Deepseek’s Implications for Urban Real Asset Investors and Cities
What does Deepseek mean for Urban Real Asset Investors and Local Governments? A lot.
Much has been written about Deepseek1 and the future of artificial intelligence, geopolitics and the value of the world’s largest tech companies. But what does it mean for urban real asset2 investors and local electeds?
We see several significant implications.
Deepseek signals that “inference AI”3 may be closer to reality than the market had thought. So too then may be AI assistants, autonomous driving at scale and other Jetsons-level technologies. The rub is that these types of uses need data centers (server farms/networked computing power) that are close to where the action is—cities. Yet the existing stock (so-called “edge data centers”) is low-vacancy and/or outdated.
In short, we’re starting to think that edge data centers are the ‘new office building’. The next battle line in the war for jobs and talent among major U.S. cities. We believe investors and policymakers should be taking steps in anticipation.
Some context: whether it was deep water ports, mass transit, multi-story construction, and yes, even office buildings…the most important industries and businesses have grown around mission critical real assets. These are typically at scale and entail massive–public and private– investment.
Indeed, hard as it is to believe in 20254, this century’s single largest urban reinvention investment— the Hudson Yards and #7 subway extension— was catalyzed in part by the concern that New York City would run out of office space and lose out on job creation to regional competitors (including New Jersey).5
So if artificial intelligence is, indeed, driving an economic and social reordering a city that doesn’t position itself for the revolution will be left behind.
Readers who are steeped in this world could reasonably wonder if we’ve heard this all before. Remember what we thought 2025 would be like 2015? Self-driving cars, the metaverse, everyone draped in wearables (Business Insider)?
Dare we say this time is different 😬?
The Implications: Shifts in Power, Users, and Investment Strategy
Real asset investors, including us at Group11, are getting a “middle market” entry point into AI infrastructure. So-called “hyperscale” data centers (the very large facilities that are in suburban and rural areas and tend to serve the large tech firms) have consumed the overwhelming share of capital and market attention in recent years. Yet, only “the 1%” of the investment industry can participate in these. Each facility requires—literally—billions of dollars; that leaves these as the province of the largest institutions on the planet. The rest are locked out by, among other things, fund concentration limits (not to mention investing prudence).
Yet, at the very same time, existing edge facilities are comparatively ignored in the capital markets. They are generally small, tend to require significant improvements and capex to meet current demand, and are extremely operationally intensive. At the same time, so many of the data center companies that had the teams to run these types of assets have migrated upstream to the scaled projects.
Group11 is actively working on opportunities to capitalize on this mismatch.
Power industry expertise and relationship ‘soft tissue’ with its key players will increasingly be central to economic development. The gating factor in data center development is access to electric power. Just as Cities are often not in charge of their mass transit, they do not control their electric grids. Coordinating upgrades and expansions with priority local development projects is very difficult. Near term, soft influence — building bridges— with electric utilities and their regulators may be the coin of the realm.
Electeds, developers, and thought leaders need a new tool-kit to rationalize the impacts of edge data center development. The typical rationales for supporting the impacts of new development —public goods like good permanent job creation, affordable housing, etc.—aren’t a perfect fit. Data center developments more closely resemble warehousing impacts, which have also been stymied in/around population centers.
In conclusion: Investors may be in a position to buy ‘value’ today because of inattention to edge centers—with an embedded upside as inference AI materializes and edge facilities are in demand increasingly for inference AI. Since data centers take years to realize, however, cities that want to compete as AI talent hubs and/or inference technology deployment locations should consider near term steps to broaden their aperture on how to catalyze data centers now. Just by way of example, in our combined decades of public-private RFP experience we can count on one hand where edge data centers were included in the city’s plan.
Deepseek, in its simplest terms, is a free, downloadable artificial intelligence resource for consumer and industrial use. It was produced by a Chinese company at a much lower cost—to its makers and its customers— than anything commercially released, to-date, in the U.S. As of the time of writing, it is also considered superior by some , in certain respects, to the competitive products.
We use the term ‘real assets’ to encompass the categories of real estate and infrastructure. This is increasingly consistent with how PE investment teams are organized.
This means, in essence, that computers can reason—draw their own conclusions and make their own predictions—from facts.
To the surprise of many, there are markets, including Midtown in NYC, that are seeing something of a run on A/A+ office space. Park Avenue availability, per Avison Young, has collapsed to almost 8%, levels that haven’t been seen since pre-COVID.