$5.56	$6.03	$6.81	$8.13
P/E	28.5x	43.4x	40.1x	35.5x	29.7x
Cash Flow From Operations	$17,165.0	$18,673.0	$19,126.0	$24,550.0	$29,901.0
Capex	($8,695.0)	($6,866.0)	($7,855.0)	($35,301.0)	($38,215.0)
FCF	$8,470.0	$11,807.0	$11,271.0	($10,751.0)	($8,314.0)
Capex % of Sales	17.4%	13.0%	13.7%	52.4%	45.8%

A company’s value is the present value of future cash flows discounted at an appropriate rate. Today, estimating free cash flow is increasingly uncertain. Will these capex‑heavy projects earn attractive ROIs higher or lower than history? When does the payoff arrive? How large is maintenance capex, and what does it imply for future ROIC? From the outside, the magnitude of spending is enormous, and the headline capex figures omit M&A and investments in private AI companies. The good news is that the funders are among the most profitable businesses in the world. Still, it is unclear whether anyone has high‑confidence ROI estimates. Consider a few plausible paths: search migrates from Google to ChatGPT‑like agents; Apple’s on‑device AI splits usage between desktop and mobile assistants; or Meta’s glasses become the dominant interface. The range of outcomes is vast; the technology will change the world. The likely winners are the suppliers NVIDIA, Corning, Micron, Broadcom, Marvell, and others. From a valuation perspective, wider FCF ranges typically argue for higher discount rates. Instead, the market is assigning lower ones, judging from multiples. Investors may believe the addressable market is so large that many will win. We are more cautious.

Hyperscaler spending is massive; 2026 capex could approach $600B. At the moment, no amount seems “too much.” Meta’s recently announced $29B financing with PIMCO and Blue Owl is another warning sign. With ~$54B of 2024 FCF and expanded access to debt markets, industry participants are determined not to lose this arms race.

Today brings to mind the mid‑to‑late 2010s, when VC‑funded ride‑hailing and food delivery created significant consumer surplus to gain market share. A decade later, the true costs to consumers and operators are clearer. Similarly, the value we gain from AI tools (for example, our $200/month ChatGPT plan) surpasses the price many times over, as discussed below. We can’t say what stage we’re in, but spending is nearing levels where economic returns might struggle to meet necessary thresholds. Our investments in NVIDIA, Corning, and other AI winners position us in capital‑light parts of the expansion; even so, we are strategically trimming some holdings. This isn’t an announcement of a “bubble,” but speculative activity is increasing, so caution is advisable. With uncertain returns, debt amplifies left-tail outcomes, whether the “debt” is financial or leases, warranting higher discount rates.

Portfolio Position Update

During the quarter, we exited three positions: Donnelley Financial (DFIN), Lazard (LAZ), and Capstone Copper (CSCCF), and initiated one new holding: Topgolf Callaway Brands (MODG).

Donnelley Financial (DFIN). For ~$200, we executed an analysis that would have been impractical a year ago. We purchased DFIN in 2022 (original write‑up link to be added) in the low‑$30s and exited in the low‑$50s, an acceptable outcome over ~3.5 years. Our thesis centered on software‑driven growth, higher margins, improved ROIC, and a more recurring revenue base. Over the past two quarters, however, we grew concerned that DFIN was losing share to Workiva as ActiveDisclosure retention weakened, contrary to expectations. Lacking in‑house developers, we devised another way to examine what was happening under the hood.

Using AI and off‑the‑shelf programming tools, we parsed thousands of SEC filings to identify which vendors companies used to file annual reports. Manually compiling this dataset would have taken weeks. The results suggested that while roughly half of IPO filers select DFIN (a stable share), the ongoing post‑IPO annual‑filing share has been slipping to Workiva. A deeper look at customer profiles reinforced the concern: Workiva dominates larger issuers, while DFIN’s base skews smaller (many <$100m market cap), a cohort more prone to delisting, failure, or acquisition. What we once viewed as a temporary lull in software momentum now looks more structural, so we exited. A detailed write‑up with data will follow on Substack.

Lazard (LAZ). Lazard has improved AUM trends, recently turning to net inflows while peers faced outflows. Since October 2023, CEO Peter Orszag has executed well on a turnaround. With the stock approaching our estimate of fair value, the risk‑reward became less compelling, and we exited.

Capstone Copper (CSCCF) – From our initial purchase in 2016 to today, Capstone has been nothing short of a rollercoaster. In 2015-2016, the price of copper fell by ~1/3, from $3/lb to $2/lb, resulting in Capstone’s stock declining from $1.30 to as low as $0.20. Sentiment around the economy, especially for cyclicals tied to capex, exports, and commodities, was grim into early 2016. Contracting PMIs, falling industrial output, and collapsing energy prices resulted in significant pressure. The stock traded at a fraction of its Tangible Book Value, and the cure to low prices is lower prices, which results in supply curtailments, and ultimately, a rebound in the industrial economy would drive demand and prices for copper higher. After a series of corporate actions, the pinnacle of which was the merger of Capstone Copper with Mantos Copper in 2021-2022, and a significant rebound in copper prices stemming from energy and data center build-outs, the risk/reward was no longer favorable.

Topgolf Callaway Brands (MODG) – We have followed MODG for years; it is a straightforward business in a domain where we have real familiarity. Since acquiring Topgolf in 2021, near the peak of the pandemic‑stimulus era, Topgolf venues have underperformed, and the stock fell from ~$30 at closing to $9–10, where we initiated. The legacy clubs, balls, and apparel business continues to perform well. A planned spin‑off of Topgolf was delayed after Topgolf CEO Artie Starrs departed to lead Harley‑Davidson; we believe this raises the probability of an outright sale. With the market assigning little value to Topgolf, selling it at a reasonable price could be accretive. That said, leverage and balance‑sheet complexity warrant caution.

Our Thesis (full write-up found here) centers around:

1. Post separation, the standalone Clubs/Balls, Apparel, and Toptracer business would trade near ~6x EPS, a valuation far too cheap for a growing, premium brand, where transactions have occurred at multiples significantly higher.
2. Toptracer, though small, could add meaningful earnings as installed bays expand and pricing increases.
3. A cleaner, less leveraged business could broaden the investor base; golf equipment sales are relatively resilient to mild downturns.
4. If Topgolf is sold, proceeds and existing cash could support significant share repurchases.

Research Trip

During the quarter, we toured our second mine of the year, the world’s largest salt mine in Goderich, Ontario, owned by Compass Minerals. The visit aimed to deepen our understanding of initiatives undertaken by an essentially new management team over the past two years. Seeing the asset firsthand and speaking with other investors proved invaluable.

Key Takeaways:

1. Mine maintenance costs are significant. Management has the opportunity to lower these by closing unproductive areas of the mine.
2. Food‑grade salt (Canada‑only) is effectively restarting from $0 in revenue after an October 2024 recall.
3. Safety is the top priority and is emphasized across the company.

We were encouraged by management actions and discussions that historically poor capital allocation decisions and a lack of preventative maintenance are no longer occurring. Back to Basics reverberated throughout the tour and discussions. This new management team appears to be the right people for the job, and while much of the basics are complete, basic blocking and tackling are key to driving consistent performance from here, a task seemingly effortless, yet done so poorly by prior management. Living in the northeast, owning Compass gives us one reason to cheer for a snowy winter, often something we suffer through with no benefit. At least now, when we are snowed in, Compass is there to say all is not lost.

Regards,

Dominick D’Angelo, CFA Dominick@okeefestevens.com 585-497-9878

Disclaimer

This document is for informational purposes only. O’Keefe Stevens Advisory is not providing any investment recommendations with the publishing of this document, and no firm performance data is included in this document. Advisory services offered through O’Keefe Stevens Advisory, an investment adviser registered with the U.S. Securities & Exchange Commission.