To serve its Prime
members, the company-built warehouses strategically, mostly outside big cities.
Now data centers — an energy and water hungry sector — are looking to do
likewise. But for their host cities, the stakes are far higher.
EXPERT OPINION BY HOWARD TULLMAN, GENERAL MANAGING PARTNER, G2T3V
AND CHICAGO HIGH TECH INVESTORS @HOWARDTULLMAN1
OCT 1, 2024
One of the lessons from the pandemic
that’s likely to stick around for quite a while has to do with what I have
called the new three Rs — reserves, redundancy, and
resilience. During the pandemic, we paid a price for our decades of reliance on
fragile and foreign supply chains driven by the gospel of just-in-time
everything, as well as an infinite consumer desire for an abundance of cheaply
made and readily disposable goods. We were left with bare shelves, empty-handed
workers with nothing to assemble, layoffs, and ultimately a shattered economy. It became
increasingly obvious that on-site and nearby storage and availability of all
manner of materials, inventory, and supplies is essential to avoid similar
problems in the future.
As with so many things, Amazon, even
before the pandemic, was already focused on a fourth and equally critical R:
rapidity. In the right-now world of the new digital economy, having to wait for
anything is a mortal sin. Everyone wants everything instantly, if not sooner,
especially when ready alternatives are just a click away.
Even more compelling was the Bezos
realization that the faster you can deliver, the more products the online buyer
is likely to buy. Instant gratification is just around the corner — no muss and
no fuss — so why not stock up on a few other things your heart desires while
you’re at it. Amazon’s success during the pandemic and thereafter in sustaining
its delivery capabilities and the speed with which it has operated since its
inception is a model for logistics firms across the world.
In further response to the teachings
of the pandemic, Amazon began a massive program to build additional, enormous,
well-stocked, and strategically located warehouses. These distribution centers
are typically located close to, but carefully outside of, major urban centers
(except in New York, where a major facility is located in Staten Island).
Proximity to huge populations of Prime customers is crucial, but dodging the
pitfalls and the extortionate politics of our major U.S. cities was equally
critical.
Collar counties and small exurban
cities are ideal targets for these mammoth facilities, which promise to bring
hundreds, if not thousands, of jobs. By and large, with some occasional whining
about the associated truck traffic, these burgs are happy to have Amazon as
their big new neighbor and employer, especially because there appeared to be no
economic or environmental downsides. Funding for social projects, schools, and
other local government initiatives also helped to smooth the path for these
facilities.
But a new and far more problematic
wave of development, modeled on Amazon, is gathering speed across the U.S.,
which communities have endorsed and supported based on their prior assumptions
and overall positive experiences with the DCs. But this new activity — the
building of self-standing and substantial data centers — is a far different and
more complicated undertaking. Data centers come with near-term and potentially
far-reaching environmental and other adverse consequences.
Everyone hears Chicken Little stories
about the evils of AI, but in terms of the general public’s awareness, these
tales are almost exclusively focused on the generative content created by the huge,
large language modules that support the processing. Very few civilians and
almost no city or state regulatory authorities have any real understanding of
the ancillary costs in terms of power requirements, water consumption, and
carbon footprint that these facilities will spawn. Common agreement — while
noting that there are varying methodologies and measurement approaches — is
that an AI inquiry takes many times more compute (power) than a comparable
Google search does. Google, in the meantime, has also built its own AI stack.
The companies proposing and building
these centers are making representations about control, remediation, and
elimination with far distant timeframes, which are speculative at best and most
likely unrealistic in any event. Meta’s DeKalb, Illinois, data center is a good
example. Its plant envelope is about 2.4 million square feet, and cost about $1
billion to construct. Meta claims that it will support several hundred jobs
when fully operational, has provided more than $1 million in direct funding to
local schools and nonprofits, and has made more than 60 grants and sponsorships
for other community organizations and programs.
At the same time, the company notes
that it “expects” to be “water positive” by 2030 (at which point the project
will restore more water than it consumes). Meta also notes that its global
operations have reached net zero emissions. Needless to say, there are no
remedies or redos for missing any of these objectives, and no regulatory tools
to enforce or assure compliance — or even progress — in the interim.
Aurora, Illinois is another city that
has quietly become the home for a number of co-located data centers for major
firms such as the CME Group and Google Cloud, which are constructing multiple
buildings to house trading systems. It wasn’t lost on anyone that Chicago’s new
mayor has been threatening for some time to impose new taxes on traders, but
that wouldn’t be possible if the trades are being conducted on computers and
systems located outside city limits. Like in Aurora. Edged Energy — a division
of Endeavour — is another national data center builder and operator that is
building a new net-zero waterless data center campus in Aurora. And finally,
Amazon itself has two fulfillment centers in Aurora with about 1,000 employees.
The bottom line is an exciting and
very mixed blessing for smaller cities in many states where no one can offer
anything more concrete than their best guesses as to the long-term
environmental concerns of the new technologies being tested and implemented.
Data center power use will reach 35GW by 2030, according to one recent report, up from 17 GW in 2022. If we thought that
crypto processing ran away with the world before anyone figured out exactly how
much heat it generated and how much power it consumed, we may be looking at
another version of the same old entrepreneur’s credo — build first and figure
things out later.