Hornbeck said:
I had five main points arguing against the bitcoin / ai / whatever data center in Midtown CS.
1. Power. We don't have that kind of power. The Dansby Plant makes 206MW running full blast. The CS data centers were going to use 3x that amount of power.
2. Water. They could not give us a straight answer on how much. "We'll work that out in the due diligence phase"... Ummmm, no. We're already having to drill two new wells at a cost of $70M. Are we going to need more?
3. Noise. If they use all the noise abatement in the world, the backup generators (diesel or gas turbines) will make noise when testing and in power outages.
4. Neighborhood Integrity. Putting this thing next to residential neighborhoods, and across the street from a retirement community is lunacy. Common sense should dictate this is a bad idea. There's a very good chance property values around this thing would tank.
5. Trust. In CS, several recent issues have made us distrust anything that is coming out of City Hall. This was just the latest example.
Now, out near RELLIS, a couple of these points won't be issues, but the water, power, etc., sure will be.
I used to do some work in datacenter planning and design, and can speak to a few of these. Please note, I'm not pro or anti datacenter; I do think that the vast majority of crypto, bitcoin and AI are absolute wastes of resources.
1. Power. In the size of datacenter being talked about, you aren't talking about being a typical customer of a utility; you become your own utility. You're an entity pulling from the high voltage transmission system with your own substation and power handling equipment. In effect, you're similar to a new "city" coming online. Knowing this, the comments about power available at Dansby really don't work that way. BTU owns generation, transmission and distribution. All of these typically function as separate entities in utilities that have all three. The power that Dansby puts on the transmission grid isn't limited to this area or any area in particular. BTU generates power, puts it on the grid, and gets money from the utilities that consume it, even the part of BTU that owns distribution. Note: I don't know if they have a net metering setup with ERCOT, or if the two entities are fully separated to ERCOT with Generation getting revenue in and Distribution paying power purchase fees out, but the utility I worked with in the datacenter I worked on was fully separated. In fact, I'd bet willing to bet that BTU service area's peak load is in the 300MW range, so during such a time, even if Dansby was generating to 100%, BTU was still pulling power in from elsewhere. In the immediate area, there's the Blue Jay 1 Solar at 210MW and Tenaska just outside of Shiro at almost 1GW. Power generation goes onto the HV grid and large scale data centers pull directly from there, sometimes with a utility partner acting as their official representative to the ISO or sometimes that do that entirely themselves. The availability of power typically isn't restricted by generation in the area but by HV transmission in the area, as the actual generation can take place fairly far away, as long as their load is accounted for (within some reason; there is planning and infrastructure involved in long distance electric transmission). Heck, the power from Oak Grove part Hearne could probably absorb most of that formerly-proposed datacenter's load. Large transmission level customers and utilities both do the same thing and pay for load availability; they pay what are essentially power brokers to ensure that someone on the ERCOT grid, typically within the area that the transmission network is built up enough to serve the power load, generates power and puts it on the grid to cover their expected load.
That would be why some datacenters look to co-locate with on-site generation, especially if that generation isn't grid connected or is recently retired. The grid planners typically have taken that generation out of planning for cities and such, so the datacenter can spin it back up to generate power directly without having to go through brokers or load planning as much, and they can (in some cases) exempt themselves from load shed events (think Uri) since their power source isn't part of the grid planning. Additionally, if they really want to extra headache, they can even use any surplus generation to feed onto the grid and have a secondary revenue source. Off-grid datacenters are the new hotness last I saw, and do solve the problems of the datacenter stressing the electrical grid where they're located. Which is good, because the planning for those is much more rural usually.
2. Water. This is a much tougher problem, but also not nearly as bad as it used to be. Newer datacenters use a closed-loop system which has a very minimal loss and after initial filling has a minimal draw. The system vendors claimed "less ongoing water usage after install then a typical fast food restaurant". I'm not sure I'd believe that claim entirely, but if it was a newer, high-efficiency closed loop system, it should've been able to get to that or close to that. What they were proposing, who knows.
Water is a lot harder on a local utility system. They're much more consume-what-you-produce than the electric grid. Most utilities don't have a strong interconnection with water sources outside of what they own or can directly pipe from. If they do have an interconnection, it's usually minimal, and emergency-only type of connections. Honestly, this level of separation is good for us, because water is a lot harder to "generate" than electricity, and if we did have a wide interconnection, then it would've been a lot easier for San Antonio or Georgetown to just pipe our water here out. Inland desalination with piped seawater may have to be the future for water, both for cities and datacenters.
3. Noise. That is right that the primary source of noise from modern datacenters is from the backup generators running tests or running during outage events. The datacenter itself can be made very quiet and you'd have more noise typically from the HVAC system's heat exchangers outside than the actual computers and related equipment inside. Now inside; definitely a different story.
I would say though that even the generators aren't typically too bad. Running the 1MW class generators which would likely be the largest there (since in a true outage, the one here couldn't have had enough generation to run the site in production mode; outage mode would've been just enough equipment running to keep the network live and the monitoring systems and HVAC in standby modes) can definitely be made to be quiet from any reasonable distance.
From my office I've heard lawnmowers from the groundscrew that were louder than the gensets running from across the parking lots. And in a true outage, about half of my neighbors in my neighborhood are getting their own generators out, putting them on the driveway and running them, so the occasional din of a 30 minute genset test once a week or, more likely, a month isn't a super high concern to me. I understand that others would and that's totally valid as well. I would just say that in the datacenters I've worked at, the external noise was virtually non-existent, though I do know that some aren't that way.
4 and 5 - These are completely valid, especially the almost farcical level of secrecy it had until the very last moment. As far as what it would do to property values; I can't entirely say. If this area does develop into a higher activity, higher traffic commercial area, that bring with it a separate set of nuisances that the very low traffic a datacenter has would receive. Its an academic point now, but is one that definitely should've been researched and fleshed out long before this came to a vote like this.
