Bill Chen
Key Views & Dialogues
Bill Chen’s follow up on REITs and $ALX
Alexander’s ($ALX) restructured a $300M nonrecourse retail-condo loan at an effective 56% haircut, with ALX buying back the $132.5M A-tranche for 44 cents on the dollar and saving $17.2M in interest annually. Bill estimates $339-340 per share versus a $240 stock, with near-100% dividend coverage after 2028; a Rego 1 sale could trigger a $90M special dividend, while thin liquidity and governance remain watchpoints.
View summary (7)
- Bill Chen returns two days after his last appearance to lay out Alexander’s ($ALX), where a December 29 restructuring saw the lender on a $300M nonrecourse retail-condo loan under the Bloomberg Tower take an effective 56% haircut. The loan was split into A/B/C tranches and ALX bought the $132.5M A-tranche back — “pay us 44 cents on the dollar” — saving $17.2M a year of interest (a 13-14% effective yield), while a $167.5M PIK C-tranche can be forgiven on a below-par refi. Bill: “I basically spent all New Year’s Eve thinking about this deal.”
- The sum-of-the-parts pencils to roughly $339-340 per share against a $240 stock: at today’s price you pay for the Bloomberg office condo net of debt — a 6 cap on a $78.7M triple-net lease running to 2040, ~$1.31B — plus the cash, and get everything else free. “Everything else” is the retail condo ($132M), the Alexander apartment building ($160M), Rego Park 2 ($433M at a 6 cap on ~$26M estimated NOI), and the Rego 1 development site ($100-200M).
- On the obvious governance discount for Vornado control, Bill argues incentives run the other way: Roth, Wight and Mandelbaum own ~46% of ALX versus ~10.3% of VNO. Roth receives ~$12M a year in ALX dividends, while ALX paid him roughly $1M over three years in compensation. Bill says management has ruled out a take-under and Roth has floated selling every asset except the Bloomberg Tower. “Show me the incentives and they’ll show you the results.”
- Bill bought the position up to 13% of his portfolio on New Year’s Eve because he can’t find a surviving short thesis. Dividend coverage is ~90% now and near-100% after the 2028 step-up to $88.3M; the lone sell-side model still carries $47M of cash interest versus his $32M post-deal; and last year’s half-billion of maturities is resolved. His theory: the stock sits on Goldman Marquee as a recommended short, bucketed into “short office, short mall, short New York City.”
- The 2040 lease extension embeds a 2030 rent reset collared to Class A NYC rents — a floor of $85.7M and a ceiling of $104M versus $78.7M today — plus another escalator in 2035. Nearest catalyst: a Rego 1 sale (“under advanced negotiation” per a footnote), which via REIT payout rules on a ~$50M cost basis would force roughly a $90M special dividend, 7-8% of the market cap.
- Follow-on idea: Whitestone (~$14) is the last remaining sub-$3B grocery-anchored shopping-center REIT after Blackstone bought ROIC and Alexander & Baldwin, worth ~$20-21 at those deal cap rates versus a lowball $15.20 private-equity bid. Bill flags one of the highest one-mile household incomes among public REITs and names Blackstone, KKR, Regency, Kimco and InvenTrust as natural acquirers.
- On REIT buybacks the two split: Bill reads them as a high-hurdle signal — rating agencies may put repurchasing REITs on negative watch or issue a negative rating, so “when you do see them actually engage in it, treat that as a signal” — and wants Camden to lift Q3’s $50M to $150M in Q4. Andrew’s counter from Park Hotels: aggressive NAV-discount buybacks (“we think our NAV is 25, we trade for 10”) at ~$15, stock now $10.93 near literal COVID lows, buyback halted.
Bill Chen on the current set up for REITs
Bill Chen’s 6% cap-rate framework yields roughly 5.1–5.2% true unlevered free cash flow; adding 2–3% rent growth and 20% LTV leverage supports about 10–11% permanent-hold returns without a takeout. NAREIT’s 7.3% two-year return reflected 2021 overbuilding and 2024–25 deliveries, but below-average starts suggest the cycle is turning; five simultaneous liquidations and 110–130% gross exposure offer catalysts while leaving picked-over-tail risk.
View summary (8)
- Bill Chen’s rebuttal to the “cap-rate arb never closes” objection is a full total-return algorithm: buy at a 6 cap, knock off 40–50bps of G&A and ~$1,000/door maintenance capex to get ~5.1–5.2% true unlevered free cash flow, add 2–3% rent growth, and layer conservative 20% loan-to-value debt. Under his simplifying assumption that the debt is effectively borrowed at the cap rate, the permanent-hold math reaches about a 10% total return; since many multifamily REITs now trade around the mid-6% range, he says the return approaches 10–11% — no takeout required.
- The reason that algorithm produced only a 7.3% two-year total return for NAREIT is classic capital-cycle theory, and Bill argues the cycle is now turning. Free money in 2021 put “a shovel in the ground” across every asset class except office, deliveries hit in 2024–25 and produced very muted rent growth, but the supply wave is now being absorbed and new-start activity is trending below 20-year averages. AI data centers and infrastructure-related projects are the main exceptions. Because there has been “very very minimal” cap-rate compression, Bill says he is more excited today than on his podcast appearance two years ago.
- Andrew’s case that REITs are “probably a little bit closer to a bond than an equity”: his multifamily NOI forecasts from two years ago landed within 2–3%, versus plus-or-minus 30% for a chemicals EBITDA call. Andrew’s label — “capital cycles for beginners” — and Bill’s kicker: the most multifamily deliveries in roughly 40 years produced NOI down only about 3% over the cycle.
- The corporate-governance and buyback fight centered on managements and boards owning little stock, and Camden’s 10-Q shows $630M of nine-month operating cash flow against just $50M of buybacks versus $310M of development and $334M of acquisitions. Bill defended large blue-chip REITs by citing Camden’s asset sales and buybacks, plus AvalonBay’s roughly $150M of buybacks, while Andrew argued the incremental dollar should go to repurchases. Bill also invoked a fiduciary duty not to let the portfolio age, which Andrew challenged.
- Bill’s anti-buyback Griffin story: the CEO recycled Hartford land through 1031 exchanges into Lehigh Valley warehouses at about 7% unlevered returns, rents later doubled, and the company expanded to Charlotte. Value investors had pressed for buybacks, but after Gordon DuGan joined as chairman the stock popped 20% in one day and the CEO’s strategy received recognition. Griffin was later taken private by Senue and the government of Singapore at double Bill’s cost basis, after growing from under 2M to about 15M square feet. Bill’s point: “not every solution to an undervalued stock is some form of share buyback.” Andrew’s challenge was that the outcome depended on an exceptional operator.
- The takeout wave is real and Bill thinks the private-equity bar is very low: public holders get a 25–40% bump and PE can still win easily. His list includes Inco, Elme Communities, ROIC and Alexander & Baldwin, plus Dream Residential and a hotel-and-resort name; the first four were described as liquidating or taken out, with ROIC and A&B bought by Blackstone. Andrew considered going activist on A&B but rotated capital instead because “it’s such a target-rich environment.”
- Most dislocated corners right now: life science and cold storage (Lineage, Americold, Alexandria at high implied cap rates), anything at 40–50% LTV, and self-storage — “we love our self-storage exposure right now.” Event-driven liquidations are the most dislocated of all on a risk-adjusted basis.
- Five real-estate liquidations are running simultaneously — Bill can’t recall a precedent — and for the first time he’s running 110–130% gross exposure, since asset sales and cash on the balance sheet can return half the capital within one to five months. His underwriting floor is roughly 20% upside, so a 20-to-19.50 low-end revision cuts the MOIC from about 1.22x to roughly 1.15–1.17x. Both flagged that in net-lease office liquidations, the easy-to-underwrite, long-lease assets and multifamily-conversion candidates sold first, leaving a picked-over tail.
⚡️GPT5-Codex-Max: Training Agents with Personality, Tools & Trust — Brian Fioca + Bill Chen, OpenAI
Codex Max combines 24-hours-or-more runtime with self-managed context, showing how performance increasingly depends on a model–harness package optimized for terminal-shaped tools. Packaged agents that Zed, GitHub, and VS Code can invoke may reduce retuning costs, while applied evals, traces, graders, and guardrails remain essential to earning trust as coding agents expand into general computer automation.
View summary (6)
- Codex Max is pitched as both long-running and faster: it can work for “24 hours or more,” manage its own context through compaction inside the Codex harness, and sometimes reach the right answer faster on the same class of problem. Brian has run it locally for longer than a day across several days, with his laptop closed and nobody opening it; “Max” means “speed and maximization, like maximalist,” not merely slower deliberation.
- OpenAI’s product direction increasingly treats performance as a model–harness package rather than an interchangeable raw model. Codex is optimized for terminal-shaped tools, while mainline GPT-5 is broader and more steerable; Brian says Codex is open source and its model is available in the API. Partners found that naming a search tool “rg” rather than “grep” made tool-call performance “jump by a lot.” The host says generalization remains the end goal, framing GPT-5’s generality and Codex’s coding focus as two horizons.
- “Personality” here means reliable engineering conduct—planning, context gathering, progress updates, and checking work—because trust constrains delegation. GPT-5.1 can preface tool calls and be prompted to suppress chatter, while Codex relies on a reasoning summarizer. Communication tokens may be wasteful in a 24-hour cloud run, but they help engineers follow, interrupt, or stop bad trajectories early.
- The abstraction layer is moving from models to packaged agents that other products can invoke, reducing the need to retune a harness for every release. Zed, GitHub, and VS Code are cited as surfaces that can use a whole agent. Codex Max’s context management also supports agents handing work to or spawning parallel subagents, although the pattern is still emerging.
- Applied evals are the operating system for trust, product improvement, and model training—not a leaderboard sidecar. OpenAI wants customer-specific failures and evals, using traces, graders, guardrails, and metaprompting to improve behavior; Bill’s analogy is a “PhD in an API” who still needs a job description, mentorship, and performance reviews. Brian says initial Codex adoption at OpenAI was around 50%; Bill says those users use it every day.
- Coding agents are breaking out of coding into general computer automation, although vision and application access remain bottlenecks. The guests frame a coding agent as “a computer use agent but for the terminal,” already useful for email, media snippets, experiment directories, and desktops. Bill names Devin and Cascade as targets to beat and describes building a Devin for non-coding/email; their 2026 hopes include more computer use, broader subagent systems, UI-only integrations, and enough trust to give ordinary teams capabilities associated with top-tier firms.
Rhizome Partner’s Bill Chen’s post-NAREIT takeaways
Blue-chip public REITs enter the next cycle with 3-8x interest coverage, roughly 25% loan-to-value and seven-year debt available from 4.9% to the high-5% range, unlike private owners underwriting near 1.25x debt-service coverage. With market-rate multifamily starts down 70-80% and supply constrained through at least 2028-29, Bill underwrites roughly 15-18% IRRs for Mid-America and Camden, while governance and capital-allocation discipline remain the key risks to monitor.
View summary (7)
Bill Chen’s central post-NAREIT takeaway is that blue-chip public REITs are entering the next cycle with balance sheets private owners cannot match. The management teams he met reported no distress, interest coverage generally ran 3-8x—7.2x at Mid-America and 6.8x at Camden—and seven-year unsecured debt was available around 4.9% to the high-5% range. Private-market new-debt underwriting, by contrast, often starts at 1.25x debt-service coverage.
A 70-80% collapse in market-rate multifamily starts is setting up a supply-light window through at least 2028-29. Affordable and mission-driven projects may still proceed, while AI data centers are a separate category; Bill thinks rent growth could run at least 3% in 2026-28 if market-rate supply remains constrained. Public REITs can fund development from diversified, roughly 95%-occupied portfolios while private owners have their “hands tied behind their backs.”
Bill still underwrites roughly 18% three-year and 15% four-year IRRs for Mid-America and Camden without heroic assumptions. His model uses 3-4% rent growth and a 5% exit cap rate, while Andrew’s pushback is that a new property trading at a 5% cap does not prove an older portfolio deserves the same mark. Bill’s rebuttal: scale, liquidity, diversification and acquisition optionality arguably warrant a “liquidity premium,” not today’s public-market discount.
New York residential fundamentals are unusually strong precisely because regulation and physical constraints suppress new construction. Bill separated that operating strength from political risk: proposals for indefinite rent freezes or price controls could damage affordable-housing providers, as Washington, D.C.’s eviction restrictions illustrated, yet they might also increase the scarcity value of existing high-end, largely unregulated portfolios. “You can’t really build anything here.”
Grocery-anchored shopping centers are Bill’s next major theme after more than 15 years of underbuilding. Annual supply growth fell from roughly 3% in 1999-2008 to below 0.5% recently, while Sun Belt portfolios can combine 3% contractual escalators, 10-20% renewal lifts and roughly 20% new-lease increases into 4-4.5% NOI growth. Competing projects may require rents to double, making existing centers bought at 7-9% cap rates especially compelling.
Management quality remains the largest caveat to cheap REIT valuations. Andrew highlighted controlled structures and missed buybacks; Bill contrasted Camden’s modest $40 million repurchase with Eurofins buying roughly 5% of its shares in five months, and called Hudson Pacific’s buyback-near-$20/issuance-in-the-mid-$2s sequence “doing everything wrong.” Face-to-face meetings remain crucial because defensiveness, frustration and capital-allocation intent do not appear cleanly in filings.
Bill now sees enough dislocation to construct an entire roster rather than rely on one or two ideas. Mid-America, Camden and FRPH are the “offensive linemen”; grocery centers are running backs; Bill later identified Clipper and COPT as 2%-3% wide-receiver allocations, while also discussing a small Seaport position; and Dream Residential is special teams, with a potential sale against a stated $13.30 NAV. “In today’s environment, we could build a whole team.”