DiamondRock posted $0.27 in adjusted FFO per diluted share for Q4 2025, beating consensus by $0.03. Net income hit $23.8 million for the quarter, up 273.7% year-over-year. The board raised the quarterly dividend to $0.09 from $0.08. The headline reads like a victory lap. The 2026 guidance reads like a warning label.

Full-year 2026 adjusted EBITDA is projected at $287 million to $302 million. The midpoint of that range is $294.5 million. Full-year 2025 actual was $297.6 million. That's a midpoint decline of roughly 1%. RevPAR growth guidance is 1% to 3%, which sounds fine until you remember that 2025 comparable RevPAR grew just 0.4%. So the company is guiding for acceleration in revenue per room while simultaneously guiding for flat-to-lower EBITDA. The only way those two numbers coexist is if cost to achieve is rising faster than revenue. That's the number behind the number.

The preferred stock redemption is the move worth studying. DRH retired all 4.76 million shares of its 8.25% Series A preferred in December, spending $121.5 million in cash. At 8.25%, that preferred was costing roughly $9.8 million annually. Eliminating that obligation is pure accretion to common equity... but it also burned a significant cash position. Pair that with 4.8 million common shares repurchased during 2025 at an average of $7.72, and you're looking at a company that deployed over $158 million in capital on balance sheet cleanup rather than acquisitions. That's a statement about where management sees better value: in their own stock versus what's available in the transaction market. At $7.72 average repurchase against a portfolio trading at $257K per key versus $440K adjusted replacement cost, the math supports the buyback. But it also means DRH is choosing financial engineering over portfolio growth at a point in the cycle where others are buying.

An owner I sat across the table from once told me, "I'm not worried about the quarter. I'm worried about the year after the quarter everyone celebrates." He was talking about a different REIT, but the pattern is identical. DRH's 2025 was strong on earnings per share because of share count reduction and preferred elimination, not because of NOI growth. Adjusted EBITDA was essentially flat year-over-year (down 0.1%). Free cash flow per share grew 6%, but decompose that and the growth came from fewer shares outstanding, not from more cash flow. That's not a critique of the strategy... it's a description of the mechanism. Investors pricing DRH on FFO per share growth should understand that the growth engine is capital return, not operating improvement. Those are different durability profiles.

The Altman Z-Score sitting at 0.97 is the line item that should keep asset managers honest. Below 1.8 is the distress zone. DRH isn't in crisis, but a Z-Score under 1.0 for a lodging REIT with 35 properties and flat EBITDA guidance means the margin for error on cost management in 2026 is thin. If RevPAR comes in at the low end of guidance (1%) and labor costs track the industry projection of 3% growth, the EBITDA floor of $287 million starts looking optimistic. Check again.

J.P. Morgan analyst Daniel Politzer says 2026 is the year hotel AI spending starts paying off. The source article doesn't break out the exact capital allocation, but the major brands are directing meaningful portions of their technology budgets at AI-adjacent transformation. Let's decompose that.

The bull case relies on a few data points that keep circulating. Hyatt claims 20% greater productivity in group sales teams using AI tools. Wyndham says AI-powered call centers are cutting labor costs for franchisees. A Deloitte study (sourced from vendor-friendly research, which I always flag) claims 250% ROI within two years, driven by 15-20% staffing savings and up to 10% RevPAR lift. Those numbers are doing a lot of heavy lifting. A 10% RevPAR boost from AI-based pricing at a 200-key select-service running $95 RevPAR is $9.50 per room per night... $693K annually. Against what implementation cost? The research doesn't say. Nobody's showing the denominator.

Here's what the headline doesn't tell you. "Productivity gains" in group sales don't flow directly to EBITDA unless you reduce headcount or close incrementally more business with the same team. Hyatt hasn't specified which one. A 20% productivity number without a corresponding revenue or labor line item is a metric without a home on the P&L. I've audited management companies that reported "efficiency improvements" for three consecutive years while GOP margins stayed flat. The improvements were real. The earnings impact wasn't. Same structure here... until someone shows me the flow-through, the productivity number is a press release, not a finding.

The franchise owner's math is where this gets uncomfortable. Wyndham's AI call center savings accrue to the franchisee, which is genuinely interesting... if the franchisee isn't simultaneously absorbing a technology fee increase that offsets the labor reduction. I analyzed a portfolio last year where the management company rolled out an "AI-enhanced" revenue management layer. The software cost $4.20 per room per month. The incremental RevPAR gain over the existing RMS was $1.80 per occupied room at 68% occupancy... roughly $1.22 per room per month. The owner was paying $2.98 per room per month for the privilege of saying they had AI. Check again.

The real number here is not whether AI creates value in hotels. It does. Dynamic pricing has been creating value for 15 years (we just called it revenue management). The real number is whether 2026 AI spending generates returns that exceed the cost of capital for the owners funding it. J.P. Morgan is a sell-side firm covering publicly traded hotel companies. Their job is to tell investors the story is getting better. The owner at a 150-key branded property writing checks for technology mandates needs a different calculation... one that starts with total cost deployed and ends with actual incremental free cash flow. That calculation is conspicuously absent from every AI earnings narrative I've read this quarter.