AI Is Filling the Offices It Was Meant to Empty

The short thesis on office REITs was clean: artificial intelligence eliminates white-collar jobs, white-collar workers vacate offices, office landlords spiral into insolvency. Three steps, no ambiguity. It convinced options traders, short sellers, and most sell-side desks. The office REIT index reflected that conviction, trading near its worst levels since the depth of the 2008 to 2009 financial crisis.

The operating data for Q2 2026 does not support a single premise of that trade.

BXP reported Q2 2026 revenue of $895.7 million against a consensus estimate of $860.3 million, a $35.4 million beat. FFO per share came in at $1.78, up from $1.71 a year earlier. The company raised its full-year 2026 FFO guidance to a range of $6.99 to $7.05 per share, with the midpoint landing above the analyst consensus of $6.96. JLL reported that leasing activity across the U.S. office market established a new post-pandemic high in Q2 2026, running 27% above the trailing five-year average. Colliers put Q2 net absorption at 16.9 million square feet nationally, the strongest single quarter in seven years. The supposed executioner of the office market is, by every verifiable metric, its most active demand driver.

Bullet Summary

• BXP Q2 2026 revenue of $895.7 million beat the $860.3 million consensus by $35.4 million; FFO of $1.78 per share rose from $1.71 year-over-year.
• Full-year 2026 FFO guidance raised to $6.99 to $7.05 per share, with the midpoint above analyst consensus of $6.96.
• JLL: U.S. office leasing hit a new post-pandemic high in Q2, up 27% over the trailing five-year average; national vacancy declined 60 basis points quarter-over-quarter.
• Colliers: Q2 net absorption of 16.9 million square feet was the strongest quarterly reading in seven years; year-to-date total reached 24.1 million square feet.
• San Francisco Class A rents rose 4.2% year-over-year to $49.10 per square foot as AI companies drove seven consecutive quarters of leasing above 2 million square feet.
• SL Green reported Q2 2026 Manhattan office occupancy of 94.7% with 53 new leases signed; revenue was $264 million.
• Office CMBS delinquency peaked at 12.34% in January 2026, a record, before easing to 11.71% in March; distress is concentrated in pre-2022 vintage loans on commodity-quality assets, not Class A trophy portfolios.
• BofA raised its BXP price target to $85; Ladenburg raised to $85; Zacks 17-analyst average target stands at $80.94, ranging from $62 to $100.

Market Context: Why This Moment Is Not What the Bears Priced

The macro environment entering H2 2026 presents a specific challenge for office REIT bears: the data points they anchored to are moving in the wrong direction. National office vacancy declined 60 basis points quarter-over-quarter in Q2 per JLL, with availability falling for eight consecutive quarters. Colliers places national vacancy at 18%, down from its 2025 peak, driven by a near-record-low construction pipeline of just 23.4 million square feet under development nationally. Newmark’s Q2 report put vacancy at 19.9%, down 60 basis points year-over-year, with 48 of 62 tracked markets posting positive net absorption. Trophy vacancy fell faster still, dropping 190 basis points year-over-year to 16.7%, per Newmark. The flight to quality that analysts described in 2023 and 2024 as a theoretical dynamic is now showing up in the absorption data.

The RTO backdrop has shifted materially. CBRE’s 2025 Americas Office Occupier Sentiment Survey found that 66% of respondents required employees to be in the office three or more days per week, up from 53% the prior year and 49% two years prior. Average office utilization on the CBRE panel reached 53%, compared to 38% in 2024 and 35% in 2023. Kastle’s badge-access data showed weekly occupancy at 56.3%, with peak days hitting 66%. The labor market context matters: employee resistance to RTO mandates has measurably softened. A survey tracking worker attitudes found that the share of employees who said they would leave over a full-time mandate dropped from 91% in early 2025 to 40% by late 2025, as a cooling job market reduced bargaining leverage. Amazon’s 350,000-person full return in early 2025 removed the operational uncertainty that had kept other large employers cautious.

Interest rates remain the primary macro overhang. The Fed has not pivoted. Elevated borrowing costs extend refinancing pressure on commercial real estate broadly, and BXP’s own debt management requires ongoing execution, not just stated intention. The company has raised $1.2 billion in asset sale proceeds and secured a $1.2 billion construction loan for 343 Madison Avenue. That is progress, but higher-for-longer rates mean the balance sheet discipline must continue. Any macro deterioration that forces a broad credit tightening would spread sentiment risk across the quality spectrum regardless of individual operators’ fundamentals.

Sector Breakdown: Two Markets, One Ticker Symbol

Understanding the office REIT universe in mid-2026 requires accepting that “office” is not a coherent investable category. It is two entirely separate businesses sharing a sector label. The first is trophy Class A space in gateway markets, where AI-driven demand is accelerating, RTO mandates are tightening, and available supply is structurally constrained by a near-record-low development pipeline. The second is commodity B and C space in secondary markets, where structural obsolescence, hybrid work patterns, and refinancing pressure are producing genuine and durable distress. Conflating the two is the analytical error that has kept office REIT multiples depressed through a period of operationally improving trophy assets.

The capital rotation story is playing out in real time. CBRE expects total office investment volume to increase 16% in 2026, with private and institutional investors increasing their office holdings since 2023. The buyers are not buying commodity space. They are purchasing distressed assets with a low basis cost in markets where demand fundamentals support eventual stabilization, or they are acquiring trophy assets where near-term cash flows justify the entry. Either way, the institutional capital moving back into the sector is doing so with a selective, quality-first lens that mirrors what the listed REITs have been saying about their own portfolios for two years.

Sector leadership is concentrated in three publicly traded names. BXP operates 51.1 million square feet across 164 properties in six gateway markets: Boston, Los Angeles, New York, San Francisco, Seattle, and Washington, D.C. SL Green is Manhattan’s dominant landlord. Cousins Properties focuses on Class A Sun Belt assets across Austin, Atlanta, Phoenix, Charlotte, Tampa, Houston, Dallas, and Nashville. These are not interchangeable with the distressed secondary-market operators that have largely exited the public markets. Paramount Group, City Office, and Office Properties Income are not the same trade as BXP. Treating them as such because they share a sector classification is a category error with real cost.

Stock-Specific Financial Breakdown

BXP: The Operational Case That Has Not Re-Rated

BXP’s Q2 2026 numbers are the clearest challenge to the extinction thesis. Revenue of $895.7 million beat consensus by $35.4 million. FFO of $1.78 per share rose year-over-year and beat guidance. The company’s pipeline under letters of intent reached approximately 3.5 million square feet as of the Nareit REITweek conference in June, supporting management’s stated target of raising occupancy from 87% to 89% by end of 2026 and by another two percentage points in 2027. BXP CEO Owen Thomas noted that leasing activity continued to strengthen across all markets, led by New York City, with over 1 million square feet leased in Q1 and over 800,000 square feet signed in Q2 alone.

The Q1 2026 data was already telling. BXP completed 68 leases totaling more than 1.1 million square feet, including approximately 140,000 square feet at 360 Park Avenue South and roughly 104,000 square feet at 680 Folsom Street in San Francisco. BXP President Douglas Linde described AI-oriented companies as absorbing the majority of incremental space and noted they were aggressively hiring while treating in-person work as central to their business model. That characterization aligns exactly with the leasing volume data. Midtown South in Q1 2026 captured as much AI demand as the entire first half of 2025.

Valuation context: BXP’s stock opened near $68 as of mid-August. BofA raised its target to $85. Ladenburg is at $85. Zacks reports a 17-analyst average price target of $80.94, with a range from $62 to $100. The stock is trading approximately 20% below the consensus target and more than 25% below the most bullish institutional targets, during a period when the company is beating on revenue, raising guidance, and delivering its strongest leasing pipeline in years. BXP’s development pipeline stands at $3.6 billion. Available sublease inventory nationally has fallen 28% from its cyclical peak. New office deliveries hit a 14-year low. The supply constraint that should underpin premium rent growth is already in place.

SL Green: Manhattan’s Occupancy Story

SL Green reported Q2 2026 revenue of $264 million and Manhattan office occupancy of 94.7%, with 53 new leases signed during the quarter. The occupancy number is the relevant figure for traders evaluating the gap between the stock and its fundamentals. A 94.7% occupied Manhattan trophy portfolio is not a distressed asset. The company reported a wider net loss of $20.51 million, which reflects the interplay of depreciation, interest expense, and non-recurring items rather than operating cash flow deterioration. SLG continued its buyback program with $14.1 million repurchased in Q2, bringing total share retirement since 2016 to approximately 48.6% of the share count. That per-share metric concentration, combined with the SUMMIT observatory expansion into Tokyo, reflects management’s confidence in the long-term asset quality even while reported earnings remain under pressure from debt costs.

The San Francisco Signal and Why It Matters

San Francisco was designated ground zero for AI-driven office apocalypse. The 2026 data makes that designation look like the market’s most expensive analytical error of the cycle. Kidder Mathews research confirmed that AI companies drove seven consecutive quarters of leasing activity above 2 million square feet in San Francisco. Financial District Class A and B asking rents rose 4.2% year-over-year to $49.10 per square foot per their Q2 2026 report. JLL’s Bisnow-cited research found that the nearly 770,000 square feet of Class A space taken by AI companies in Q1 alone drove positive net absorption of all office space to 1.6 million square feet, the strongest quarter in eight years, and pulled overall vacancy to 32.6%.

The tenant list is not abstract. Anthropic signed a 420,000-square-foot lease at 300 Howard Street in January 2026. OpenAI’s San Francisco footprint moved toward 1 million square feet in Mission Bay. Together.ai took 149,000 square feet at 2 Henry Adams Street in Q2. Reflection AI leased 25,000 square feet at 140 New Montgomery in South of Market, helping push that building to 94% occupancy. Hex Technologies more than doubled its Financial District footprint. JLL’s Alexander Quinn noted that robotics and AI-centric drone companies are expected to sign deals for an additional 1.5 million square feet in 2026, drawn by the concentration of AI and robotics engineering talent in the Bay Area. JLL has forecast that AI companies will lease 7 million additional square feet in San Francisco between 2026 and 2030.

In 2025, the Bay Area claimed 14 of the 100 largest office leases in the United States, totaling 4.3 million square feet, per CBRE data cited by the Silicon Valley Business Journal. AI firms now occupy at least 6 million square feet across Silicon Valley alone. VTS forecast San Francisco office demand growth of 15% in 2026, placing it at the top of the U.S. market ranking. The city the bears wrote off is leading the national recovery by nearly every measure that matters to Class A landlords.

The CMBS Distress Question: Lagging Signal, Not Leading One

The CMBS delinquency data deserves specific treatment because it is the most frequently cited pillar of the bear case and the most frequently misread. Office CMBS delinquency peaked at 12.34% in January 2026, per Trepp, a record high exceeding the prior cycle peak of roughly 10.7% reached in late 2012 after the global financial crisis. It eased to 11.71% in March but remains at historically elevated levels. The MBA reported that overall CMBS loan delinquency stood at 5.21% as of Q1 2026, up from 4.97% the prior quarter, with office, lodging, and retail seeing the largest increases.

These numbers are real. The question is what they are measuring. Colliers and Trepp both characterize the current distress as concentrated in refinancing pressure on pre-2022 vintage loans, not in operating cash flow failure at the asset level. Per CRE Daily’s Trepp analysis, the record January delinquency rate was heavily influenced by two large single loans: Worldwide Plaza at $940 million and One New York Plaza at $835 million moving to delinquency. Concentration risk, not a broad-based operating collapse, is moving the headline rate. Morningstar DBRS estimated that more than $100 billion in fixed and floating CMBS loans would come due in 2026 across all asset types, with office maturity defaults representing the primary driver of elevated rates. But maturity defaults on 2018 to 2021 vintage loans, originated during the peak of pre-pandemic optimism and now unable to refinance at 2026 rates, are a different problem than demand deterioration at Class A assets. Excluding CMBS, the MBA found that delinquency rates across banks, life insurance companies, and GSEs remain below 1.25%.

The pricing reset that follows distressed asset resolution lowers the ownership basis for new buyers. That lower basis allows competitive rental pricing while still generating acceptable returns. The distress wave is not destroying the office sector’s economics. For quality assets, it is recapitalizing the sector at levels that the prior vintage of overleveraged loans had prevented.

Technical and Trading Framework

BXP has been trading in the middle of its 52-week range and above its 200-day simple moving average as of mid-August, per CNN Markets data. That technical positioning reflects a stock that has not yet broken out from its post-earnings range despite delivering a material beat. Volume patterns following the Q2 earnings release showed a muted response relative to the magnitude of the operational outperformance, which is consistent with a market still weighted by sector-level sentiment rather than company-specific fundamentals.

Key levels for traders: The $68 to $70 range has served as near-term support since the earnings release. The analyst consensus cluster sits between $80 and $85 for the more bullish houses, with BofA and Ladenburg both at $85 following post-earnings target raises on August 3 and July 30, respectively. Citi raised its target to $69 from $58 on August 4, a meaningful upward revision but still below the stock’s pre-rate-cycle levels. Deutsche Bank holds at Hold. The divergence in analyst positioning, with targets ranging from $62 to $100, reflects genuine uncertainty about the pace of re-rating rather than disagreement about whether the assets are performing.

Implied volatility on office REIT names remains elevated relative to the broader REIT universe, consistent with a market that is pricing ongoing headline risk from CMBS delinquency data and AI labor displacement uncertainty. When IV is elevated against a backdrop of operationally improving fundamentals, premium sellers have a structural advantage over premium buyers. The expected move framework for Q3 earnings must account for the asymmetry: three consecutive quarters of operational beats have not produced a meaningful re-rating, which either means the market is right that risk is higher than the operating data suggests, or it means the next catalyst carries compressed re-rating potential that has accumulated across multiple quarters of underperformance relative to fundamentals.

Scenario Modeling

Bull Case

Conditions: AI leasing velocity in San Francisco and New York sustains through H2 2026. BXP’s 3.5 million square foot LOI pipeline converts at the historical rate. The Fed signals a rate normalization path that eases refinancing pressure on the sector. A string of in-line-to-better Q3 and Q4 quarters breaks the pattern of operational beats failing to produce stock re-rating.

Price levels: BXP moves toward the $80 to $85 consensus cluster over a six to nine month horizon, consistent with the BofA and Ladenburg targets. A defined-risk bull call spread targeting the mid-$70s through January 2027 expiry captures the re-rating move at a fraction of the outright equity exposure, and limits downside to the premium paid. The risk to this structure is that IV compression erodes the value of the long leg if no catalyst arrives before decay accelerates.

Base Case

Conditions: The operational improvement at BXP and SLG continues at the current pace. AI leasing remains robust but does not accelerate meaningfully beyond the current trajectory. The Fed holds rates steady. CMBS distress resolves gradually through modifications and asset transfers without a contagion event that spreads to trophy-asset pricing.

Price levels: BXP continues to trade in the $65 to $75 range, with occupancy improving toward the 89% target by year-end and the signed-not-occupied pipeline providing near-term visibility. The re-rating is gradual rather than event-driven. Quarterly dividend payments and occupancy trajectory progress are the primary return drivers. A cash-secured put at the $65 strike through September or October expiry captures premium decay against a stock that has demonstrated operational resilience at current levels.

Bear Case

Conditions: A large CMBS liquidation event in Q3 or Q4 triggers indiscriminate sector selling that depresses even trophy-asset names. AI VC funding velocity decelerates sharply, reducing the primary demand driver in San Francisco. Generative AI begins to produce measurable white-collar headcount reductions at major BXP tenants in financial services and legal, reducing renewal probability.

Price levels: BXP tests the low-$60s, with the $62 floor from Deutsche Bank’s most cautious target providing reference. Bear exposure is better concentrated in names with higher leverage, near-term debt maturities, and genuine commodity-quality exposure. The short is already crowded: mortgage and office REITs carry the highest short interest among small-cap real estate stocks per July 2026 data. A squeeze remains possible even on negative headlines if those headlines are less severe than the short base has priced. Isolating bear positions to the genuinely impaired operators rather than the trophy-asset names is where the fundamental support for the trade is strongest.

Active Trader Strategy Framework

The central positioning consideration for office REIT exposure in H2 2026 is sizing against the elevated IV environment while targeting expirations that allow Q3 earnings to serve as a catalyst. BXP’s signed-not-occupied pipeline conversion through H2 is the near-term occupancy catalyst; it is also the most visible and trackable data point between now and the next earnings print.

• Distinguish trophy-asset exposure from commodity exposure before sizing any position. BXP (Class A gateway markets), SLG (Manhattan premier), and CUZ (Sun Belt Class A) carry fundamentally different risk profiles than distressed secondary-market operators. They should not be sized or hedged the same way.
• Track CMBS maturity resolution. Large distressed sales in Q3 or Q4 can create entry points on quality names through indiscriminate sector selling, providing lower-basis opportunities for traders who have done the work on asset quality differentiation in advance.
• Monitor AI VC funding velocity in San Francisco and New York. The direct line from venture funding to hiring to leasing demand is the most reliable leading indicator for gateway-market absorption, but it is not instantaneous. Watch for any deceleration in funding rounds as a forward signal.
• Watch BXP’s LOI pipeline conversion. The 3.5 million square feet under letters of intent provides management’s occupancy targets with near-term visibility. Pipeline conversion into executed leases is the event that turns guidance into cash flow.
• For any bear positioning on the sector, isolate it to names with high leverage, near-term debt maturities, and genuine commodity-quality exposure. Shorting trophy-asset operators at crisis-era multiples while AI firms sign flagship leases in their core markets is the crowd trade. It is also the most fundamentally exposed position in the sector right now.
• Volatility expectations: heading into Q3, IV is likely to remain elevated given the CMBS maturity calendar and ongoing macro uncertainty. That elevated IV environment favors premium-selling structures on names with demonstrated operational resilience over outright directional bets that pay for the same elevated volatility.

Risk Factors: What the Contrarian Case Must Hold Honestly

Three legitimate risks demand precise acknowledgment, not dismissal. First, the CMBS maturity calendar is real and unresolved. Morningstar DBRS estimated that more than $100 billion in fixed and floating CMBS loans come due across all property types in 2026. Any large office asset liquidation or forced sale can generate headline risk that produces indiscriminate selling across the quality spectrum, even when the liquidated asset is a fundamentally different product from BXP’s trophy portfolio. Markets do not always discriminate quickly. Second, the AI labor displacement thesis is a probability distribution, not a binary. BXP management acknowledged on the Q2 call that the long-term labor impacts of AI remain difficult to predict. If generative AI materially reduces white-collar headcount at law firms, financial services companies, and professional services tenants over the next three to five years, the demand base for office space could contract in ways that AI firm leasing does not fully offset. The uncertainty is real; the magnitude and timing are not knowable with current data.

Third, the debt management story requires execution. BXP has raised $1.2 billion in asset sale proceeds and secured major construction financing. Those are positive steps. But a higher-for-longer rate environment extends the refinancing pressure cycle and demands ongoing capital discipline that cannot be assumed in advance. The company’s balance sheet management is better than the worst-case scenarios the market has priced, but it is not yet resolved.

The counterargument is pricing. Office REITs are trading at multiples that imply a severe deterioration in fundamentals. The Q2 2026 data, across multiple independent research providers, shows the opposite. When the market prices the worst-case scenario and the operating data delivers something materially better, the gap is where the return lives. That gap has persisted through several consecutive quarters of operational beats without closing. The persistence of the discount is itself a risk factor, because it implies the market has information the operating data does not fully capture. Disciplined traders account for that possibility rather than dismissing it.

Forward Outlook

The defining dynamic for office REITs through the remainder of 2026 is supply scarcity meeting accelerating demand. Colliers places the national construction pipeline at a record low of 23.4 million square feet. Newmark puts construction at 16.3 million square feet, roughly 85% below its Q1 2020 peak. The pipeline for new Class A space is effectively nonexistent in most gateway markets. Cushman and Wakefield’s rolling four-quarter absorption total hit 14.3 million square feet in Q2 2026, the strongest reading since 2020 and the seventh consecutive quarter of improvement. CBRE put Q2 net absorption at 12.6 million square feet, the ninth consecutive quarter of positive demand, with a four-quarter total of 38.9 million square feet.

BXP’s occupancy target of 89% by year-end 2026 and 91% by end of 2027 is supported by a 3.5 million square foot LOI pipeline that gives the trajectory near-term visibility. San Francisco Class A rents rising 4.2% year-over-year to $49.10 per square foot signals that the pricing environment at the quality end of the market is tightening, not loosening. JLL expects AI companies to lease 7 million additional square feet in San Francisco through 2030, while robotics and physical AI firms are forecast to add 1.5 million square feet of additional Bay Area leasing in 2026 alone. Total office investment volume is expected to increase 16% in 2026, per CBRE, as institutional capital re-engages with a sector that has repriced to levels that make the economics work.

The crowd’s thesis was not irrational. AI will displace some white-collar work. The probabilities around that displacement are genuinely uncertain. But the companies building AI are simultaneously signing some of the largest office leases in the history of several gateway markets. The market has priced the first half of that sentence and largely ignored the second. When that pricing gap corrects, the people who sold office REITs because AI kills offices will discover they sold them to the companies that are filling them.

Preparation Over Prediction

The office REIT trade in H2 2026 is not about predicting whether AI ultimately destroys or saves the sector. That question will resolve over years, not quarters. The trade is about whether the current price reflects the current operating reality. By every verifiable metric available as of Q2 2026, it does not. BXP beat on revenue by $35 million, raised guidance, and posted its strongest leasing pipeline in years. SL Green is 94.7% occupied in Manhattan with 53 new leases signed. Colliers recorded the highest quarterly net absorption in seven years nationally. The market is pricing a severe scenario that the data has not delivered for at least two consecutive years.

Disciplined risk management means sizing against that uncertainty, not ignoring it. The CMBS calendar, the interest rate environment, and the AI labor displacement probability distribution are all real considerations. They belong in every risk framework applied to this sector. But preparation means knowing the levels, understanding the structures that limit downside while preserving upside optionality, and distinguishing trophy-asset quality from commodity distress before the next catalyst arrives rather than after. The data is available. The work is knowable. The gap between price and operating reality is measurable. What happens next is not certain. What is certain is that uncertainty and mispricing are not the same thing, and at current levels, they appear to be coexisting in the same sector.

For informational and educational purposes only. Not investment advice. Trading involves risk, including loss of principal.