Why You Shouldn't Be Paying the AI Startup Rate

The San Francisco office market is white hot. Demand is at a record high. Huge AI companies like OpenAI and Anthropic grab headlines when they take down entire buildings. But it’s the steady surge of smaller startups that pushes demand to its current highs.  

38% of the leases completed in Q1 2026 were with AI companies. At the top end, these companies are competing for large blocks (such as Anthropic’s lease of all of 300 Howard Street), a market that is increasingly scarce.  At the low end, series A and B startups are scrambling to secure well-located, pre-built (even furnished) spaces they can lease immediately. Speed is a key driver. They will pay a premium for occupancy-ready space.

Both ends of the AI demand spectrum present risk to the landlord.  Many of these companies end up paying a risk premium.  The rents they pay define the market.  Beneath the surface, landlords are aware of the risk. Even OpenAI, the largest AI tenant in San Francisco, is high risk due to its massive compute spend, which requires it to continue raising large rounds of funding.

Your profitable, stable company presents a decidedly different risk profile to the landlord. You should not be paying the risk premium. We’re beginning to see landlords favor stability. After all, this is San Francisco. The market has a long history of boom-and-bust tied to tech demand. The key is to negotiate from a position of strength, aligning your occupancy with stability. Let the guys with 12 months of burn pay the premium.

Landlord as Bank: The Hidden Cost of "Convenient" TI Financing

The cost to build office space is at an all-time high, forcing companies to make deliberate decisions about how tenant improvements are funded. These decisions directly impact cash, balance sheet, and EBITDA—and should not be left solely to the real estate team.

For companies where valuation matters, structure matters. A business preparing for a sale, for example, may choose to fund all or a portion of the improvements with cash to preserve EBITDA, given valuation is often tied to an EBITDA multiple.

Sometimes when there is a shortfall between the tenant improvement allowance a landlord has offered as a concession to the lease and the total cost to build the space, the landlord will offer to finance the difference.

At first glance, this may appear to be an efficient solution. But the details matter.

If structured as a true loan—separate from the lease—the tenant can capitalize the improvements, record debt, and keep rent lower. This is typically more favorable from an EBITDA perspective.

But most landlords aren’t truly interested in acting as a lender. Their real motivation is to optimize for asset value.

They do so by embedding the additional funding into the lease as rent. This step makes the cost of the financing more expensive to the tenant while turbo-charging the value it creates for the landlord. How? By adding the loan value to rent, it becomes subject to the annual rent escalations common in most leases (typically 3%), further compounding the cost of the loan. Most importantly, the increased rent drives higher net operating income, which directly increases the landlord’s asset value upon sale.

Tenants must carefully assess the implications of landlord offers to finance additional tenant improvements, as the proposed structures often carry hidden costs.

Why Non-Tech Companies Need More Time to Lease Office Space in San Francisco

At roughly 86 million square feet, the San Francisco office market is not particularly large. When you break it down by submarkets, building class, or premium view space, it becomes even smaller. With approximately 8 million square feet of active demand, much of it concentrated in the best submarkets and best buildings, the leasing environment can become challenging for companies that want to make thoughtful, well-informed decisions.

Technology companies, especially AI firms, represent the largest share of that demand. But San Francisco is home to many companies outside the tech sector. These businesses must often operate in a market shaped by the behavior of fast-moving technology tenants.

That dynamic creates friction.

Tech companies frequently move faster and are often willing to pay more to secure the right space. Historically they have absorbed space quickly and sometimes with less sensitivity to deal terms. It is not that terms do not matter to them. Their priorities are simply different. In the technology economy, speed often determines the winners. Companies race to scale and investors continue to provide enormous capital to the firms they believe will get there first.

For more mature, non-tech businesses, this can make the leasing process difficult. Space they carefully evaluate can disappear overnight when a technology company decides to move faster or pay more.

So what should these companies do?

Allow more time for the leasing process.

Time creates flexibility. It allows companies to evaluate options thoroughly, negotiate with multiple landlords, and pivot when opportunities disappear. When tenants lose space to faster-moving competitors, the real problem is rarely the leasing strategy. The problem is usually a lack of time to recover and pursue alternatives.

Starting early does not mean starting blindly. Begin too early and the process can lose momentum. But if companies are going to make a mistake on timing, it is far better to err on the side of starting too early rather than too late.

Because in San Francisco’s office market, time is not just part of the leasing process.

It is often the single greatest source of negotiating leverage an occupier has.

Block’s Layoffs: Validating the Citrini Thesis, or Solving for Gross Mismanagement?

Last week, Jack Dorsey, CEO of Block, Inc., announced the company is laying off a whopping 40% of its workforce, more than 4,000 employees. Coming on the heels of the Citrini Memo, it is difficult not to at least consider the parallels between Block’s actions and the fictional scenarios portrayed therein. Indeed, Dorsey’s commentary on the matter reads as if taken directly from Citrini’s dystopian narrative:

“The core thesis is simple. Intelligence tools have changed what it means to build and run a company. I don’t think we’re early to this realization. I think most companies are late. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes.”

In the aftermath of this announcement, a number of people, including former Block employees, have argued the layoffs are really about eliminating corporate bloat. AI, they suggest, is simply a convenient narrative that creates better optics by making the company appear to be getting ahead of a meaningful trend, rather than correcting for poor management decisions that resulted in massive overhiring.

I have questions.

If Dorsey’s stated case for the layoffs is valid, does this not align squarely (pun intended) with Citrini’s doomsday scenario? Alternatively, if this is really about correcting corporate bloat, how did Block management get so far off track as to add 40% more employees than necessary to run the company effectively?

To be sure, Dorsey makes clear that “gross profit more than doubled from the first quarter to the fourth quarter of 2025.” He goes on to write, “We believe this financial performance is just beginning to reflect the product development velocity improvements we drove this year.”

Is it possible Block generated $2.87 billion in profit while carrying $235 million in excess labor spend? Or is it more plausible that AI has already automated workflows that previously required large teams, making certain roles expendable?

The answer may lie somewhere in the middle. Yes, Block likely over hired. And yes, AI may now be enabling the company to automate work previously done by humans.

Either way, we will all be watching closely for signs that Dorsey’s prediction proves correct: that “the majority of companies will reach the same conclusion and make similar structural changes.”

One thing is certain. If AI-driven workforce reductions approach anything close to the scale of Block’s recent layoffs, and if similar levels of job elimination become commonplace, we should all be concerned.

A Wild Ride

Speculation about the economic impact of AI is everywhere. From economists to technologists to the barista at your local coffee shop, everyone has an opinion about what the future holds. Just this week we saw two sharply contrasting perspectives: one from Alap Shah and Citrini Research projecting widespread job loss and severe economic fallout, and another from Citadel pushing back on the logic behind the Citrini thesis.

Few technologies have generated this level of speculation and fear. Even Citrini’s admittedly fictional portrayal of the next two years had an immediate, very real impact on markets, sending stocks sharply lower. Forecasts about a future in which AI reaches its full potential swing wildly from dystopian to utopian. Investors do not know what to believe.

Take Nvidia. Imagine Jensen Huang huddled with his executive team before an earnings call: “Great work, everyone. With this historic performance, the stock should respond positively.” And then it drops 9 percent. The numbers are strong, yet the reaction is negative. That is not about earnings. It is about nerves. Investors want to believe in an AI-driven future, but the doomsday narrative is hard to ignore. What if the pessimists are right?

Uncertainty is the defining theme of this moment. That likely makes it one of the most opportunistic periods in modern economic history. It is a time when a bold bet could generate enormous gains or painful losses. There appears to be little middle ground. The stakes feel binary.

My view is that the future sits well above most of our pay grades, certainly mine. Most of us are too far removed from the underlying data and too unfamiliar with the technical architecture to see what the builders see. We experience AI through consumer outcomes, which are undeniably impressive and advancing quickly. More recently, we have begun hearing about agentic systems and breakthroughs from tools like Claude and Moltbot. Yet for all the progress, AI adoption in everyday life has not fully crossed into game-changing territory.

This is a wild ride. Buckle up. The future feels more uncertain than at any point I can remember.

Defining Work and Life

Companies have long sought to enhance employee productivity by reducing daily friction. This friction arises both in the execution of work itself and in the activities tangential to work such as commuting, eating, childcare, and laundry. Interestingly, medieval economies in Europe were organized around self-sufficient manors where life and work were inseparable. Similar models existed throughout the world, including in China and the Middle East. For most of human history, working was living.

That began to change in the nineteenth century. As new modes of transportation were adopted, including rail and later the automobile, white-collar workers gained the ability to live farther from where they worked. The expansion of highways accelerated this trend and gave rise to the modern commuter suburb. At the same time, advances in technology and economic systems raised living standards and created more discretionary time outside of work. Work and life began to evolve as distinct concepts. As that separation took hold, people increasingly perceived friction when work encroached on life.

In more recent decades, ambitious employers have sought to maximize employee engagement by reducing the tangential friction surrounding work. The technology sector has led this effort, developing campus environments where employees can address daily needs on site, from healthcare and childcare to meals and laundry.

The pandemic temporarily disrupted this trajectory, creating the largest separation between work and life ever experienced at scale. Today, however, the AI sector is once again redefining the relationship. With enormous stakes and a rapid pace of change, many AI companies are promoting an intense work culture, often embracing a 996 model in which employees work from 9 a.m. to 9 p.m., six days per week. Such a commitment leaves little room for life outside work and is incompatible with long commutes. It is therefore unsurprising to see some AI companies renting apartments to provide employees with housing close to the workplace.

These are unusual times. On one hand, many are experiencing a profound imbalance in which work dominates life. On the other, AI holds the promise of replacing labor at scale, potentially creating more free time than humanity has ever known. How this tension resolves is difficult to predict. For now, as we race toward an uncertain future, many are experiencing life as work once again, echoing patterns from long ago.

Rising Tide Not Lifting All Boats

AI companies are leasing office space in San Francisco at an accelerating pace. Demand spans the spectrum, from Series A startups to Anthropic, which recently signed a lease for just under 500,000 square feet. Yet the market remains deeply bifurcated. The best buildings are seeing real competition. Commodity assets are not.

This creates an existential question for owners of smaller buildings that lack views, scale, or high-end amenities: how do you stay relevant?

There are two viable paths.

The first is to lean into AI, but with discipline. Owners of commodity buildings should not expect to win mature AI tenants. That ship has sailed. The opportunity lies earlier. Early-stage startups struggle to find space that is move-in ready, flexible, and available on short terms. That segment is underserved. It carries more risk and higher churn, but it also delivers faster velocity and higher revenue per lease. Buildings that solve this problem can remain competitive by embracing flexibility rather than fighting it.

The second path is to zig while the market zags. Ignore AI entirely. Position the building as a refuge for tenants being priced out of a tech-dominated market. This strategy works best for Tier 3 assets with a low cost basis. The advantage is price. By leasing at a meaningful discount, these owners can capture demand from companies shut out of Tier 1 and Tier 2 buildings. The key is focus. Compete on affordability, not amenities you cannot credibly deliver.

The takeaway is simple. AI is not lifting all boats. It is widening the gap. Owners who understand which side of that divide they’re on, and act accordingly, can still differentiate their offering and achieve leasing success.

What Lies Ahead

It’s hard not to witness the rapid emergence of AI in the workplace without pondering our near future. Technology has long served as both catalyst and accelerant in shaping how we work. But the major catalysts of prior eras, things like the train, the telephone, and even the internet, took years to deliver change at scale.

With the recent introduction of Clawdbot, now Moltbot, people are beginning to experience the next iteration of AI beyond LLMs and vibe coding. We are starting to see how an AI agent can be deployed to do many of the things we do, only faster and better. While technologists have been working on AI for decades, advancements over the past several years have been swift. Indeed, the average information economy worker is barely keeping pace. Many still think AI is simply about better search. Frankly, search is AI’s least interesting and least impactful application.

I remember the many ways the internet crept into my life, both personal and professional. It did not feel like there was a single game-changing moment. Instead, there was a steady progression until one day it was everywhere. To be sure, it displaced workers and reshaped entire industries, think Amazon and bookstores. But I never recall thinking, “Wow, this is going to change everything.”

I think we could be in such a moment now. A moment when we can barely wrap our heads around what lies ahead. It is one thing to contemplate the next decade of my own career. But what about my young children? What will work look like for them? I have yet to hear anyone offer a coherent explanation. Instead, I read provocative comments from people like Elon Musk suggesting that people should forget about saving for retirement because it will not be necessary. Others suggest that large percentages of the work we do will soon be done by AI. And then what?

The LLMs felt like a meaningful advancement, but not a true game changer. While I have not yet been able to use Moltbot, I have read reports from those who have. Many are breathless and genuinely blown away by its effectiveness. Just this past week, publicly traded software companies saw their stocks hammered in reaction to agents like Moltbot. Maybe that response is reasonable. After all, why would we continue to buy one-size-fits-all software that is expensive and exposes us to variables outside our control, when an AI agent could build a custom solution that we own and control?

I do not know what lies ahead. But I do know one thing. It will not be boring.

San Francisco Office Is Hot (Again)

It can be confusing to understand the leasing dynamic in a city like San Francisco. The data suggests a market in which occupiers should enjoy outsized leverage. Vacancy remains north of 30%, after all. Yet many companies are surprised to encounter real competition for space and rental rates at or near all-time highs. How can both be true?

The answer lies in how that vacancy is distributed. Excess vacancy is concentrated in assets that are either fundamentally inferior or burdened by a distressed capital stack. Occupiers are largely bypassing inferior buildings as they focus on higher-quality environments to support a return to the office. At the same time, buildings with broken capital stacks often cannot transact at market terms, rendering them largely irrelevant to active tenants. As a result, effective vacancy for quality space is far lower than headline figures suggest. In the Class A premium segment, vacancy is closer to 5%.

As more companies work to bring employees back, a new narrative is emerging. The office must be well designed, well located, and rich in amenities. Employers recognize the friction inherent in return-to-office initiatives and want to get it right. Increasingly, they are willing to pay more for space that delivers a compelling experience. This same logic explains why several large-scale office developments are actively pursuing tenants with plans to build in the near term.

Leasing volume reached historic highs in 2025, and anecdotal evidence from early 2026 points to another year of strong demand.The “generational” leverage tenants enjoyed from 2022 through 2024 has largely evaporated. We expect tenant leverage to continue eroding in cases where occupiers are competing for high-quality space.

A New Product

Office space is a product. In the US, that product has been offered and consumed in largely the same way for decades. That is beginning to change.

Investors are exploring more creative ways to monetize office space. One example is the speculative construction and furnishing of space that can be leased on more flexible terms than those traditionally offered, including short-term leases of less than three years. Tenants like this approach and are often willing to pay a premium to avoid the friction inherent in the traditional leasing process. That friction includes the need to design, build, and furnish space, along with the long-term commitments required to support business needs that are constantly evolving.

Beyond changes to the physical product, landlords are also rethinking pricing. Historically, office leases have required tenants to pay a base rent plus their pro rata share of operating expenses. Those expenses are structured in different ways, such as full-service leases with a base year, NNN leases, or variations in between. Regardless of structure, the outcome is the same: fixed rent paired with variable costs, making total occupancy expense difficult to predict.

Some landlords are now simplifying this model by collapsing base rent and operating expenses into a single, fixed rent. While this rent typically increases annually by a defined measure, it is otherwise predictable and easy for tenants to understand and budget.

There is meaningful demand for office space that is occupancy-ready, offered on flexible terms, and priced with clarity. Thoughtful investors will begin carving out portions of their buildings for these offerings. While this space cannot be monetized in the same way as traditional long-term leases, over comparable time horizons it can generate significantly higher cash flow through premium rents and stronger residual value that reduces downtime between occupancies.

Flexibility Can Be Expensive

These days, many companies place a premium on flexible leasing. This is understandable in a time marked by uncertainty about the office. It is also a byproduct of a shifting market in which landlords have increased the extent to which they offer flexible solutions. Less obvious is how flexible leasing can end up costing significantly more than longer-term leasing.

A flexible lease is usually under three years in term. It is almost always a pre-built space, often furnished and ready for occupancy. Flexible leases come in the form of both direct and sublease offerings. Subleases may be discounted to market because, absent protections such as a recognition agreement from the landlord, they present a risk of losing the space if the sublandlord defaults. Yet high-quality subleases that align with the type of flexible offerings otherwise being provided by landlords are usually priced similarly. These spaces are not discounted and are often priced at a premium to the longer-term market.

In other words, flexibility costs more. This is true in part because landlords must rationalize the expense of building new space over a shorter term, and it is more difficult for them to create a positive impact on asset valuation with short-term leases. As a result, flexible leases carry a higher cost.

There is also an element of market risk, especially when the rent trajectory is rising. For example, rents in San Francisco appear to have hit bottom and are beginning to trend upward. A tenant that signs a three-year lease today is likely to face higher leasing costs in roughly two years when addressing the expiration, whereas it could have locked in a historically lower cost for a longer period. As markets improve, leasing options also diminish, making it harder for tenants to find high-quality space. Finally, as leverage begins to shift back toward landlords, concessions are often reduced quickly.

To be sure, choosing a flexible lease may be the right answer, regardless of the longer-range cost implications. But it is important for tenants to recognize that flexibility can be expensive.

Is Your Window Open?

Among the more routine elements of our business advising tenants is the art and science of analyzing opportunity, or what we call “looking for open windows.” When markets shift, as they did post-pandemic, we think it is important for occupiers to take stock of their lease, the market, and the specific dynamic in the building in which they lease space.

Somewhat surprisingly, this analysis is not intuitive to many of the companies we speak with. Instead, they believe they are stuck with their lease until the end of term. Sometimes that ends up being true.

But when it is not, when the so-called window is open, tenants can derive significant benefit from a lease restructure.

But just as windows open, they close.

The circumstances that lead to opportunity typically involve a distressed capital stack. The equity may be wiped out, a loan may be maturing, or both, all against the backdrop of steep declines in the market value of the asset. When this type of distress is present, a tenant who is willing to commit to extended term may find a landlord who is ready to reduce the existing rent immediately and contribute concessions in exchange for that term.

We have written about this strategy in the past. We even have a name for it: “End and Extend,” meaning end the high in-place rent now and extend the term.

What causes the window to close?

It is usually a reset of the capital stack, which may come in the form of an asset sale, an infusion of fresh equity, and/or a new loan. A reset takes the pressure off and allows the landlord to lease at market. After a reset, if the building has vacancy, it is common to see a flurry of new leasing activity.
When all of this is happening, the landlord loses its motivation for disrupting the tenant’s in-place cash flow.

In other words, there is no longer a reason to reduce the existing rent, irrespective of an extension offering.

While the San Francisco office market is in the early stages of a recovery, there remains plenty of distress and, potentially, open windows. We refer to the process of analyzing specific situations as part art and part science. It is art in that one must aggregate a great deal of data, historical, current, and projected, to paint the right picture and craft the negotiation strategy.

The science is in running the models that inform the opportunity.