Quarterly Investor Letter Q3 2026
October 30th, 2026
The S&P 500 returned 2.3% in Q3 2026, a number that hides more than it shows. About six in ten members of the index fell; the equal-weight S&P lost 1.9%, and the Russell 2000 lost 7.2% after gaining 21.5% in Q2. A handful of mega-caps, led by Microsoft (+37.5%), did the work. Meanwhile, on September 16, the Fed raised rates for the first time since 2023, and the 10-year Treasury ended the quarter at 5.29%, its highest level since 2002.
During the quarter, we made three changes to the portfolio. We bought Octave Intelligence (OCTV), an industrial and public safety software business spun off by Hexagon in May. We sold M&T Bank (MTB) in July at ~$248. And on July 30, we repurchased the options we had put on against our Corning position.
Perrigo, NVIDIA, and Baxter were among the portfolio’s best-performing securities in the quarter. Corning, Callaway, Sphere, Weyerhaeuser and our Fannie Mae preferred positions were a negative contribution in the quarter.
A Narrow Market
Q2’s rally was broad; Q3’s was not. The S&P 500 returned 2.3% (+12.7% YTD) and the Nasdaq 2.6% (+16.1% YTD), while the equal-weight S&P fell 1.9% (+10.0% YTD), the Dow fell 2.3% (+7.2% YTD), and the Russell 2000 fell 7.2% (+13.9% YTD). By sector, Energy (+17.2%) and Technology (+7.2%) led, while Industrials (-9.7%) and Utilities (-12.4%) lagged.
Under the surface, about 311 of the 500 members fell; over 1/3 of S&P 500 companies sit 20% or more below their 52-week highs. The ten largest companies are ~39% of the index against a historical average near 24%, and the Magnificent 7 are ~34.6%. At 19.0x forward earnings (FactSet), the S&P sits right at its 10-year average of 19.1x, though today, alternatives (bonds) offer substantially higher yields than the average of the last 10 years.
In Q2, we described our view that market participants were short software to fund long positions in semiconductors, and wrote that if it flipped, the whiplash could be violent. Q3 saw semiconductors fall more than 11%, while the iShares Software ETF (IGV) rose 17%. Octave is where our watching and waiting turned into a purchase. Volatility is the opportunity, but only if one can underwrite and differentiate a cheap business from one that deserves to be cheap.
AI Is a Price Question
In our Q3 2025 letter, we wrote that 2026 hyperscaler capex could approach $600B. By Q1, the figure was closer to $700B. Following summer earnings, the four largest spenders (Microsoft, Alphabet, Amazon and Meta) guided to roughly $720B to $745B for 2026. NVIDIA’s July-quarter revenue was $96.2B (+106%), and Microsoft’s commercial backlog reached $678B (+84%), roughly a third of which is estimated to be tied to OpenAI.
We continue to highlight the falling free cash flow: combined trailing free cash flow for the four is down 24% y/y, Alphabet’s was -$5.9B in Q2, and Oracle’s was -$5.4B in its fiscal first quarter, paired with a $20B equity sale. Hyperscalers issued ~$220B of bonds through August 10, against about $70B in all of 2025, and smaller borrowers like CoreWeave pay yields near 15%. A year ago, we walked through Corning in 1998 to 2000, when vendor financing and debt-funded capacity drove the telecom buildout. The pattern is starting to rhyme, with one important difference: today’s largest spenders continue to have fortress-like balance sheets amid all the issuance. However, our job as investors is to see where the puck is going, and it’s likely these balance sheets will continue to worsen over the coming years.
The next source of funding may be public equity. OpenAI raised at an $852B valuation in March, and Anthropic raised at $965B in May and reportedly confidentially filed for an IPO in June. Supply of that size competes for the same dollars that own today’s AI winners. Just as equity indices are top-heavy in the same bet, so are AI spend and usage; at what point will investors begin to say they are full on AI? The amount we are discussing is no longer a drop in the bucket. As of mid-September, Goldman tracked $578B of AI-related gross debt issuance YTD globally, with hyperscalers about 40% of it. Through September, corporate debt issuance for the year was ~$2.2T (+25% y/y), meaning ~25% of corporate debt issuance is tied to one overarching theme. Given the expectation of further growth in 2027-2028, this share of total issuance will likely rise. While AI compute demand appears insatiable today, we have questions about what the bond (and equity) market’s capacity to absorb this issuance will be.
We continue to find ourselves in a yes-but mindset. We use AI extensively in our day-to-day research and operations, and continue to find new and transformative ways to automate or improve processes. We try to square that with the fact that it feels like we are barely scratching the surface of what AI can do for organizations, and therefore the financial impact and efficiency improvements are still in the early innings. On the other hand, the sheer dollar size of spending by major players, coupled with the concern that hyperscalers are planning for up-and-to-the-right demand and the reality that new technology takes time and know-how to implement, raises the question of whether an air pocket will arise as companies slowly transition processes from legacy to modern tech. A final note: many market participants find safety in the hyperscaler take-or-pay (contractual guarantee) of revenue over the coming years. As prior investors in GrafTech International (EAF), we are cautious that, should demand deteriorate, these contracts will be restructured or headed for bankruptcy claims.
Corning (GLW)
Corning was our biggest winner in Q2, and we trimmed and hedged the position. In Q3, the stock fell 39.8%, from $255.43 to $153.76, the largest decline in the portfolio.
In Q2, core sales rose 17% to $4.74B, core EPS rose 30% to $0.78, and Optical Communications grew 32%. Management guided Q3 core sales to $4.9B to $5.0B and core EPS to $0.85 to $0.89, and reiterated its target of a $20B annualized sales run rate by the end of 2026. We continue to own Corning, a position we first bought nearly a decade before the AI boom. Our job is not to be right about every quarter’s price; it is to size the position so that a quarter like this is painful rather than permanent. Position sizing does not eliminate the risk of loss.
Corning’s decline also highlights how the actions of a single investment fund can have an outsized influence on security prices, particularly when leverage and crowded positioning are involved. According to press reports, in July, Situational Awareness, the AI-focused hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, suffered a 67% decline following significant losses across its concentrated portfolio of AI infrastructure investments. Facing mounting liquidity pressures, the fund was forced to sell most of its approximately $16 billion public equity portfolio to Citadel. Citadel subsequently reduced more than 80% of the acquired portfolio’s risk through nearly 100 block trades totaling over $4 billion.
We believe this liquidation likely contributed to the selling pressure (and buying pressure on the way up) across AI infrastructure stocks, including Corning. The impact of a large fund extends beyond its direct holdings. As prominent investors build substantial positions, others often follow their investments, creating crowded trades in which many shareholders own the same securities for similar reasons. When leverage forces one of these investors to liquidate, selling pressure can quickly spread as other investors reduce exposure, liquidity deteriorates, and declining share prices trigger additional selling. We believe these dynamics likely amplified the volatility experienced across AI-related securities during the quarter. While Corning’s direct exposure to the liquidation cannot be quantified, its 39.8% decline, despite continued fundamental growth, demonstrates how market positioning and forced selling can overwhelm underlying business performance in the short term. It also reinforces the importance of valuation discipline, prudent position sizing, and avoiding excessive leverage.
New Position: Octave Intelligence (OCTV)
We bought Octave Intelligence during Q3. Octave is the industrial and public safety software business Hexagon AB spun off on May 22, 2026, at one Octave share for every ten Hexagon shares. When-issued trading began May 21 and regular-way trading May 28, with depositary receipts in Stockholm from May 25. The spin included Hexagon’s Asset Lifecycle Intelligence and Safety, Infrastructure & Geospatial divisions, plus ETQ and Bricsys.
Last quarter, we explained why we waited years to buy Sotera: we do not buy good businesses while forced sellers lean on the stock. Octave is the same idea from the other side, a typical orphaned spin. It came out at ~$6B against Hexagon’s ~$29B pre-spin market value. Hexagon holders, mostly Swedish, woke up with a US-listed stock a fraction of the size of their Hexagon stake, in a software business unrelated to the measurement company they chose to own, in a category everyone thinks AI will displace. They could buy 5x to 10x more to make it matter in their portfolio, or sell it and keep Hexagon. Selling is the easier choice. We like situations where sellers are price-insensitive and want out.
The structure adds to it. A US-only primary listing, a depositary line in Stockholm and two share classes keep index funds and mandate-constrained buyers on the sidelines. Melker Schörling AB, Hexagon’s anchor holder, owns 21.8% of the shares and 42.9% of the votes, and its registration rights open May 22, 2027. After closing its first regular-way day at $17.05, the stock traded in a range of roughly $15.41 to $27.39 and closed Q3 at $18.88.
Octave sells software across four workflows: Design (~40% of revenue), Operate (~30%), Protect (~20%), and Build (~10%). The pitch is that the cost of an error compounds at each stage of an asset’s life: 1x in design, 5x in construction, 25x in operation, and 500x+ when a refinery, power plant, or emergency system fails. Octave’s software sits where those errors get caught. For FY25, revenue was $1.638B, ARR was above $1.1B, adjusted operating income was ~$506m (a 31% margin), and free cash flow was $320m (20%). Gross retention is 97% (99% for large customers), and net retention is 105%. Customers include BMW, Phillips 66, Cargill, Fluor, Bechtel, and United Airlines. We think the market is missing four things.
- Cross-sell and low-hanging fruit is the thesis. 86% of customers use a single workflow. ARR per customer runs ~$150k on one workflow, ~$500k on two, and $2m+ on three or more. Under Hexagon, each workflow’s sales team was siloed; as one former Hexagon director put it to us, the divisions shared customers but not information. Renewals ran out of finance, so pricing was an afterthought. A Chief Revenue Officer hired in 2025 moved renewals into sales. On the Q2 call, management said more than 100 customers added a second solution in the quarter, an early sign of improvement.
- AI enhances these products; it does not displace them. The “software is dead” debate we described in Q2 is the reason the stock is cheap. We do not think it applies here. Octave sells mission-critical workflows priced at a small fraction of the cost of failure, built on decades of plant data. That is not something a company insources to save money. Our channel checks suggest customers want Octave to build AI workflows on their own engineering data, which entrenches it as the system of record.
- The revenue “miss” is largely irrelevant. Q2 organic revenue fell 1%, and management cut FY26 organic revenue guidance to 0% to 2% from 3% to 4% after public safety perpetual license deals slipped. ARR guidance was held at 6% to 8%, SaaS revenue grew 21%, and recurring revenue rose to 69% of LTM revenue from 65% a year earlier (the target is 75%). Moving customers from perpetual licenses to SaaS lowers reported revenue now and raises lifetime value.
- The balance sheet buys time. Net debt was ~$340m at June 30, about one year of free cash flow, even after the $625m payment to Hexagon. This balance sheet optionality will allow them to make tuck-in acquisitions or repurchase stock should the valuation remain disconnected from the stock price.
We also built our own check. Public safety procurement leaves a paper trail, so from council minutes, RFP awards, and contract disclosures, we tracked 80 municipal deals for OnCall, Octave’s dispatch software. Since 2024, OnCall has won 13 dispatch and records deals against 6 confirmed losses, and 2026 is 4-0 with seven bids still open. A subset, but a positive indication.
We purchased the stock in mid-August around ~$19/share. Based on our estimates, Octave trades at ~3x 2026 sales and ~12x forward earnings, 2.4 to 3 turns of EV/sales below Autodesk, Bentley, and PTC. Part of that gap is deserved due to the lower recurring mix (69% vs 93% to 97% at Bentley and PTC), a controlling shareholder, and a dual-class structure. We believe the discount could narrow over time if ARR growth remains steady and the recurring mix rises, although there is no assurance that it will.
What can go wrong. Melker Schörling’s stake and its 2027 registration rights are a known overhang. Protect is lumpy, still sold largely on perpetual licenses to governments. AI could weaken switching costs over time. Net retention may not rise from 105% toward management’s 108% target. We are heavily in the weeds with AI and our daily use of it; in conjunction with OCTV results, this informs our view on the underlying stickiness in a rapidly changing landscape.
Sold: M&T Bank (MTB)
We sold M&T on July 29 at $248 as the risk-reward was no longer attractive. The stock traded at 2.1x tangible book, up from 1.72x at year-end 2025. M&T is a high-quality bank and a neighbor down the Thruway, one we would happily own again at the right price. When a business we like reaches fair value, we sell and wait. While credit costs are low and wealth management (markets) are strong, it’s when the business landscape is rosy, and a stock is priced for that to continue, that the best time to sell often arrives, as any hiccup is unexpected and causes market participants to reassess their prior outlook. We make no claim of timing. The stock ended the quarter at $217.60.
Balance Sheets in a 5% World
With the 10-year above 5% and small caps down 7.2%, the natural question is how our more levered holdings are holding up. Our three most levered positions (PRGO, BAX, SHC) in the portfolio all happen to be in the Healthcare space.
Perrigo (PRGO) +42.1%. Our biggest loser in 2025 and our Q1 laggard was our best performer in Q3. While the business is far from fixed, when things go from bad to less bad, the stock usually reacts. The business is not fixed: Q2 core sales fell 3%, and core adjusted EPS fell 21%. Perrigo sold Dermacosmetics for ~$359m, paid down $759m of debt in the first half, and extended its credit agreement to 2031. Net debt is ~$2.9B, with a target at or slightly below 4.0x. The infant formula review is ongoing with no outcome, and an interim CEO is in place. Given the ongoing turmoil at the management level, it’s unsurprising that buyout rumors are swirling. A sale of the company is possible but speculative, and there is no assurance that any transaction will occur. Additionally, this dividend aristocrat does not necessarily have the healthiest balance sheet to keep this raise going. Therefore, the easy way out is to sell, avoiding this potentially negative reaction to the stock. We continue to hold Perrigo, most recently adding to the position in February 2026. What’s interesting about Perrigo (PRGO) is that it’s down significantly since our initial purchase and has been largely uncorrelated with many of our other positions. In factor terms, it would have been a natural funding short for anyone running a long momentum book, since the stock has done little but fall for the last decade. But when that factor reverses, the rebound can be sharp. PRGO is no longer an easy funding short, and with the buyout rumors it could become a natural long with a catalyst.
Underperformers During the Quarter
Several of the stocks that worked best in Q2 were among the weakest performers in Q3.
Sphere Entertainment (SPHR) -22.3%, after a 47.5% gain in Q2. The venue business is still growing (Q2 Sphere revenue +29%). The Wizard of Oz has slowed from its early run, and Las Vegas visitation fell 4.3% y/y in August. In Q2, we wrote that great assets often have longer right tails than one underwrites. The path along that tail is rarely smooth.
Callaway Golf (CALY) -25.0%. While the mid-quarter Good Good controversy was not expected, by and large everything appears to be status quo. Callaway raised full-year guidance after a solid Q2 (adjusted EBITDA +36%) but guided to a softer second half on a later iron launch and SKU cuts. The balance sheet did what we said it would: $1.4B of debt retired after the Topgolf sale, ~$204m of net cash at June 30, and ~$84m of stock repurchased through June under a $200m authorization. Golf + buybacks is still the thesis. In our Q4 letter, we put fair value at $20+ (an estimate that is not a prediction or guarantee of future results); the stock ended Q3 at $14.09. A risk moving forward continues to be the price of gas/diesel. Disposable income is lower as the consumer spends more on daily goods, which may cut into discretionary spending.
Weyerhaeuser (WY) -20.4%. Our Q4 and Q1 letters argued that softwood lumber was near the end of a multi-year downturn. Unsurprisingly, when mortgage rates are near multi-decade highs, the demand and outlook for new home construction is weak. Lumber futures fell 12.6% ($617.00 to $539.50), mortgage rates moved above 7%, and housing remains stuck. Should demand recover on the margin, lumber prices could increase materially as every part of the supply chain is lean today, although there is no assurance that demand will recover.
| Winners | Q3 2026 Security Return* | Losers | Q3 2026 Security Return* | |
| PRGO | 42.1% | GLW | -39.8% | |
| NVDA | 14.1% | FNMAS | -31.7% | |
| BAX | 12.6% | CALY | -25.0% | |
| BKRIF | 10.4% | SPHR | -22.3% | |
| HCC | 8.5% | WY | -20.4% |
*Returns shown are for the individual securities discussed, for the period indicated. They are not the returns of any client account, strategy or composite and do not reflect advisory fees, transaction costs or taxes, which would reduce returns. Past performance is not indicative of future results.
Closing
Our approach has not changed: own durable businesses at reasonable prices, hold cash when prices are not reasonable, sell when a position reaches fair value, and size positions so that no single outcome defines the portfolio. As always, we thank you for your trust.
Portfolio Top 5 Holdings
As of September 30, 2026, our five largest positions by market value, including cash, were Cash, NVDA, QCOM, GLW, AER.
As always, we thank you for your trust.
Regards,
Dominick D’Angelo, CFA
585-497-9878
Disclaimer
The performance information included in this letter reflects selected investments and portfolio activity discussed for illustrative and informational purposes only. It does not represent the performance of any specific client account or a composite of client accounts. Individual client results will vary based on factors including, but not limited to, timing of investment, cash flows, portfolio composition, and market conditions.
Returns discussed herein are historical in nature and are not indicative of future results. Past performance does not guarantee future performance. There can be no assurance that any investment strategy or security discussed will achieve similar results or be profitable in the future.
Index returns referenced are provided for general market comparison purposes only. Indices are unmanaged, do not reflect the deduction of fees or expenses, and cannot be invested in directly. Comparisons to indices are not intended to imply that client portfolios mirror or track any benchmark.
References to specific securities, including those identified as winners or losers, do not constitute investment advice, a recommendation to buy or sell, or an indication of all securities purchased or sold for client accounts. The securities discussed may not be held by all clients, and not all investments made by O’Keefe Stevens Advisory were profitable.
This material is not intended as a solicitation or offer to provide investment advisory services except where such services are offered in compliance with applicable law. All investing involves risk, including the potential loss of principal.
O’Keefe Stevens Advisory is an SEC-registered investment adviser. Registration does not imply a certain level of skill or training. Additional information about the firm, including its Form ADV Part 2A and Form CRS, is available upon request and at adviserinfo.sec.gov.
Returns shown for individual securities are for the securities themselves for the period indicated, are not the returns of any client account, strategy or composite, and do not reflect the deduction of advisory fees, transaction costs or taxes, which would reduce an investor’s returns. The securities shown were selected for discussion in this letter and do not represent all securities held, purchased or sold.
This letter contains forward-looking statements, estimates, valuation views and opinions, including statements about possible future events, fair value estimates and expectations for individual securities or markets. These reflect our current beliefs and assumptions, are inherently uncertain, may prove incorrect, and may change without notice. They are not guarantees or predictions of actual results.
Certain information in this letter, including market, index, industry and company data, was obtained from third-party sources, company filings and press reports that we believe to be reliable but have not independently verified. We do not guarantee its accuracy or completeness.
O’Keefe Stevens Advisory, its employees and its clients may own, or may have owned, securities discussed in this letter and may buy or sell them at any time, including in a manner inconsistent with the views expressed here. Holdings and positions are as of the date indicated, vary by account, and are subject to change.
Our portfolios are concentrated in a limited number of securities and may be more volatile than diversified portfolios or broad market indices. Investments in smaller companies, companies with significant leverage, and securities sensitive to interest rates may involve additional risks, including the risk of significant loss.
Advisory services offered through O’Keefe Stevens Advisory, an investment adviser registered with the U.S. Securities & Exchange Commission.
No responses yet