WEBVTT
00:00:00.000 --> 00:00:02.399
As always, this content is for educational and
00:00:02.399 --> 00:00:04.700
informational purposes only. It is not legal,
00:00:04.740 --> 00:00:08.800
financial or professional advice. Right. So welcome
00:00:08.800 --> 00:00:11.800
to this special deep dive on scaling legends.
00:00:12.580 --> 00:00:15.779
You know, the destination for contractors and
00:00:15.779 --> 00:00:17.519
construction business owners who are really trying
00:00:17.519 --> 00:00:20.199
to capture the next massive wave in the industry.
00:00:20.339 --> 00:00:22.820
And it is a massive wave. Oh, it's huge. Because
00:00:22.820 --> 00:00:25.660
look. If you want to build the world's most advanced
00:00:25.660 --> 00:00:28.579
AI supercomputer today, you'd think the first
00:00:28.579 --> 00:00:30.960
step is, you know, building a highly secure server
00:00:30.960 --> 00:00:33.539
room. Right. Some high -tech warehouse. Exactly.
00:00:33.700 --> 00:00:36.159
But it's not. The first thing you actually have
00:00:36.159 --> 00:00:38.820
to build is an $80 million municipal wastewater
00:00:38.820 --> 00:00:42.340
plant. Which is just, I mean, it completely upends
00:00:42.340 --> 00:00:44.460
the traditional business model for anyone bidding
00:00:44.460 --> 00:00:46.359
on these projects. Right. You aren't just putting
00:00:46.359 --> 00:00:48.520
up a tilt wall warehouse for computers anymore.
00:00:48.759 --> 00:00:51.469
You're basically... You're essentially building
00:00:51.469 --> 00:00:54.229
an industrial -grade public utility. An industrial
00:00:54.229 --> 00:00:57.030
public utility, yeah. And that reality right
00:00:57.030 --> 00:00:59.530
there, that's what's driving the single largest
00:00:59.530 --> 00:01:01.789
capital allocation shift the construction industry
00:01:01.789 --> 00:01:04.349
has seen in a generation, honestly. Without a
00:01:04.349 --> 00:01:07.150
doubt. So today we are tearing into the U .S.
00:01:07.170 --> 00:01:10.010
data center construction super cycle. We're going
00:01:10.010 --> 00:01:12.549
to look at the staggering numbers behind this
00:01:12.549 --> 00:01:15.670
boom, map out exactly where the money is flowing,
00:01:15.890 --> 00:01:18.450
and really unpack why heavy civil engineering,
00:01:18.530 --> 00:01:23.379
specifically around water management, The water
00:01:23.379 --> 00:01:52.719
wall. It is smartbusinessautomator .com. It's,
00:01:52.719 --> 00:01:54.400
you know, it's the kind of intelligence platform
00:01:54.400 --> 00:01:57.359
you absolutely need when market conditions are
00:01:57.359 --> 00:01:59.299
shifting this violently. Shifting violently.
00:01:59.480 --> 00:02:01.420
That's a good way to put it. Because the numbers
00:02:01.420 --> 00:02:03.920
we're looking at, they aren't just reflecting
00:02:03.920 --> 00:02:06.120
a busy season. I mean, they indicate a fundamental
00:02:06.120 --> 00:02:09.740
rewiring of the commercial construction sector
00:02:09.740 --> 00:02:11.919
entirely. Okay, let's jump straight into those
00:02:11.919 --> 00:02:14.479
numbers then. Pulling the latest data from smart
00:02:14.479 --> 00:02:16.580
business automators market intelligence report,
00:02:16.840 --> 00:02:19.840
year to date U .S. data center construction spending
00:02:19.840 --> 00:02:25.539
has hit $58 .1 billion. Yep. 58 .1 billion. And
00:02:25.539 --> 00:02:27.520
we need to give that number some context because
00:02:27.520 --> 00:02:29.919
that figure is more than four times the prior
00:02:29.919 --> 00:02:32.460
year record over the same period. Wow. Right.
00:02:32.520 --> 00:02:35.060
Like we aren't talking about, oh, 20 % organic
00:02:35.060 --> 00:02:38.219
growth. We are looking at a 400 % multiplier
00:02:38.219 --> 00:02:40.919
on a sector that was already operating at record
00:02:40.919 --> 00:02:43.460
highs. Yeah. And a sudden 400 % jump. That isn't
00:02:43.460 --> 00:02:45.780
growth, right? It's a market shock. Absolutely
00:02:45.780 --> 00:02:47.939
a shock. When a sector expands that rapidly,
00:02:48.120 --> 00:02:51.830
it acts like a localized black hole. It pulls
00:02:51.830 --> 00:02:54.710
massive quantities of skilled trades away from
00:02:54.710 --> 00:02:56.909
other commercial projects. Oh, for sure. Your
00:02:56.909 --> 00:02:59.689
electricians, your pipe fitters, heavy equipment
00:02:59.689 --> 00:03:02.550
operators. Yeah, exactly. It just monopolizes
00:03:02.550 --> 00:03:06.150
regional supplies of structural steel, specialized
00:03:06.150 --> 00:03:10.009
concrete, copper. And look at the velocity. Yeah.
00:03:10.129 --> 00:03:13.889
One recent month alone accounted for $7 .9 billion
00:03:13.889 --> 00:03:17.870
in spending. Wait, $7 .9 billion in a single
00:03:17.870 --> 00:03:20.979
30 -day window? One month. That makes it one
00:03:20.979 --> 00:03:23.319
of the highest single months on record in the
00:03:23.319 --> 00:03:25.360
history of the sector. It really does. It's unprecedented.
00:03:25.599 --> 00:03:28.000
But wait, hold on. The same SBA report shows
00:03:28.000 --> 00:03:31.560
there are nearly 100 megaprojects currently in
00:03:31.560 --> 00:03:33.580
pre -construction with year -end start dates.
00:03:33.759 --> 00:03:36.919
And the planned spend on those is over $101 billion.
00:03:37.539 --> 00:03:40.139
$101 billion. Just waiting in the wings. Right.
00:03:40.180 --> 00:03:42.199
And I look at that number and I immediately think
00:03:42.199 --> 00:03:44.699
about bonding. I think about risk. Oh, definitely.
00:03:44.939 --> 00:03:46.719
Because historically, the construction industry
00:03:46.719 --> 00:03:49.560
operates on incredibly tight bandwidths for bonds.
00:03:49.580 --> 00:03:58.180
Right. You can't just infinitely scale. They'd
00:03:58.180 --> 00:04:06.740
get laughed off the phone. Exactly. Are general
00:04:06.740 --> 00:04:10.340
contractors even scaled to absorb $101 billion
00:04:10.340 --> 00:04:13.139
in a few quarters like this? Or is this just
00:04:13.139 --> 00:04:15.240
a fire hose hitting a teacup? Well, the short
00:04:15.240 --> 00:04:17.519
answer is no, not in their traditional formats.
00:04:17.740 --> 00:04:19.720
The surety market, you know, the companies that
00:04:19.720 --> 00:04:21.639
actually guarantee these massive projects will
00:04:21.639 --> 00:04:23.939
get built. They're not going to underwrite that
00:04:23.939 --> 00:04:26.699
kind of localized risk for a single entity. The
00:04:26.699 --> 00:04:29.060
exposure is just way too high. Way too high.
00:04:29.160 --> 00:04:31.720
So what happens? You see this massive rise in
00:04:31.720 --> 00:04:35.759
joint ventures. Top tier national general contractors
00:04:35.759 --> 00:04:38.899
are partner up just to syndicate the risk. And
00:04:38.899 --> 00:04:42.079
they're also being forced to leverage. really
00:04:42.079 --> 00:04:44.959
massive supply chain financing because the procurement
00:04:44.959 --> 00:04:47.399
cycles are so extreme. Well, sure. You can't
00:04:47.399 --> 00:04:49.759
wait on standard 90 day payment terms when you
00:04:49.759 --> 00:04:52.660
are bulk ordering $50 million worth of specialized
00:04:52.660 --> 00:04:54.860
cooling towers, right? Right. Especially when
00:04:54.860 --> 00:04:57.060
those towers have an 18 month lead time. Wow.
00:04:57.870 --> 00:05:00.089
That drastically changes the unit economics for
00:05:00.089 --> 00:05:02.310
anyone involved. Completely. And you mentioned
00:05:02.310 --> 00:05:04.470
procurement, which leads directly to what it
00:05:04.470 --> 00:05:06.370
actually costs to put a shovel in the ground
00:05:06.370 --> 00:05:08.189
for these things. Yeah, let's talk about the
00:05:08.189 --> 00:05:10.629
cost per square foot. So the median build cost
00:05:10.629 --> 00:05:14.850
is currently sitting at $473 per square foot.
00:05:14.990 --> 00:05:17.970
It's staggering. It is. I mean, we all know standard
00:05:17.970 --> 00:05:20.290
tilt -up industrial warehousing might run you
00:05:20.290 --> 00:05:23.509
somewhere between $100 to maybe $200 a foot,
00:05:23.569 --> 00:05:25.230
depending on the market, the finishes, whatever.
00:05:25.410 --> 00:05:29.850
But $473 a foot, that is approaching like hospital
00:05:29.850 --> 00:05:33.009
or advanced laboratory levels of density. At
00:05:33.009 --> 00:05:35.870
$473 a foot, we aren't just putting up tilt -wall
00:05:35.870 --> 00:05:39.209
warehouses. These are highly engineered fortresses.
00:05:39.250 --> 00:05:41.629
It's like comparing the construction of a basic
00:05:41.629 --> 00:05:46.939
pole barn to a nuclear submarine. is highly accurate,
00:05:47.100 --> 00:05:50.100
actually, because of what that $473 represents.
00:05:50.540 --> 00:05:52.819
Right. You're building structural shells engineered
00:05:52.819 --> 00:05:55.819
to withstand extreme seismic and weather events.
00:05:56.180 --> 00:05:58.920
Because, look, uptime is non -negotiable for
00:05:58.920 --> 00:06:00.480
these hyperscalers. Right. They can't afford
00:06:00.480 --> 00:06:03.120
to go offline for a hurricane. Exactly. But the
00:06:03.120 --> 00:06:06.120
real cost driver is the MEP. The mechanical,
00:06:06.240 --> 00:06:08.220
electrical, and plumbing traits. Oh, the MEP
00:06:08.220 --> 00:06:11.259
must be insane. It is. In a standard commercial
00:06:11.259 --> 00:06:15.620
build, MEP might be what? 20 to 30 % of the budget.
00:06:15.680 --> 00:06:17.860
Yeah, around there. In a hyperscale data center,
00:06:18.000 --> 00:06:22.459
it can easily eclipse 50 or 60%. Jeez. You're
00:06:22.459 --> 00:06:25.439
installing massive, redundant power substations.
00:06:27.669 --> 00:06:30.149
uninterruptible power supply rooms that are just
00:06:30.149 --> 00:06:32.629
filled to the brim with lithium -ion batteries.
00:06:32.970 --> 00:06:35.170
And layers of concurrent maintainability, I'm
00:06:35.170 --> 00:06:37.910
assuming. Yes, exactly. Meaning they have to
00:06:37.910 --> 00:06:39.709
be able to shut down and perform maintenance
00:06:39.709 --> 00:06:42.430
on any single piece of infrastructure without
00:06:42.430 --> 00:06:45.310
ever dropping power to the server floor. Which
00:06:45.310 --> 00:06:47.310
means the geographic footprint of where this
00:06:47.310 --> 00:06:49.870
money is landing isn't random. Not at all. Like
00:06:49.870 --> 00:06:51.529
we said, you can't just drop a submarine in the
00:06:51.529 --> 00:06:53.490
middle of a desert without a supply line. Nope.
00:06:53.569 --> 00:06:55.649
According to the Smart Business Automator Market
00:06:55.649 --> 00:06:58.759
Intelligence Report, Roughly 60 % of all new
00:06:58.759 --> 00:07:01.199
starts are captured by just five states. Just
00:07:01.199 --> 00:07:04.139
five. Virginia, Texas, Louisiana, Illinois, and
00:07:04.139 --> 00:07:07.279
North Carolina. Now, I'd assume this hyper -concentration
00:07:07.279 --> 00:07:11.060
is just about cheap land and grid access in places
00:07:11.060 --> 00:07:14.699
like Texas or Louisiana, but there has to be
00:07:14.699 --> 00:07:17.100
a fiber -optic component driving Virginia and
00:07:17.100 --> 00:07:20.370
Illinois, right? Spot on. It's the intersection
00:07:20.370 --> 00:07:22.610
of path of least resistance energy grids and
00:07:22.610 --> 00:07:25.009
legacy fiber networks. Okay, that makes sense.
00:07:25.189 --> 00:07:27.910
So Northern Virginia, for example, is the undisputed
00:07:27.910 --> 00:07:30.949
data center alley of the world. And that's because...
00:07:31.240 --> 00:07:33.639
Decades ago, the foundational fiber backbone
00:07:33.639 --> 00:07:36.279
of the internet was laid there. Oh, right. And
00:07:36.279 --> 00:07:38.220
then Illinois serves as the critical switching
00:07:38.220 --> 00:07:41.240
hub for the entire Midwest network. Got it. Meanwhile,
00:07:41.379 --> 00:07:43.620
you've got Texas and Louisiana offering vast
00:07:43.620 --> 00:07:46.759
tracts of developable land and deeply entrenched,
00:07:46.819 --> 00:07:49.620
heavily deregulated energy infrastructure. So
00:07:49.620 --> 00:07:52.160
you combine land, power, and fiber, and boom,
00:07:52.220 --> 00:07:55.050
you get that 60 % concentration. Exactly. But
00:07:55.050 --> 00:07:57.990
the scale of the individual projects within those
00:07:57.990 --> 00:08:00.750
zones and the immediate spillover markets is
00:08:00.750 --> 00:08:02.649
just staggering. Right. Let's look at a named
00:08:02.649 --> 00:08:04.629
example from the intelligence report. Skansky
00:08:04.629 --> 00:08:08.449
USA recently booked a $255 million contract for
00:08:08.449 --> 00:08:11.310
a Georgia data center. But the kicker here, the
00:08:11.310 --> 00:08:14.089
really wild part, is the classification of that
00:08:14.089 --> 00:08:17.750
contract. It is a supplemental contract. Yeah.
00:08:17.810 --> 00:08:20.410
Let that sink in for a second. A $255 million
00:08:20.410 --> 00:08:24.240
change order, essentially. Right. I mean... For
00:08:24.240 --> 00:08:26.920
a regional general contractor, a quarter of a
00:08:26.920 --> 00:08:29.600
billion dollars is a massive company -defining
00:08:29.600 --> 00:08:32.399
anchor project. You retire on that. Yeah. But
00:08:32.399 --> 00:08:35.000
in the hyperscale data center world, it is simply
00:08:35.000 --> 00:08:38.059
an add -on to an existing scope of work. The
00:08:38.059 --> 00:08:40.779
project specifics on that Skanska build highlight
00:08:40.779 --> 00:08:44.100
exactly why these costs inflate so much. What
00:08:44.100 --> 00:08:47.320
are the specs? So it's 245 ,000 square feet comprising
00:08:47.320 --> 00:08:49.899
five data halls, and it's targeted for completion
00:08:49.899 --> 00:08:52.940
by the first quarter of 2028. Wow. We are looking
00:08:52.940 --> 00:08:55.580
at multi -year phasing here. And that extended
00:08:55.580 --> 00:08:57.419
timeline, that's driven by the commissioning
00:08:57.419 --> 00:08:59.519
process. Oh, sure. Because you don't just build
00:08:59.519 --> 00:09:02.740
245 ,000 square foot shell, flip a switch and
00:09:02.740 --> 00:09:05.519
hand over the keys. I wish. Yeah, right. No,
00:09:05.580 --> 00:09:08.080
you build Hall 1. You rigorously test the cooling
00:09:08.080 --> 00:09:09.980
and power loads under artificial stress. They
00:09:09.980 --> 00:09:12.240
often call this Level 5 commissioning. And you
00:09:12.240 --> 00:09:14.480
hand it over while you're literally still pouring
00:09:14.480 --> 00:09:17.379
concrete for Hall 3. Right. So by stretching
00:09:17.379 --> 00:09:20.240
these timelines out to 2028, contractors are
00:09:20.240 --> 00:09:22.759
locking up their best superintendents, their
00:09:22.759 --> 00:09:27.059
top project managers, local labor pools for half
00:09:27.059 --> 00:09:30.279
a decade. Yes. you are effectively building a
00:09:30.279 --> 00:09:32.799
temporary mini -city of construction logistics
00:09:32.799 --> 00:09:36.600
that just hovers over a single geographic pocket
00:09:36.600 --> 00:09:40.340
for years. Which creates an absolute vacuum for
00:09:40.340 --> 00:09:42.840
local subcontractors. A total vacuum. If you're
00:09:42.840 --> 00:09:45.100
an electrical or mechanical sub in Georgia or
00:09:45.100 --> 00:09:47.320
one of those top five states, you are suddenly
00:09:47.320 --> 00:09:50.179
competing against these multi -year behemoth
00:09:50.179 --> 00:09:52.980
projects that can afford to pay heavy premiums
00:09:52.980 --> 00:09:54.899
for labor. They absolutely will pay the premium.
00:09:55.279 --> 00:09:57.879
Because the end client, usually some massive
00:09:57.879 --> 00:10:00.720
tech conglomerate, is losing millions in potential
00:10:00.720 --> 00:10:03.379
AI revenue for every day that facility isn't
00:10:03.379 --> 00:10:05.659
online. Right. They will pay to accelerate the
00:10:05.659 --> 00:10:07.879
schedule. Exactly. So if you are a construction
00:10:07.879 --> 00:10:09.840
business owner operating in one of these target
00:10:09.840 --> 00:10:12.240
states, your strategic planning has to account
00:10:12.240 --> 00:10:14.840
for labor hoarding. Labor hoarding. You either
00:10:14.840 --> 00:10:17.240
figure out how to bid into these hyperscale supply
00:10:17.240 --> 00:10:20.100
chains yourself, or you have to completely adjust
00:10:20.100 --> 00:10:22.220
your pricing models for local commercial work.
00:10:22.509 --> 00:10:24.990
Because your labor costs are going to be artificially
00:10:24.990 --> 00:10:27.370
inflated by the data center campus down the road.
00:10:27.529 --> 00:10:29.070
Yep. They'll pull all your guys if you don't
00:10:29.070 --> 00:10:31.730
adjust. Wow. Okay, that brings us to the most
00:10:31.730 --> 00:10:34.450
critical pivot of this entire deep dive. Here
00:10:34.450 --> 00:10:37.409
we go. You have the joint venture formed. You
00:10:37.409 --> 00:10:40.049
have the bonding secured. You found the land
00:10:40.049 --> 00:10:43.539
in Virginia or Georgia. And the $101 billion
00:10:43.539 --> 00:10:46.559
in pre -construction capital is ready to deploy.
00:10:46.860 --> 00:10:48.639
You're ready to dig. You're ready. And then you
00:10:48.639 --> 00:10:51.000
hit a wall. Yeah. And it's not the local power
00:10:51.000 --> 00:10:53.200
utility telling you they don't have the megawatts.
00:10:53.200 --> 00:10:55.899
It's the water board. The water wall. The water
00:10:55.899 --> 00:10:58.279
wall. This is where the physics of artificial
00:10:58.279 --> 00:11:01.519
intelligence completely alters the construction
00:11:01.519 --> 00:11:03.940
blueprint. To illustrate how massive this wall
00:11:03.940 --> 00:11:07.360
is, let's look at Memphis. XAI has committed
00:11:07.360 --> 00:11:10.940
to resuming an $80 million graywater recycling
00:11:10.940 --> 00:11:13.860
plant for its Colossus campus in Memphis. $80
00:11:13.860 --> 00:11:16.399
million. $80 million just for the water facility.
00:11:16.759 --> 00:11:19.000
The capacity specs on that facility tell the
00:11:19.000 --> 00:11:21.080
real story, though. Oh, yeah. It is designed
00:11:21.080 --> 00:11:24.179
to process up to 13 million gallons of water
00:11:24.179 --> 00:11:27.440
a day. 13 million gallons a day isn't a data
00:11:27.440 --> 00:11:29.960
center. That is municipal scale. It really is.
00:11:30.059 --> 00:11:32.240
That is enough water to supply a small city.
00:11:32.879 --> 00:11:35.840
Why on earth does a supercomputer need the water
00:11:35.840 --> 00:11:38.620
pressure of a municipality? Well, it comes down
00:11:38.620 --> 00:11:41.299
to thermal dynamics, right? Yeah. And the shift
00:11:41.299 --> 00:11:44.559
from traditional cloud computing to artificial
00:11:44.559 --> 00:11:47.139
intelligence. Okay, break that down for us. So,
00:11:47.159 --> 00:11:50.159
traditional server racks can generally be cooled
00:11:50.159 --> 00:11:53.149
with ambient air. You pump chilled air up through
00:11:53.149 --> 00:11:55.710
a raised floor, it passes over the servers, and
00:11:55.710 --> 00:11:57.830
you vent the hot air out. Standard AC stuff.
00:11:58.210 --> 00:12:01.929
Right. But the new GPU clusters used for AI training,
00:12:02.090 --> 00:12:05.639
they run so incredibly hot. The air cooling is
00:12:05.639 --> 00:12:08.360
physically incapable of rejecting the heat fast
00:12:08.360 --> 00:12:11.039
enough. Oh, wow. So they have to use liquid cooling.
00:12:11.240 --> 00:12:12.919
Exactly. They have to use direct -to -chip liquid
00:12:12.919 --> 00:12:15.480
cooling and massive rear -door heat exchangers.
00:12:15.600 --> 00:12:17.159
Wait, so they pump it directly to the chips?
00:12:17.320 --> 00:12:19.379
Yep. You are pumping fluids directly through
00:12:19.379 --> 00:12:21.740
the server chassis itself, and then that heated
00:12:21.740 --> 00:12:24.340
fluid is piped out to massive evaporative cooling
00:12:24.340 --> 00:12:26.700
towers on the roof or the perimeter of the facility.
00:12:26.960 --> 00:12:29.779
And to keep those processors from melting down,
00:12:30.350 --> 00:12:33.129
Those towers evaporate millions of gallons of
00:12:33.129 --> 00:12:35.450
water into the atmosphere every single day. Which
00:12:35.450 --> 00:12:38.129
leads directly to the core conflict in Memphis.
00:12:38.860 --> 00:12:41.379
The primary reason stated for this $80 million
00:12:41.379 --> 00:12:44.559
water plant is to ensure the supercomputer's
00:12:44.559 --> 00:12:47.600
immense cooling needs do not draw down the city
00:12:47.600 --> 00:12:50.620
of Memphis' drinking water aquifer. Right. Because
00:12:50.620 --> 00:12:53.059
Memphis relies on the Memphis sand aquifer. Yeah.
00:12:53.159 --> 00:12:55.639
It is a pristine source of municipal drinking
00:12:55.639 --> 00:12:58.440
water. Right. You cannot have a private corporation
00:12:58.440 --> 00:13:02.220
drawing 13 million gallons of pristine aquifer
00:13:02.220 --> 00:13:05.460
water a day just to cool... Graphic processing
00:13:05.460 --> 00:13:07.919
units. People would riot. The municipal grid
00:13:07.919 --> 00:13:10.500
would fail, and yet the public backlash would
00:13:10.500 --> 00:13:13.580
be catastrophic. Yeah. So the only viable engineering
00:13:13.580 --> 00:13:15.919
solution is gray water. Gray water being treated
00:13:15.919 --> 00:13:18.440
wastewater. Exactly. Water that isn't safe for
00:13:18.440 --> 00:13:20.419
human consumption, but is perfectly acceptable
00:13:20.419 --> 00:13:22.539
to run through an industrial heat exchanger.
00:13:22.639 --> 00:13:25.360
Exactly right. But getting 13 million gallons
00:13:25.360 --> 00:13:28.179
of gray water to a data center every day, that
00:13:28.179 --> 00:13:30.259
requires heavy infrastructure. Yeah, you can't
00:13:30.259 --> 00:13:34.149
just run a hose. No. You have to intercept municipal
00:13:34.149 --> 00:13:37.289
wastewater. You have to build a dedicated treatment
00:13:37.289 --> 00:13:41.309
facility on or near the campus, filter it, chemically
00:13:41.309 --> 00:13:44.389
treat it to prevent scale and biological growth
00:13:44.389 --> 00:13:46.730
in the cooling loops. Oh, because algae and stuff?
00:13:46.909 --> 00:13:48.409
Right. You can't have algae in a supercomputer.
00:13:48.649 --> 00:13:51.309
You use it, and then you often have to treat
00:13:51.309 --> 00:13:53.450
the blowdown water before discharging it back
00:13:53.450 --> 00:13:55.470
into the municipal sewer system. See, that is
00:13:55.470 --> 00:13:58.809
the ultimate aha moment here. The future of tech
00:13:58.809 --> 00:14:01.149
construction isn't digital, it's heavy civil.
00:14:01.370 --> 00:14:04.230
Yes. If an $80 million water treatment plant
00:14:04.230 --> 00:14:06.669
is now just a prerequisite for building an AI
00:14:06.669 --> 00:14:10.049
campus, the very definition of a data center
00:14:10.049 --> 00:14:12.549
contractor has fundamentally changed. It really
00:14:12.549 --> 00:14:15.009
has. I mean, referencing Smart Business Automator
00:14:15.009 --> 00:14:17.070
again here, you look at the trends. If a build
00:14:17.070 --> 00:14:19.210
can drain a city's drinking aquifer, you are
00:14:19.210 --> 00:14:21.529
no longer just bidding on clean rooms, steel
00:14:21.529 --> 00:14:23.950
framing, and heavy copper wire. Right. You're
00:14:23.950 --> 00:14:26.049
bidding on miles of large diameter underground
00:14:26.049 --> 00:14:28.710
trenching. Yep. You are pouring concrete for
00:14:28.710 --> 00:14:30.850
massive clarifiers. You're building chemical
00:14:30.850 --> 00:14:33.769
dosing stations, complex pump houses, and biological
00:14:33.769 --> 00:14:36.450
membrane filtration systems. You are building
00:14:36.450 --> 00:14:39.950
an industrial wastewater treatment plant. right
00:14:39.950 --> 00:14:42.169
next to the most sophisticated computing hardware
00:14:42.169 --> 00:14:45.409
on Earth. And the tolerances are zero. Zero.
00:14:45.570 --> 00:14:47.929
If the water plant goes offline, the cooling
00:14:47.929 --> 00:14:50.549
loops run dry. If the cooling loops run dry,
00:14:50.750 --> 00:14:53.210
the supercomputer fries in a matter of minutes.
00:14:53.490 --> 00:14:55.750
It's all connected. So the contractors who are
00:14:55.750 --> 00:14:58.070
actually going to capture a piece of this $101
00:14:58.070 --> 00:15:01.409
billion pre -construction pipeline aren't just
00:15:01.409 --> 00:15:03.649
the ones who know how to build a highly secure
00:15:03.649 --> 00:15:07.090
warehouse. They are the ones with deep, heavy,
00:15:07.129 --> 00:15:10.899
civil... earthwork, and utility expertise. Or,
00:15:11.039 --> 00:15:13.860
you know, the ones strategic enough to form the
00:15:13.860 --> 00:15:16.580
right partnerships. Very true. If you're a premier
00:15:16.580 --> 00:15:19.679
commercial builder. You need to be actively joint
00:15:19.679 --> 00:15:22.419
venturing with the heavy civil contractors who
00:15:22.419 --> 00:15:24.559
traditionally build municipal infrastructure.
00:15:24.679 --> 00:15:26.700
Absolutely. Because this water wall is the new
00:15:26.700 --> 00:15:29.059
reality. You can secure all the silicon. You
00:15:29.059 --> 00:15:31.019
can lock down all the electricity. But if you
00:15:31.019 --> 00:15:33.419
cannot engineer and execute the water rights
00:15:33.419 --> 00:15:36.279
and the physical water infrastructure, your mega
00:15:36.279 --> 00:15:38.179
project will never make it out of pre -construction.
00:15:38.299 --> 00:15:40.639
It's dead in the water. Literally. Literally.
00:15:40.740 --> 00:15:44.019
Okay, let's synthesize this actionable intelligence
00:15:44.019 --> 00:15:46.179
for the construction business owners listening
00:15:46.179 --> 00:15:48.360
right now. Let's do it. We started by looking
00:15:48.360 --> 00:15:52.960
at a $58 .1 billion tidal wave of year -to -date
00:15:52.960 --> 00:15:55.860
spending, which is moving at four times the speed
00:15:55.860 --> 00:15:58.940
of previous records and demanding totally new
00:15:58.940 --> 00:16:01.480
risk management and bonding strategies. Right.
00:16:01.539 --> 00:16:04.720
And we followed that capital into a $101 billion
00:16:04.720 --> 00:16:07.850
pre -construction pipeline. We know from the
00:16:07.850 --> 00:16:09.750
Smart Business Automator Market Intelligence
00:16:09.750 --> 00:16:13.350
Report that 60 % of this work is hyper -concentrated
00:16:13.350 --> 00:16:16.309
in five key states. Virginia, Texas, Louisiana,
00:16:16.629 --> 00:16:19.129
Illinois, and North Carolina. Exactly. And we
00:16:19.129 --> 00:16:22.309
established that at a median build cost of $473
00:16:22.309 --> 00:16:25.470
a square foot. These projects are incredibly
00:16:25.470 --> 00:16:28.389
dense, multi -year commitments. Yeah. As we saw
00:16:28.389 --> 00:16:31.590
with Skanska's $255 million supplemental contract
00:16:31.590 --> 00:16:34.549
in Georgia, a single phase of a single campus
00:16:34.549 --> 00:16:37.230
can tie up labor and project management resources
00:16:37.230 --> 00:16:40.250
until 2028. Such a massive footprint. It really
00:16:40.250 --> 00:16:42.090
is. And ultimately, we uncovered that the barrier
00:16:42.090 --> 00:16:43.889
to entry has evolved from electrical capacity
00:16:43.889 --> 00:16:46.570
to municipal water management. The big pivot.
00:16:46.690 --> 00:16:49.750
To compete at this scale, contractors must possess
00:16:49.750 --> 00:16:53.450
or partner for heavy utility and water treatment
00:16:53.450 --> 00:16:56.649
expertise to handle the massive evaporative cooling
00:16:56.649 --> 00:16:59.980
requirements of AI infrastructure. It's a wildly
00:16:59.980 --> 00:17:02.879
lucrative space, but only for the firms agile
00:17:02.879 --> 00:17:05.039
enough to realize that building a data center
00:17:05.039 --> 00:17:07.839
today means building a water plant tomorrow.
00:17:08.200 --> 00:17:11.299
The traditional commercial blueprint just doesn't
00:17:11.299 --> 00:17:13.839
apply here anymore. The physics of the build
00:17:13.839 --> 00:17:15.859
have been completely rewritten. They really have.
00:17:16.019 --> 00:17:18.119
So I want to leave you with a final thought to
00:17:18.119 --> 00:17:20.720
mull over. Something that wasn't explicitly spilled
00:17:20.720 --> 00:17:22.819
out in the data, but is the absolute logical
00:17:22.819 --> 00:17:25.380
conclusion of this water constraint. We know
00:17:25.380 --> 00:17:27.579
municipalities like Memphis are forcing developers
00:17:27.579 --> 00:17:30.009
to build their own greywater infrastructure to
00:17:30.009 --> 00:17:32.819
protect drinking aquifers. But what happens to
00:17:32.819 --> 00:17:36.460
that massive $101 billion pre -construction pipeline
00:17:36.460 --> 00:17:39.380
if the municipalities in those top five states
00:17:39.380 --> 00:17:42.019
suddenly decide that even gray water is too precious
00:17:42.019 --> 00:17:44.460
and they cap commercial water usage entirely?
00:17:44.900 --> 00:17:47.839
Right. Will the next great data center boom force
00:17:47.839 --> 00:17:49.920
the construction industry to pioneer entirely
00:17:49.920 --> 00:17:53.099
new geographies near oceans or major river systems?
00:17:53.279 --> 00:17:55.920
Or will it force us to invent entirely new ways
00:17:55.920 --> 00:17:58.000
of building massive infrastructure without a
00:17:58.000 --> 00:17:59.039
drop of water at all?
00:00:00.000 --> 00:00:02.399
As always, this content is for educational and
00:00:02.399 --> 00:00:04.700
informational purposes only. It is not legal,
00:00:04.740 --> 00:00:08.800
financial or professional advice. Right. So welcome
00:00:08.800 --> 00:00:11.800
to this special deep dive on scaling legends.
00:00:12.580 --> 00:00:15.779
You know, the destination for contractors and
00:00:15.779 --> 00:00:17.519
construction business owners who are really trying
00:00:17.519 --> 00:00:20.199
to capture the next massive wave in the industry.
00:00:20.339 --> 00:00:22.820
And it is a massive wave. Oh, it's huge. Because
00:00:22.820 --> 00:00:25.660
look. If you want to build the world's most advanced
00:00:25.660 --> 00:00:28.579
AI supercomputer today, you'd think the first
00:00:28.579 --> 00:00:30.960
step is, you know, building a highly secure server
00:00:30.960 --> 00:00:33.539
room. Right. Some high -tech warehouse. Exactly.
00:00:33.700 --> 00:00:36.159
But it's not. The first thing you actually have
00:00:36.159 --> 00:00:38.820
to build is an $80 million municipal wastewater
00:00:38.820 --> 00:00:42.340
plant. Which is just, I mean, it completely upends
00:00:42.340 --> 00:00:44.460
the traditional business model for anyone bidding
00:00:44.460 --> 00:00:46.359
on these projects. Right. You aren't just putting
00:00:46.359 --> 00:00:48.520
up a tilt wall warehouse for computers anymore.
00:00:48.759 --> 00:00:51.469
You're basically... You're essentially building
00:00:51.469 --> 00:00:54.229
an industrial -grade public utility. An industrial
00:00:54.229 --> 00:00:57.030
public utility, yeah. And that reality right
00:00:57.030 --> 00:00:59.530
there, that's what's driving the single largest
00:00:59.530 --> 00:01:01.789
capital allocation shift the construction industry
00:01:01.789 --> 00:01:04.349
has seen in a generation, honestly. Without a
00:01:04.349 --> 00:01:07.150
doubt. So today we are tearing into the U .S.
00:01:07.170 --> 00:01:10.010
data center construction super cycle. We're going
00:01:10.010 --> 00:01:12.549
to look at the staggering numbers behind this
00:01:12.549 --> 00:01:15.670
boom, map out exactly where the money is flowing,
00:01:15.890 --> 00:01:18.450
and really unpack why heavy civil engineering,
00:01:18.530 --> 00:01:23.379
specifically around water management, The water
00:01:23.379 --> 00:01:52.719
wall. It is smartbusinessautomator .com. It's,
00:01:52.719 --> 00:01:54.400
you know, it's the kind of intelligence platform
00:01:54.400 --> 00:01:57.359
you absolutely need when market conditions are
00:01:57.359 --> 00:01:59.299
shifting this violently. Shifting violently.
00:01:59.480 --> 00:02:01.420
That's a good way to put it. Because the numbers
00:02:01.420 --> 00:02:03.920
we're looking at, they aren't just reflecting
00:02:03.920 --> 00:02:06.120
a busy season. I mean, they indicate a fundamental
00:02:06.120 --> 00:02:09.740
rewiring of the commercial construction sector
00:02:09.740 --> 00:02:11.919
entirely. Okay, let's jump straight into those
00:02:11.919 --> 00:02:14.479
numbers then. Pulling the latest data from smart
00:02:14.479 --> 00:02:16.580
business automators market intelligence report,
00:02:16.840 --> 00:02:19.840
year to date U .S. data center construction spending
00:02:19.840 --> 00:02:25.539
has hit $58 .1 billion. Yep. 58 .1 billion. And
00:02:25.539 --> 00:02:27.520
we need to give that number some context because
00:02:27.520 --> 00:02:29.919
that figure is more than four times the prior
00:02:29.919 --> 00:02:32.460
year record over the same period. Wow. Right.
00:02:32.520 --> 00:02:35.060
Like we aren't talking about, oh, 20 % organic
00:02:35.060 --> 00:02:38.219
growth. We are looking at a 400 % multiplier
00:02:38.219 --> 00:02:40.919
on a sector that was already operating at record
00:02:40.919 --> 00:02:43.460
highs. Yeah. And a sudden 400 % jump. That isn't
00:02:43.460 --> 00:02:45.780
growth, right? It's a market shock. Absolutely
00:02:45.780 --> 00:02:47.939
a shock. When a sector expands that rapidly,
00:02:48.120 --> 00:02:51.830
it acts like a localized black hole. It pulls
00:02:51.830 --> 00:02:54.710
massive quantities of skilled trades away from
00:02:54.710 --> 00:02:56.909
other commercial projects. Oh, for sure. Your
00:02:56.909 --> 00:02:59.689
electricians, your pipe fitters, heavy equipment
00:02:59.689 --> 00:03:02.550
operators. Yeah, exactly. It just monopolizes
00:03:02.550 --> 00:03:06.150
regional supplies of structural steel, specialized
00:03:06.150 --> 00:03:10.009
concrete, copper. And look at the velocity. Yeah.
00:03:10.129 --> 00:03:13.889
One recent month alone accounted for $7 .9 billion
00:03:13.889 --> 00:03:17.870
in spending. Wait, $7 .9 billion in a single
00:03:17.870 --> 00:03:20.979
30 -day window? One month. That makes it one
00:03:20.979 --> 00:03:23.319
of the highest single months on record in the
00:03:23.319 --> 00:03:25.360
history of the sector. It really does. It's unprecedented.
00:03:25.599 --> 00:03:28.000
But wait, hold on. The same SBA report shows
00:03:28.000 --> 00:03:31.560
there are nearly 100 megaprojects currently in
00:03:31.560 --> 00:03:33.580
pre -construction with year -end start dates.
00:03:33.759 --> 00:03:36.919
And the planned spend on those is over $101 billion.
00:03:37.539 --> 00:03:40.139
$101 billion. Just waiting in the wings. Right.
00:03:40.180 --> 00:03:42.199
And I look at that number and I immediately think
00:03:42.199 --> 00:03:44.699
about bonding. I think about risk. Oh, definitely.
00:03:44.939 --> 00:03:46.719
Because historically, the construction industry
00:03:46.719 --> 00:03:49.560
operates on incredibly tight bandwidths for bonds.
00:03:49.580 --> 00:03:58.180
Right. You can't just infinitely scale. They'd
00:03:58.180 --> 00:04:06.740
get laughed off the phone. Exactly. Are general
00:04:06.740 --> 00:04:10.340
contractors even scaled to absorb $101 billion
00:04:10.340 --> 00:04:13.139
in a few quarters like this? Or is this just
00:04:13.139 --> 00:04:15.240
a fire hose hitting a teacup? Well, the short
00:04:15.240 --> 00:04:17.519
answer is no, not in their traditional formats.
00:04:17.740 --> 00:04:19.720
The surety market, you know, the companies that
00:04:19.720 --> 00:04:21.639
actually guarantee these massive projects will
00:04:21.639 --> 00:04:23.939
get built. They're not going to underwrite that
00:04:23.939 --> 00:04:26.699
kind of localized risk for a single entity. The
00:04:26.699 --> 00:04:29.060
exposure is just way too high. Way too high.
00:04:29.160 --> 00:04:31.720
So what happens? You see this massive rise in
00:04:31.720 --> 00:04:35.759
joint ventures. Top tier national general contractors
00:04:35.759 --> 00:04:38.899
are partner up just to syndicate the risk. And
00:04:38.899 --> 00:04:42.079
they're also being forced to leverage. really
00:04:42.079 --> 00:04:44.959
massive supply chain financing because the procurement
00:04:44.959 --> 00:04:47.399
cycles are so extreme. Well, sure. You can't
00:04:47.399 --> 00:04:49.759
wait on standard 90 day payment terms when you
00:04:49.759 --> 00:04:52.660
are bulk ordering $50 million worth of specialized
00:04:52.660 --> 00:04:54.860
cooling towers, right? Right. Especially when
00:04:54.860 --> 00:04:57.060
those towers have an 18 month lead time. Wow.
00:04:57.870 --> 00:05:00.089
That drastically changes the unit economics for
00:05:00.089 --> 00:05:02.310
anyone involved. Completely. And you mentioned
00:05:02.310 --> 00:05:04.470
procurement, which leads directly to what it
00:05:04.470 --> 00:05:06.370
actually costs to put a shovel in the ground
00:05:06.370 --> 00:05:08.189
for these things. Yeah, let's talk about the
00:05:08.189 --> 00:05:10.629
cost per square foot. So the median build cost
00:05:10.629 --> 00:05:14.850
is currently sitting at $473 per square foot.
00:05:14.990 --> 00:05:17.970
It's staggering. It is. I mean, we all know standard
00:05:17.970 --> 00:05:20.290
tilt -up industrial warehousing might run you
00:05:20.290 --> 00:05:23.509
somewhere between $100 to maybe $200 a foot,
00:05:23.569 --> 00:05:25.230
depending on the market, the finishes, whatever.
00:05:25.410 --> 00:05:29.850
But $473 a foot, that is approaching like hospital
00:05:29.850 --> 00:05:33.009
or advanced laboratory levels of density. At
00:05:33.009 --> 00:05:35.870
$473 a foot, we aren't just putting up tilt -wall
00:05:35.870 --> 00:05:39.209
warehouses. These are highly engineered fortresses.
00:05:39.250 --> 00:05:41.629
It's like comparing the construction of a basic
00:05:41.629 --> 00:05:46.939
pole barn to a nuclear submarine. is highly accurate,
00:05:47.100 --> 00:05:50.100
actually, because of what that $473 represents.
00:05:50.540 --> 00:05:52.819
Right. You're building structural shells engineered
00:05:52.819 --> 00:05:55.819
to withstand extreme seismic and weather events.
00:05:56.180 --> 00:05:58.920
Because, look, uptime is non -negotiable for
00:05:58.920 --> 00:06:00.480
these hyperscalers. Right. They can't afford
00:06:00.480 --> 00:06:03.120
to go offline for a hurricane. Exactly. But the
00:06:03.120 --> 00:06:06.120
real cost driver is the MEP. The mechanical,
00:06:06.240 --> 00:06:08.220
electrical, and plumbing traits. Oh, the MEP
00:06:08.220 --> 00:06:11.259
must be insane. It is. In a standard commercial
00:06:11.259 --> 00:06:15.620
build, MEP might be what? 20 to 30 % of the budget.
00:06:15.680 --> 00:06:17.860
Yeah, around there. In a hyperscale data center,
00:06:18.000 --> 00:06:22.459
it can easily eclipse 50 or 60%. Jeez. You're
00:06:22.459 --> 00:06:25.439
installing massive, redundant power substations.
00:06:27.669 --> 00:06:30.149
uninterruptible power supply rooms that are just
00:06:30.149 --> 00:06:32.629
filled to the brim with lithium -ion batteries.
00:06:32.970 --> 00:06:35.170
And layers of concurrent maintainability, I'm
00:06:35.170 --> 00:06:37.910
assuming. Yes, exactly. Meaning they have to
00:06:37.910 --> 00:06:39.709
be able to shut down and perform maintenance
00:06:39.709 --> 00:06:42.430
on any single piece of infrastructure without
00:06:42.430 --> 00:06:45.310
ever dropping power to the server floor. Which
00:06:45.310 --> 00:06:47.310
means the geographic footprint of where this
00:06:47.310 --> 00:06:49.870
money is landing isn't random. Not at all. Like
00:06:49.870 --> 00:06:51.529
we said, you can't just drop a submarine in the
00:06:51.529 --> 00:06:53.490
middle of a desert without a supply line. Nope.
00:06:53.569 --> 00:06:55.649
According to the Smart Business Automator Market
00:06:55.649 --> 00:06:58.759
Intelligence Report, Roughly 60 % of all new
00:06:58.759 --> 00:07:01.199
starts are captured by just five states. Just
00:07:01.199 --> 00:07:04.139
five. Virginia, Texas, Louisiana, Illinois, and
00:07:04.139 --> 00:07:07.279
North Carolina. Now, I'd assume this hyper -concentration
00:07:07.279 --> 00:07:11.060
is just about cheap land and grid access in places
00:07:11.060 --> 00:07:14.699
like Texas or Louisiana, but there has to be
00:07:14.699 --> 00:07:17.100
a fiber -optic component driving Virginia and
00:07:17.100 --> 00:07:20.370
Illinois, right? Spot on. It's the intersection
00:07:20.370 --> 00:07:22.610
of path of least resistance energy grids and
00:07:22.610 --> 00:07:25.009
legacy fiber networks. Okay, that makes sense.
00:07:25.189 --> 00:07:27.910
So Northern Virginia, for example, is the undisputed
00:07:27.910 --> 00:07:30.949
data center alley of the world. And that's because...
00:07:31.240 --> 00:07:33.639
Decades ago, the foundational fiber backbone
00:07:33.639 --> 00:07:36.279
of the internet was laid there. Oh, right. And
00:07:36.279 --> 00:07:38.220
then Illinois serves as the critical switching
00:07:38.220 --> 00:07:41.240
hub for the entire Midwest network. Got it. Meanwhile,
00:07:41.379 --> 00:07:43.620
you've got Texas and Louisiana offering vast
00:07:43.620 --> 00:07:46.759
tracts of developable land and deeply entrenched,
00:07:46.819 --> 00:07:49.620
heavily deregulated energy infrastructure. So
00:07:49.620 --> 00:07:52.160
you combine land, power, and fiber, and boom,
00:07:52.220 --> 00:07:55.050
you get that 60 % concentration. Exactly. But
00:07:55.050 --> 00:07:57.990
the scale of the individual projects within those
00:07:57.990 --> 00:08:00.750
zones and the immediate spillover markets is
00:08:00.750 --> 00:08:02.649
just staggering. Right. Let's look at a named
00:08:02.649 --> 00:08:04.629
example from the intelligence report. Skansky
00:08:04.629 --> 00:08:08.449
USA recently booked a $255 million contract for
00:08:08.449 --> 00:08:11.310
a Georgia data center. But the kicker here, the
00:08:11.310 --> 00:08:14.089
really wild part, is the classification of that
00:08:14.089 --> 00:08:17.750
contract. It is a supplemental contract. Yeah.
00:08:17.810 --> 00:08:20.410
Let that sink in for a second. A $255 million
00:08:20.410 --> 00:08:24.240
change order, essentially. Right. I mean... For
00:08:24.240 --> 00:08:26.920
a regional general contractor, a quarter of a
00:08:26.920 --> 00:08:29.600
billion dollars is a massive company -defining
00:08:29.600 --> 00:08:32.399
anchor project. You retire on that. Yeah. But
00:08:32.399 --> 00:08:35.000
in the hyperscale data center world, it is simply
00:08:35.000 --> 00:08:38.059
an add -on to an existing scope of work. The
00:08:38.059 --> 00:08:40.779
project specifics on that Skanska build highlight
00:08:40.779 --> 00:08:44.100
exactly why these costs inflate so much. What
00:08:44.100 --> 00:08:47.320
are the specs? So it's 245 ,000 square feet comprising
00:08:47.320 --> 00:08:49.899
five data halls, and it's targeted for completion
00:08:49.899 --> 00:08:52.940
by the first quarter of 2028. Wow. We are looking
00:08:52.940 --> 00:08:55.580
at multi -year phasing here. And that extended
00:08:55.580 --> 00:08:57.419
timeline, that's driven by the commissioning
00:08:57.419 --> 00:08:59.519
process. Oh, sure. Because you don't just build
00:08:59.519 --> 00:09:02.740
245 ,000 square foot shell, flip a switch and
00:09:02.740 --> 00:09:05.519
hand over the keys. I wish. Yeah, right. No,
00:09:05.580 --> 00:09:08.080
you build Hall 1. You rigorously test the cooling
00:09:08.080 --> 00:09:09.980
and power loads under artificial stress. They
00:09:09.980 --> 00:09:12.240
often call this Level 5 commissioning. And you
00:09:12.240 --> 00:09:14.480
hand it over while you're literally still pouring
00:09:14.480 --> 00:09:17.379
concrete for Hall 3. Right. So by stretching
00:09:17.379 --> 00:09:20.240
these timelines out to 2028, contractors are
00:09:20.240 --> 00:09:22.759
locking up their best superintendents, their
00:09:22.759 --> 00:09:27.059
top project managers, local labor pools for half
00:09:27.059 --> 00:09:30.279
a decade. Yes. you are effectively building a
00:09:30.279 --> 00:09:32.799
temporary mini -city of construction logistics
00:09:32.799 --> 00:09:36.600
that just hovers over a single geographic pocket
00:09:36.600 --> 00:09:40.340
for years. Which creates an absolute vacuum for
00:09:40.340 --> 00:09:42.840
local subcontractors. A total vacuum. If you're
00:09:42.840 --> 00:09:45.100
an electrical or mechanical sub in Georgia or
00:09:45.100 --> 00:09:47.320
one of those top five states, you are suddenly
00:09:47.320 --> 00:09:50.179
competing against these multi -year behemoth
00:09:50.179 --> 00:09:52.980
projects that can afford to pay heavy premiums
00:09:52.980 --> 00:09:54.899
for labor. They absolutely will pay the premium.
00:09:55.279 --> 00:09:57.879
Because the end client, usually some massive
00:09:57.879 --> 00:10:00.720
tech conglomerate, is losing millions in potential
00:10:00.720 --> 00:10:03.379
AI revenue for every day that facility isn't
00:10:03.379 --> 00:10:05.659
online. Right. They will pay to accelerate the
00:10:05.659 --> 00:10:07.879
schedule. Exactly. So if you are a construction
00:10:07.879 --> 00:10:09.840
business owner operating in one of these target
00:10:09.840 --> 00:10:12.240
states, your strategic planning has to account
00:10:12.240 --> 00:10:14.840
for labor hoarding. Labor hoarding. You either
00:10:14.840 --> 00:10:17.240
figure out how to bid into these hyperscale supply
00:10:17.240 --> 00:10:20.100
chains yourself, or you have to completely adjust
00:10:20.100 --> 00:10:22.220
your pricing models for local commercial work.
00:10:22.509 --> 00:10:24.990
Because your labor costs are going to be artificially
00:10:24.990 --> 00:10:27.370
inflated by the data center campus down the road.
00:10:27.529 --> 00:10:29.070
Yep. They'll pull all your guys if you don't
00:10:29.070 --> 00:10:31.730
adjust. Wow. Okay, that brings us to the most
00:10:31.730 --> 00:10:34.450
critical pivot of this entire deep dive. Here
00:10:34.450 --> 00:10:37.409
we go. You have the joint venture formed. You
00:10:37.409 --> 00:10:40.049
have the bonding secured. You found the land
00:10:40.049 --> 00:10:43.539
in Virginia or Georgia. And the $101 billion
00:10:43.539 --> 00:10:46.559
in pre -construction capital is ready to deploy.
00:10:46.860 --> 00:10:48.639
You're ready to dig. You're ready. And then you
00:10:48.639 --> 00:10:51.000
hit a wall. Yeah. And it's not the local power
00:10:51.000 --> 00:10:53.200
utility telling you they don't have the megawatts.
00:10:53.200 --> 00:10:55.899
It's the water board. The water wall. The water
00:10:55.899 --> 00:10:58.279
wall. This is where the physics of artificial
00:10:58.279 --> 00:11:01.519
intelligence completely alters the construction
00:11:01.519 --> 00:11:03.940
blueprint. To illustrate how massive this wall
00:11:03.940 --> 00:11:07.360
is, let's look at Memphis. XAI has committed
00:11:07.360 --> 00:11:10.940
to resuming an $80 million graywater recycling
00:11:10.940 --> 00:11:13.860
plant for its Colossus campus in Memphis. $80
00:11:13.860 --> 00:11:16.399
million. $80 million just for the water facility.
00:11:16.759 --> 00:11:19.000
The capacity specs on that facility tell the
00:11:19.000 --> 00:11:21.080
real story, though. Oh, yeah. It is designed
00:11:21.080 --> 00:11:24.179
to process up to 13 million gallons of water
00:11:24.179 --> 00:11:27.440
a day. 13 million gallons a day isn't a data
00:11:27.440 --> 00:11:29.960
center. That is municipal scale. It really is.
00:11:30.059 --> 00:11:32.240
That is enough water to supply a small city.
00:11:32.879 --> 00:11:35.840
Why on earth does a supercomputer need the water
00:11:35.840 --> 00:11:38.620
pressure of a municipality? Well, it comes down
00:11:38.620 --> 00:11:41.299
to thermal dynamics, right? Yeah. And the shift
00:11:41.299 --> 00:11:44.559
from traditional cloud computing to artificial
00:11:44.559 --> 00:11:47.139
intelligence. Okay, break that down for us. So,
00:11:47.159 --> 00:11:50.159
traditional server racks can generally be cooled
00:11:50.159 --> 00:11:53.149
with ambient air. You pump chilled air up through
00:11:53.149 --> 00:11:55.710
a raised floor, it passes over the servers, and
00:11:55.710 --> 00:11:57.830
you vent the hot air out. Standard AC stuff.
00:11:58.210 --> 00:12:01.929
Right. But the new GPU clusters used for AI training,
00:12:02.090 --> 00:12:05.639
they run so incredibly hot. The air cooling is
00:12:05.639 --> 00:12:08.360
physically incapable of rejecting the heat fast
00:12:08.360 --> 00:12:11.039
enough. Oh, wow. So they have to use liquid cooling.
00:12:11.240 --> 00:12:12.919
Exactly. They have to use direct -to -chip liquid
00:12:12.919 --> 00:12:15.480
cooling and massive rear -door heat exchangers.
00:12:15.600 --> 00:12:17.159
Wait, so they pump it directly to the chips?
00:12:17.320 --> 00:12:19.379
Yep. You are pumping fluids directly through
00:12:19.379 --> 00:12:21.740
the server chassis itself, and then that heated
00:12:21.740 --> 00:12:24.340
fluid is piped out to massive evaporative cooling
00:12:24.340 --> 00:12:26.700
towers on the roof or the perimeter of the facility.
00:12:26.960 --> 00:12:29.779
And to keep those processors from melting down,
00:12:30.350 --> 00:12:33.129
Those towers evaporate millions of gallons of
00:12:33.129 --> 00:12:35.450
water into the atmosphere every single day. Which
00:12:35.450 --> 00:12:38.129
leads directly to the core conflict in Memphis.
00:12:38.860 --> 00:12:41.379
The primary reason stated for this $80 million
00:12:41.379 --> 00:12:44.559
water plant is to ensure the supercomputer's
00:12:44.559 --> 00:12:47.600
immense cooling needs do not draw down the city
00:12:47.600 --> 00:12:50.620
of Memphis' drinking water aquifer. Right. Because
00:12:50.620 --> 00:12:53.059
Memphis relies on the Memphis sand aquifer. Yeah.
00:12:53.159 --> 00:12:55.639
It is a pristine source of municipal drinking
00:12:55.639 --> 00:12:58.440
water. Right. You cannot have a private corporation
00:12:58.440 --> 00:13:02.220
drawing 13 million gallons of pristine aquifer
00:13:02.220 --> 00:13:05.460
water a day just to cool... Graphic processing
00:13:05.460 --> 00:13:07.919
units. People would riot. The municipal grid
00:13:07.919 --> 00:13:10.500
would fail, and yet the public backlash would
00:13:10.500 --> 00:13:13.580
be catastrophic. Yeah. So the only viable engineering
00:13:13.580 --> 00:13:15.919
solution is gray water. Gray water being treated
00:13:15.919 --> 00:13:18.440
wastewater. Exactly. Water that isn't safe for
00:13:18.440 --> 00:13:20.419
human consumption, but is perfectly acceptable
00:13:20.419 --> 00:13:22.539
to run through an industrial heat exchanger.
00:13:22.639 --> 00:13:25.360
Exactly right. But getting 13 million gallons
00:13:25.360 --> 00:13:28.179
of gray water to a data center every day, that
00:13:28.179 --> 00:13:30.259
requires heavy infrastructure. Yeah, you can't
00:13:30.259 --> 00:13:34.149
just run a hose. No. You have to intercept municipal
00:13:34.149 --> 00:13:37.289
wastewater. You have to build a dedicated treatment
00:13:37.289 --> 00:13:41.309
facility on or near the campus, filter it, chemically
00:13:41.309 --> 00:13:44.389
treat it to prevent scale and biological growth
00:13:44.389 --> 00:13:46.730
in the cooling loops. Oh, because algae and stuff?
00:13:46.909 --> 00:13:48.409
Right. You can't have algae in a supercomputer.
00:13:48.649 --> 00:13:51.309
You use it, and then you often have to treat
00:13:51.309 --> 00:13:53.450
the blowdown water before discharging it back
00:13:53.450 --> 00:13:55.470
into the municipal sewer system. See, that is
00:13:55.470 --> 00:13:58.809
the ultimate aha moment here. The future of tech
00:13:58.809 --> 00:14:01.149
construction isn't digital, it's heavy civil.
00:14:01.370 --> 00:14:04.230
Yes. If an $80 million water treatment plant
00:14:04.230 --> 00:14:06.669
is now just a prerequisite for building an AI
00:14:06.669 --> 00:14:10.049
campus, the very definition of a data center
00:14:10.049 --> 00:14:12.549
contractor has fundamentally changed. It really
00:14:12.549 --> 00:14:15.009
has. I mean, referencing Smart Business Automator
00:14:15.009 --> 00:14:17.070
again here, you look at the trends. If a build
00:14:17.070 --> 00:14:19.210
can drain a city's drinking aquifer, you are
00:14:19.210 --> 00:14:21.529
no longer just bidding on clean rooms, steel
00:14:21.529 --> 00:14:23.950
framing, and heavy copper wire. Right. You're
00:14:23.950 --> 00:14:26.049
bidding on miles of large diameter underground
00:14:26.049 --> 00:14:28.710
trenching. Yep. You are pouring concrete for
00:14:28.710 --> 00:14:30.850
massive clarifiers. You're building chemical
00:14:30.850 --> 00:14:33.769
dosing stations, complex pump houses, and biological
00:14:33.769 --> 00:14:36.450
membrane filtration systems. You are building
00:14:36.450 --> 00:14:39.950
an industrial wastewater treatment plant. right
00:14:39.950 --> 00:14:42.169
next to the most sophisticated computing hardware
00:14:42.169 --> 00:14:45.409
on Earth. And the tolerances are zero. Zero.
00:14:45.570 --> 00:14:47.929
If the water plant goes offline, the cooling
00:14:47.929 --> 00:14:50.549
loops run dry. If the cooling loops run dry,
00:14:50.750 --> 00:14:53.210
the supercomputer fries in a matter of minutes.
00:14:53.490 --> 00:14:55.750
It's all connected. So the contractors who are
00:14:55.750 --> 00:14:58.070
actually going to capture a piece of this $101
00:14:58.070 --> 00:15:01.409
billion pre -construction pipeline aren't just
00:15:01.409 --> 00:15:03.649
the ones who know how to build a highly secure
00:15:03.649 --> 00:15:07.090
warehouse. They are the ones with deep, heavy,
00:15:07.129 --> 00:15:10.899
civil... earthwork, and utility expertise. Or,
00:15:11.039 --> 00:15:13.860
you know, the ones strategic enough to form the
00:15:13.860 --> 00:15:16.580
right partnerships. Very true. If you're a premier
00:15:16.580 --> 00:15:19.679
commercial builder. You need to be actively joint
00:15:19.679 --> 00:15:22.419
venturing with the heavy civil contractors who
00:15:22.419 --> 00:15:24.559
traditionally build municipal infrastructure.
00:15:24.679 --> 00:15:26.700
Absolutely. Because this water wall is the new
00:15:26.700 --> 00:15:29.059
reality. You can secure all the silicon. You
00:15:29.059 --> 00:15:31.019
can lock down all the electricity. But if you
00:15:31.019 --> 00:15:33.419
cannot engineer and execute the water rights
00:15:33.419 --> 00:15:36.279
and the physical water infrastructure, your mega
00:15:36.279 --> 00:15:38.179
project will never make it out of pre -construction.
00:15:38.299 --> 00:15:40.639
It's dead in the water. Literally. Literally.
00:15:40.740 --> 00:15:44.019
Okay, let's synthesize this actionable intelligence
00:15:44.019 --> 00:15:46.179
for the construction business owners listening
00:15:46.179 --> 00:15:48.360
right now. Let's do it. We started by looking
00:15:48.360 --> 00:15:52.960
at a $58 .1 billion tidal wave of year -to -date
00:15:52.960 --> 00:15:55.860
spending, which is moving at four times the speed
00:15:55.860 --> 00:15:58.940
of previous records and demanding totally new
00:15:58.940 --> 00:16:01.480
risk management and bonding strategies. Right.
00:16:01.539 --> 00:16:04.720
And we followed that capital into a $101 billion
00:16:04.720 --> 00:16:07.850
pre -construction pipeline. We know from the
00:16:07.850 --> 00:16:09.750
Smart Business Automator Market Intelligence
00:16:09.750 --> 00:16:13.350
Report that 60 % of this work is hyper -concentrated
00:16:13.350 --> 00:16:16.309
in five key states. Virginia, Texas, Louisiana,
00:16:16.629 --> 00:16:19.129
Illinois, and North Carolina. Exactly. And we
00:16:19.129 --> 00:16:22.309
established that at a median build cost of $473
00:16:22.309 --> 00:16:25.470
a square foot. These projects are incredibly
00:16:25.470 --> 00:16:28.389
dense, multi -year commitments. Yeah. As we saw
00:16:28.389 --> 00:16:31.590
with Skanska's $255 million supplemental contract
00:16:31.590 --> 00:16:34.549
in Georgia, a single phase of a single campus
00:16:34.549 --> 00:16:37.230
can tie up labor and project management resources
00:16:37.230 --> 00:16:40.250
until 2028. Such a massive footprint. It really
00:16:40.250 --> 00:16:42.090
is. And ultimately, we uncovered that the barrier
00:16:42.090 --> 00:16:43.889
to entry has evolved from electrical capacity
00:16:43.889 --> 00:16:46.570
to municipal water management. The big pivot.
00:16:46.690 --> 00:16:49.750
To compete at this scale, contractors must possess
00:16:49.750 --> 00:16:53.450
or partner for heavy utility and water treatment
00:16:53.450 --> 00:16:56.649
expertise to handle the massive evaporative cooling
00:16:56.649 --> 00:16:59.980
requirements of AI infrastructure. It's a wildly
00:16:59.980 --> 00:17:02.879
lucrative space, but only for the firms agile
00:17:02.879 --> 00:17:05.039
enough to realize that building a data center
00:17:05.039 --> 00:17:07.839
today means building a water plant tomorrow.
00:17:08.200 --> 00:17:11.299
The traditional commercial blueprint just doesn't
00:17:11.299 --> 00:17:13.839
apply here anymore. The physics of the build
00:17:13.839 --> 00:17:15.859
have been completely rewritten. They really have.
00:17:16.019 --> 00:17:18.119
So I want to leave you with a final thought to
00:17:18.119 --> 00:17:20.720
mull over. Something that wasn't explicitly spilled
00:17:20.720 --> 00:17:22.819
out in the data, but is the absolute logical
00:17:22.819 --> 00:17:25.380
conclusion of this water constraint. We know
00:17:25.380 --> 00:17:27.579
municipalities like Memphis are forcing developers
00:17:27.579 --> 00:17:30.009
to build their own greywater infrastructure to
00:17:30.009 --> 00:17:32.819
protect drinking aquifers. But what happens to
00:17:32.819 --> 00:17:36.460
that massive $101 billion pre -construction pipeline
00:17:36.460 --> 00:17:39.380
if the municipalities in those top five states
00:17:39.380 --> 00:17:42.019
suddenly decide that even gray water is too precious
00:17:42.019 --> 00:17:44.460
and they cap commercial water usage entirely?
00:17:44.900 --> 00:17:47.839
Right. Will the next great data center boom force
00:17:47.839 --> 00:17:49.920
the construction industry to pioneer entirely
00:17:49.920 --> 00:17:53.099
new geographies near oceans or major river systems?
00:17:53.279 --> 00:17:55.920
Or will it force us to invent entirely new ways
00:17:55.920 --> 00:17:58.000
of building massive infrastructure without a
00:17:58.000 --> 00:17:59.039
drop of water at all?