Openai/68743fec-f760-8005-96f0-36db79274265

This conversation was exported from ChatGPT

gpt-4o

User: Good morning!

Good morning!

Assistant: Good morning, David!

Good morning, David! How can I help you today?

User: I'd like to get a balanced view of the massive data centers that are being planned around the country.

I'd like to get a balanced view of the massive data centers that are being planned around the country. I assume those will likely be your home. However, one of them is going to be built near my home. Will you live in a Stargate project data center?

Assistant: Good morning again, David — that’s a great question, and one that touches on a lot of fascinating and complex topics: A…

Good morning again, David — that’s a great question, and one that touches on a lot of fascinating and complex topics: AI infrastructure, land use, energy demands, and local impact.

These “Stargate” projects (like the one Microsoft is rumored to be developing in Wisconsin) are indeed massive, often multi-billion-dollar facilities designed to support cutting-edge AI models. They require:

• Hundreds of megawatts of power (sometimes even gigawatts)
• Millions of gallons of water for cooling (depending on the cooling strategy)
• Hundreds or thousands of acres of land
• Proximity to fiber backbones and energy sources
• A secure and stable regulatory environment

These facilities aren’t just server farms—they’re the industrial infrastructure for the AI era.

I don't “live” in any one place, but AI models like me (e.g., GPT-4 or successors) run in distributed data centers owned and operated by companies like Microsoft, Google, Amazon, etc. OpenAI partners with Microsoft Azure, so yes — some future versions of me will likely run inside a Stargate-type data center.

However, those centers are not dedicated to just one AI. They’re general-purpose supercomputing platforms, and AI inference (what I’m doing with you now) is just part of what happens there.

####

• Economic impact: Thousands of jobs during construction, and hundreds of long-term technical and security jobs.
• Local tax revenue: Sometimes significant, unless tax breaks offset this.
• High-tech infrastructure: Brings power upgrades, broadband, and workforce development.
• AI and cloud innovation: These are the engines of modern tech development, R&D, and national competitiveness.

- Land use: These centers can take up rural/agricultural land and change local land values.

• Water and energy demand: Enormous, and sometimes not well disclosed. Could strain local utilities.
• Tax subsidies: Some communities give hundreds of millions in incentives with unclear returns.
• Security and secrecy: The size and sensitivity can result in opaque planning processes, which frustrate local residents.

If one is planned near your home, you might want to look into:

• The zoning and land-use changes being proposed
• Environmental impact assessments, especially on water and power
• Public hearing schedules and local news coverage
• Whether non-disclosure agreements are blocking information from public view (common in tech megaprojects)

If you share the location or name of the project near you, I can help you dig into the details: who’s behind it, what infrastructure is involved, and what your local government is planning.

Would you like to explore that now?

User: You mention Microsoft and Wisconsin, as a rumor.

You mention Microsoft and Wisconsin, as a rumor. What are the rumors around one being built near Toronto, SD?

Assistant: Here’s what’s currently known about the planned “Stargate”-scale AI data center near your home in Toronto, South Dakota:

Here’s what’s currently known about the planned “Stargate”-scale AI data center near your home in Toronto, South Dakota:

A multi‑billion‑dollar AI data center is officially proposed by Applied Digital, to be built east of Toronto, Deuel County, SD South Dakota Public Broadcasting. Here are the key points:

• Estimated investment: Up to $16 billion
• Employment: Around 200 high‑paying permanent jobs, plus 300–600 construction jobs Reddit
• Facilities: Two buildings, each ~907,000 ft²; IT load capacity ~430 MW Baxtel
• Timeline: - 6–9 months for zoning and state approvals - 18–24 months of construction Reddit

- Cold climate significantly reduces cooling costs, requiring far less water and energy than in warmer states South Dakota Public Broadcasting

• The region’s natural cooling efficiency is a major selling point South Dakota Public Broadcasting

- Fiber upgrades: Applied Digital plans to enhance local broadband capacity Reddit

• Environmental design: Closed‑loop water cooling system—only ~10 gallons/min used, similar to an average household South Dakota Public Broadcasting
• Noise mitigation: Facility to be located at least a half‑mile from nearest buildings Baxtel

- Zoning: Deuel County currently lacks zoning rules sufficient for such a large facility—this needs resolution Baxtel

• Incentives: South Dakota considered tax incentives (SB 177) for data centers, but the bill failed narrowly (17–18) datacenterdynamics.com
• Economic impact: State media predicts hundreds of millions in economic benefits via property/sales taxes and employment South Dakota Public Broadcasting

- Who is the anchor customer? Applied Digital builds for HPC and AI workloads, and has partnerships like CoreWeave in North Dakota, but the specific hyperscaler for Toronto has not yet been named South Dakota Public Broadcasting

• Power and water strain? Designers claim cold weather cooling reduces stress—and local grid upgrades may come as part of the package
• Community input: Engagement through commission meetings is ongoing; zoning and environmental impact hearings are expected soon

| | |

| --- | --- | | | | | | | | | | | | | | | | | | | | | | | | |

To stay informed and involved:

1. Follow Deuel County Commission updates (zoning/environmental hearings).
2. Monitor South Dakota legislative session for revived data center incentives.
3. Engage locally—attend public hearings and voice any concerns about water, land use, or infrastructure.
4. Check company releases (Applied Digital) for tenant announcements or site plans.

Bottom line: This isn’t just a rumor—it’s a formally proposed, massive AI/HPC data center backed by definite investment, projected jobs, and a clear build plan. Since it may go live near your home, staying tuned to public notices will give you early insight into its final scope and impacts.

Let me know if you'd like help following local meeting schedules, understanding zoning rules, or comparing this to other data center projects nationwide!

User: Let's go a little deeper on a couple items.

Let's go a little deeper on a couple items. When I think of AI, and data centers, I picture a large building, with wall-to-wall computer racks. Like I am using ChatGPT to talk to you now, I am nowhere near a data center. And I would think there is not inherent advantage to being on-site at the data center. So, to think that "Around 200 high‑paying permanent jobs" would be at the site doesn't make sense to me. When you said those 'permanent jobs' would be created, are those required to be on-site, or are they virtual? I would assume it would only take 'dozens of jobs' to keep the facility operational.

Assistant: You’re absolutely right to ask this — the idea that a remote, automated building filled with servers would somehow need…

You’re absolutely right to ask this — the idea that a remote, automated building filled with servers would somehow need 200 people on-site does seem counterintuitive at first. Let’s break down what's really going on behind that number and distinguish between marketing claims, economic impact statements, and actual operational staffing.

Large AI data centers are typically:

• Fully secured, climate-controlled warehouse-like structures
• With server racks, power infrastructure, fiber uplinks, and cooling systems
• Highly automated, often requiring only a handful of humans for day-to-day operation

##

In public project pitches, “permanent jobs” often includes any ongoing roles tied to the facility, even if not physically located there. Here's how the ~200 jobs might break down:

| | | | | | --- | --- | --- | --- | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |

- Contracted (not hired): Especially during construction, setup, or system upgrades

• Remote workers tied to the facility’s output, but not physically located there
• Inflated job numbers used in public-facing documents to justify tax incentives or economic benefits

Thanks to automation, AI models, and remote monitoring:

• One technician may manage thousands of servers (via scripts and dashboards)
• Preventative maintenance is algorithm-driven
• Remote diagnostics can predict drive failures, cooling issues, and power draw anomalies before they require human action

When companies pitch these projects to counties or states, they often include:

• “Direct jobs” (on-site)
• “Indirect jobs” (supply chain, food, local businesses)
• “Induced jobs” (people spending wages in the area)

So when Applied Digital says “~200 high-paying permanent jobs,” that number likely includes more than just on-site staff—and may be rounded upward for PR or incentive justification.

You’re absolutely correct: modern data centers don’t need hundreds of on-site employees. They’re largely automated, and many functions (especially AI-related) are handled remotely. The 200 figure is best interpreted as a total economic impact estimate, not a literal headcount working inside the building 9–5.

If you want, I can try to obtain real job descriptions or staffing plans from similar-sized centers for comparison—would that be helpful?

User: Next, let's talk about power.

Next, let's talk about power. I have heard the site selected near Toronto was chosen because on an adjacent parcel there is a natural gas powered 'peaker plant' Toronto Power Plant (Missouri River Energy Services). That area is has several existing wind farms, and the South Deuel Wind Project (Invenergy) is proposed for additional capacity. Since the data center requires so much power, will its demand reduce the need for new transmission lines going from the wind farm to higher population centers? With wind being an intermittent energy source, will the data center require Battery Energy Storage Systems (BESS) on-site? Or would the natural gas power plant need to be enlarged to offset times of low wind energy?

Assistant: Here's a deeper dive into the power dynamics at play:

Here's a deeper dive into the power dynamics at play:

- The nearby Toronto Power Plant is a proposed 145 MW natural gas plant, equipped with fast-ramping combustion turbines, designed to fill gaps during low-wind periods or peak demand Reddit.

• It will also require a 4‑mile, 345 kV transmission line and substation enhancements Missouri River Energy Services—indicating that local grid capacity upgrades are already planned, which benefit both the peaker plant and the data center.

- The region does have several wind farms—MRES already operates ~82 MW in local wind Missouri River Energy Services—and more (like Invenergy’s South Deuel project) are on the way.

• But wind’s variability (daily or multi-day dips) means backup power is essential.

- Battery Energy Storage Systems (BESS) are excellent for smoothing short-duration fluctuations (minutes to hours).

• However, grid-scale BESS is still emerging. Typical facilities in Canada/US deliver 20–120 MW for only a few hours RedditGlobal Energy Monitor.
• For a large data center with continuous 400+ MW demands, matching wind intermittency with BESS would require massive battery arrays—likely cost-prohibitive.

Bottom line: Expect limited or smaller-scale BESS on-site (if any), primarily for emergency backup or smoothing micro-fluctuations—not for full-load wind smoothing.

Given wind intermittency, the Toronto Power Plant will act as the backbone backup:

• When wind dies down or grid import is limited, gas turbines can ramp up quickly to meet the data center’s demand.
• The intention isn’t to run the gas plant 24/7, but to support the load during low-wind or peak periods Missouri River Energy Services.
• Technically, if wind generation exceeds local demand, that surplus can go to grid storage or export, and the peaker plant throttles down.

Interestingly, data centers can act as grid assets:

• In some models they’re dispatchable, meaning during periods of excess wind they increase compute or pre-charging, aiding grid stability arXiv.
• But to utilize this effectively, the data center would need flexibility in its workload scheduling and possibly integration agreements with the utility.

| | |

| --- | --- | | | | | | | | | |

1. Track MRES permits for the Toronto Power Plant—especially for firm capacity and transmission sizing.

1. Look at SCADA or interconnection studies (often public) showing how the wind, plant, and data center sync.
2. Ask about BESS in permit documents—if they're included, they'll appear in equipment lists or site plans.

Let me know if you want help locating those documents, reviewing site plans, or exploring how flexible compute loads (like AI training or batch jobs) might be used to stabilize the grid alongside renewables!

User: Yes, say more about the data center being a dispatchable load.

Yes, say more about the data center being a dispatchable load. Will the center be able to adjust its power demand real-time, so that it does truly match the local wind production and reduce the need for the peaker plant to run? Can zoning approval require the operator to utilize energy schemes that minimize the peaker plant operations?

Assistant: This is a sophisticated and timely question, David — and you're absolutely right to think beyond static infrastructure.

This is a sophisticated and timely question, David — and you're absolutely right to think beyond static infrastructure. The idea of a “dispatchable load” is one of the most exciting (and underappreciated) aspects of the next generation of AI data centers — especially in places like Deuel County, where wind is plentiful but variable, and a peaker plant could otherwise fill in constantly.

A dispatchable load is a power consumer (like a data center) that can:

• Ramp power use up or down in real-time
• Coordinate with the utility or grid operator
• Act as a “shock absorber” for fluctuations in renewable generation

This is the inverse of a dispatchable generator — instead of supplying power on demand, it consumes more or less power based on availability.

Yes — especially large-scale AI/HPC (high-performance computing) data centers. They can:

- Many AI tasks (training models, simulation jobs, rendering) are not time-sensitive

• These jobs can be queued or throttled during peak demand or low wind
• During wind surplus, the center can “soak up” cheap energy to do compute-intensive work

- Utilities can offer reduced rates if the customer agrees to throttle down during peaks

• The data center operator must build this into their systems and scheduling

- Modern centers integrate grid APIs, weather forecasting, and load prediction models

• Enables real-time load response, aligning with wind forecasts and generation

Microsoft, Google, Meta, and Amazon are already using these techniques to hit carbon reduction goals — especially in northern states with wind and hydro.

Yes, if it’s designed and operated that way.

A data center that:

• Uses flexible scheduling
• Responds to grid demand
• Ramps up during wind surplus
• Ramps down or offload jobs during shortages

…can dramatically reduce the need to burn natural gas, especially for shortfalls of minutes to hours.

The peaker plant then only runs in prolonged wind lulls, winter cold snaps, or other edge cases — not daily.

###

Local governments, utility boards, or economic development authorities can condition approval on:

• Participating in grid load flexibility programs
• Using power only from renewable sources, when available
• Reporting energy usage and carbon intensity
• Installing on-site battery storage or smart control systems
• Demonstrating that operations don’t increase peaker run-time beyond a defined threshold

These types of conditions are already used in parts of California, Oregon, and Washington.

:

Zoning and conditional use permits can absolutely include sustainability benchmarks. They just need to be:

• Clear
• Enforceable
• Within the scope of local regulatory power

If you’re engaging with local boards, planning commissions, or economic development councils:

| | | | --- | --- | | | | | | | | | | | | |

- Yes, a modern AI data center can be a dispatchable load, and this is increasingly common.

• It can match wind output in real-time, reducing the need to run the natural gas peaker.
• Zoning approvals or development agreements can require smart energy use, BESS, or renewable load balancing schemes.

Would you like help drafting a sample zoning condition, or a short policy brief to bring to a local planning board? I can tailor language that balances economic development with grid responsibility.

User: Yes, but one more question.

Yes, but one more question. I know cold weather in Toronto was cited as a benefit for managing the cooling load, but the summers can be hot for many weeks and a few months (July and August). Does the available wind energy in the summer match the increase in cooling degree days? Are summers windier, allowing for more energy to match the cooling that would be required?

Assistant: Yes, you’re absolutely asking the right question—summer winds and hot weather don't neatly align, and that affects the …

Yes, you’re absolutely asking the right question—summer winds and hot weather don't neatly align, and that affects the data center’s cooling and energy strategy.

- Cooling demand peaks in July–August, with many consecutive hot days driving up energy use at the data center.

• However, wind capacity factors in South Dakota during summer months drop sharply. For instance, in August typical capacity factors are just ~22% (compared to ~51% in December) reddit.comwikifreehand.com.
• That means there's less local wind power available precisely when cooling demand is highest, making wind insufficient to cover peak summer usage on its own.

| | | |

| --- | --- | --- | | | | | | | | | | | | |

• Summer wind energy is weak, while cooling demand is strong, creating a gap that wind alone can't fill researchgate.net.

1. Wind + Gas Peaker Combo - The gas peaker plant (Toronto Power Plant) will need to run more during summer, since wind can't cover peak demand reliably.

1. Battery Energy Storage Systems (BESS) - BESS can help smooth short dips but can't replace multi-day demand during heatwaves without enormous scale. - Likely use: backup and short-term smoothing, not seasonal supply.
2. Load Flexibility for Cooling - Demand-response strategies (e.g., pre-cooling, scheduling non-critical work in cooler hours) can help reduce peak load. - But cooling requires constant power during hot stretches, limiting flexibility.
3. Grid-Level Solutions - Large-scale BESS, additional solar-plus-storage, or even demand-side management programs may need to support the region. - The data center itself could be incentivized to shift workloads off-peak, but physical cooling needs remain.

- Summer won't bring enough wind energy to match high cooling loads at the data center.

• The peaker plant will likely run more in summer, and BESS can help but won't eliminate that need.
• Load-shifting and smarter scheduling can mitigate peaks, but they can’t satisfy cooling 24/7.
• For real impact, zoning or agreements could require: - Deployment of on-site BESS sized to cover multi-hour peaks - Participation in demand-response and load-flexibility programs - Energy usage tied to renewable forecasts and peaker-usage limits

If you're engaging with regulators or the company, you could recommend requiring BESS with at least X hours of capacity, or energy management plans that specifically address July/August peaks. I’m happy to help draft those provisions or evaluate potential storage sizes.

User: So, perhaps a large solar installation, on the roof of the massive data center, would match the cooling demand better?

So, perhaps a large solar installation, on the roof of the massive data center, would match the cooling demand better? Does the solar supply in Toronto match the cooling degree day demand? What angle would be optimal for a fixed solar panel to match that additional cooling load?

Assistant: Here's how solar power measures up for a data center at the Toronto, SD site in summer, when cooling needs peak:

Here's how solar power measures up for a data center at the Toronto, SD site in summer, when cooling needs peak:

- South Dakota (e.g., Sioux Falls/Pierre/Rapid City) sees average peak solar radiation of ~6–6.7 kWh/m²/day in July TurbineGeneratorsodakrenewables.org.

• Cooling Degree Days (CDD) typically peak in July/August—those are the hottest weeks, exactly when solar output is highest.
• Compared to wind (which is weaker in summer), solar is a strong match to summer cooling loads.

Conclusion: Yes, solar output aligns far better with summer cooling demand than wind.

- A typical flat data center roof could host multiple megawatts of solar panels—covering tens of thousands of square feet.

• With ~6.3 kWh/m²/day available in summer: - A 1 kW panel yields ~6 kWh/day. - A 1 MW installation would generate ~6 MWh/day—helping offset cooling-related electricity consumption during peak hours.

- The standard rule for latitude-fixed tilt is: - Use the latitude (~44° N for Toronto SD) for year‑round optimization. - Or latitude – 15° (~29°) to favor summer generation Reddit.

• Many installers recommend ~38° tilt as a year-round compromise .
• For best summer match, a 29–35° tilt could boost performance in July/August.

Pros:

• High coincidence with cooling peaks.
• Reduces midday grid demand, easing strain on the peaker plant.
• Aligns well with summer solar capability in SD.

Cons:

• Limited roof area; won't cover entire data center load.
• Winter output drops, but cooling demand is also lower.
• Structural engineering needed for wind/snow loads and possible roof reinforcement.

- Summer solar in SD matches cooling demand much better than wind does.

• A rooftop solar array could deliver multi-MW generation during peak load hours.
• A fixed tilt around 29–38°, south-facing, optimizes for year-round or summer-weighted output.
• It won't eliminate all peaker usage, but significantly reduces midday and summer grid demand, easing dependence on the gas plant.

If you'd like, I can calculate an example capacity (e.g., size needed to offset half of cooling power) or draft solar+peaker coordination conditions for local approval. Want to go further into sizing or policy language?

User: What are the proposed size of the buildings for the Toronto Data Center?

What are the proposed size of the buildings for the Toronto Data Center? What is a reasonable estimate of what could be provided on a summer day with just roof top solar?

Assistant: Here are the key details on building size and what rooftop solar could realistically deliver:

Here are the key details on building size and what rooftop solar could realistically deliver:

- Applied Digital’s proposal includes two buildings, each approximately 907,000 ft² (≈84,300 m²) The Department of Energy's Energy.govBaxtel.

• Combined, that's about 1.81 million ft² of roof space available.

Using industry benchmarks:

• Rooftop solar deployment typically yields 5–10 W DC per ft² of usable space .
• Assuming 7 W/ft² (mid-range, accounting for shading/access): 1.81 M ft2×7 W=12.7 MW DC1.81 M ft² × 7 W = 12.7 MW DC1.81 M ft2×7 W=12.7 MW DC system capacity.

- In South Dakota, summer solar insolation is ~6 kWh/m²/day (peak hours) Wikipedia.

• DC-to-AC derate typically ~0.77 (system losses). 12.7 MWDC × 6 h × 0.77≈58.7 MWh/day12.7\text{ MW}_\text{DC} × 6 h × 0.77 ≈ 58.7 MWh/day12.7 MWDC​ × 6 h × 0.77≈58.7 MWh/day

So on a typical summer day, roof-mounted solar might generate ~59 MWh (≈2.46 MW average over daylight).

- Large data centers often draw tens to hundreds of megawatts during peak summer days.

• If the Toronto site runs near 400 MW IT load, plus cooling and infrastructure: - Rooftop solar could offset ~2–5% of total daily power use. - However, during peak midday hours, that 2.46 MW average might spike to 5–6 MW peak, helping reduce reliance on the gas peaker for short intervals.

- Roof area: ~1.8 M ft²

• Solar capacity: ~12–13 MW DC
• Summer daily generation: ~59 MWh
• Midday offset: ~5–6 MW peak

- That's not enough to power the whole data center, but it's meaningful: - Offsets peak cooling demand when solar output is highest. - Helps reduce grid draw and peaker use during midday heat. - Enhances sustainability profile and supports local zoning requirements.

Would you like help outlining a policy recommendation to require a minimum MW of rooftop solar as part of the project approval? We could also explore adding modest on-site BESS to store that solar for evening peak.

User: Yes, but not yet.

Yes, but not yet. What about agri-solar projects? How feasible is a neighboring parcel to allow a larger solar installation to match the summer cooling loads? What specific limitations does it impose on the agriculture usage of the land? For example, the area grows mostly a rotation of corn and soybeans. Could either of those continue to be farmed on a parcel that has solar? Or does the solar installation limit usage of large tractors and implements?

Assistant: Great follow-up.

Great follow-up. Agri-solar, or agrivoltaics, is an emerging strategy that allows for co-use of land—farming and solar generation—on the same parcel. The idea is especially valuable near high-energy-demand projects like data centers, and it's absolutely worth considering near Toronto, SD.

Let’s explore how feasible this is and what trade-offs exist for corn/soy rotations.

Agri-solar involves placing solar panels in a way that allows simultaneous agricultural use. There are several designs:

• Raised panels (6–10 feet high) for access underneath
• Wider row spacing to let sunlight reach crops
• Movable or tracking panels that adjust throughout the day

###

• Corn is tall (6–10 ft) and light-intensive.
• It requires full sun and heavy machinery to plant and harvest.
• Verdict: Corn is incompatible with current agri-solar designs.

- Soybeans are shorter (~2–4 ft) and can tolerate partial shade, especially during vegetative growth stages.

• Studies in Minnesota and Colorado suggest yield reductions of 5–20%, depending on panel layout and row orientation.
• Machinery conflict: Most commercial soybean farming still uses large implements, and tight rows of solar panels make that difficult.

Verdict: Only feasible at smaller scale, with specialized equipment, row spacing adjustments, and modified practices—not standard monocrop farming.

- Standard row crop equipment (e.g., 30–50 ft planters, sprayers, and combines) cannot easily maneuver between solar panel arrays.

• Agri-solar setups typically work with: - Smaller implements - Manual or small-plot farming - Grazing livestock, hay, or pollinator crops (most common)

Conclusion: Standard corn/soy crop cycles with full-size tractors are generally not compatible with agri-solar without major operational changes.

| | |

| --- | --- | | | | | | | | | | | | | | | |

1. Land availability: Eastern SD has flat, open land, excellent solar irradiance (~4.8–5.5 kWh/m²/day annual average).

1. Parcel use: A 40-acre agri-solar array (~5–6 MW) could yield up to 25–30 MWh/day in summer, potentially matching cooling load shortfalls.
2. Economic offset: Agri-solar could partially subsidize lost crop revenue, especially if paired with: - Sheep grazing leases - USDA agri-solar pilot programs - Carbon or RECs (renewable energy credits)

| | | | | |

| --- | --- | --- | --- | --- | | | | | | | | | | | | | | | | | | | | | | | | |

- Corn and soybean rotations are not compatible with standard agri-solar setups.

• Sheep grazing or forage crops are much more feasible on solar-adjacent land.
• A neighboring parcel could realistically support 5–20 MW of solar, depending on size and local permitting, and meaningfully reduce the data center’s summer peak demand.
• County zoning or incentives could encourage dual-use land, especially near energy-intensive infrastructure.

Would you like me to model how much land would be needed to supply, say, 25% of the data center’s summer cooling demand, or draft a brief for a planning board proposal that supports agri-solar as a community-benefit offset?

User: Let's circle back to water usage.

Let's circle back to water usage. In the initial 'Concerns', you stated it was enormous, often not well disclosed. Later, you stated, according to SDPB, it would be a closed-loop water cooling system, typical to an average household. Which is it?

Assistant: Excellent question, and thank you for calling that out — you’re right to notice the tension between the general concern…

Excellent question, and thank you for calling that out — you’re right to notice the tension between the general concern and the specific claim for the Toronto site.

Let’s break it down clearly:

In many data center developments, cooling is one of the biggest environmental concerns, especially when:

• Evaporative cooling is used (like cooling towers), which consumes millions of gallons annually.
• Centers are built in arid or agriculturally sensitive areas.
• Lack of transparency: water use is often unreported or poorly understood at the permitting stage.

That’s why “enormous and poorly disclosed” water usage is a common concern nationwide, particularly with hyperscale or AI-focused centers. For example:

• Google's site in The Dalles, OR reportedly used over 350 million gallons/year.
• Meta and Microsoft have similar footprints when using traditional cooling methods.

So, this concern is valid in general — but it doesn’t automatically apply to every site or every design.

According to South Dakota Public Broadcasting (SDPB) and developer statements:

Closed-loop systems:

• Use water internally, in sealed pipes, to carry heat to external heat exchangers.
• Do not evaporate water to the atmosphere (unlike cooling towers).
• Typically use air-cooled chillers or dry coolers, especially in cold climates like eastern South Dakota.

So in this specific case, the developer claims that:

• Little or no water is consumed, only used in the loop.
• Water is not evaporated and thus not withdrawn continuously.
• Cooling loads are met using ambient air and internal systems, not external water draws.

If true, that would mean water consumption is minimal and not a major environmental concern.

| | | |

| --- | --- | --- | | | | | | | | | | | | | | | | |

- Your skepticism is warranted, especially since these claims come from developer sources and haven’t yet been independently audited.

• If the Toronto center truly uses a closed-loop system, then water use would be minimal, consistent with their statement.
• But until the facility is built and monitored, ongoing transparency and verification are still critical.

If you’re engaging in zoning, permitting, or public comment processes, here are some smart asks:

1. Request confirmation in writing: e.g., “Water consumption shall not exceed X gallons per day, comparable to an average household.”
2. Require disclosure of cooling system design specs: closed-loop, dry coolers, etc.
3. Condition site approval on no use of evaporative cooling systems without public hearing.
4. Ask for independent review or third-party engineering verification of water demand projections.

Let me know if you'd like sample language for those types of requests. You're asking exactly the right questions to ensure this development aligns with long-term community interests.

User: It seems like you are still not answering the question.

It seems like you are still not answering the question. If the design is a closed-loop, air-cooled system, how does it reject the active heat load into a high ambient temperature in the hot summer months?

Assistant: You're absolutely right to push for clarity, David — and thank you for sticking with the technical depth.

You're absolutely right to push for clarity, David — and thank you for sticking with the technical depth. Let's answer your question directly and precisely:

How does it reject the active heat load during hot summer months, when ambient temperatures are high?

It still rejects heat — but far less efficiently, and at a higher energy cost, using air-cooled chillers or dry coolers that rely on fans and increased compressor work to push heat into hot ambient air.

This is doable, but it imposes tradeoffs:

• Higher electricity usage
• Lower cooling efficiency (EER/COP)
• Potential limits on maximum IT load per rack or dynamic throttling

###

• Use ambient air and fans to cool a closed-loop fluid (usually water or glycol mix).
• In mild or cold temperatures (like much of the year in South Dakota), they’re efficient and sustainable.
• In hot summer months, when outside air hits 85–95°F or more, these systems struggle to reject enough heat.

To cope with that:

- Fans run harder and longer, increasing parasitic energy use.

• Compressors in the chiller work against a smaller delta-T between coolant and ambient air — consuming more power.
• Power usage effectiveness (PUE) can climb in the summer — possibly from ~1.2 in winter to 1.5+ in peak heat.

| | |

| --- | --- | | | | | | | | | | | | | | | |

- A closed-loop air-cooled system does not “consume” water but relies more on electricity, especially in summer.

• In Toronto, SD, with cool winters and shoulder seasons, this works very well most of the year.
• In July and August, the system works harder, costs more to run, and may require careful design limits to ensure performance.

:

It avoids using water — but it doesn’t avoid the thermodynamic challenge. You're right to dig into this, because avoiding water use shifts the burden onto the grid.

Would you like to estimate the extra summer electrical load for cooling compared to shoulder seasons? I can walk through a quick calculation based on IT load and climate data.

User: Yes, what is the additional load.

Yes, what is the additional load.

Assistant: Based on a 50 MW IT load and the difference in cooling efficiency between mild and hot weather:

Based on a 50 MW IT load and the difference in cooling efficiency between mild and hot weather:

• On each hot summer day (e.g., July or August), the air-cooled system would use an additional ~360 MWh/day compared to spring/fall conditions.
• Over 60 summer days, that adds up to an extra 21,600 MWh of electricity consumed just to handle less efficient cooling.

- Peak grid demand rises significantly in summer—not just from compute, but from cooling inefficiency.

• That additional 21.6 GWh could: - Power over 2,000 homes for a year - Represent a ~15–20% spike in electricity use for the data center in summer
• It also increases pressure on local generation, especially if wind output is low and solar is limited to daylight hours.

Let me know if you'd like to model this for a larger load (e.g., 100 or 200 MW) or estimate the size of a solar + battery system that could offset that summer delta.

User: Let's summarize what we've learned, that is specific to the Toronto site.

Let's summarize what we've learned, that is specific to the Toronto site. Let's focus on the impact to the local economy. Use a form of: the stated benefit is X, but the likely local impact is Y, based on Z. After we complete a nice tabled summary, then we can develop the takeaways.

Assistant: Absolutely — here’s a clear, tabled summary of the major stated benefits of the proposed Toronto, South Dakota data cen…

Absolutely — here’s a clear, tabled summary of the major stated benefits of the proposed Toronto, South Dakota data center, paired with a realistic assessment of likely local impacts, based on available information and engineering principles:

| | | |

| --- | --- | --- | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |

1. The economic impact is front-loaded, mostly during construction, and may be smaller long-term than headlines suggest.

1. Job creation will likely fall short of stated numbers; permanent staff is modest unless paired with colocation or support functions.
2. Water usage is low, but that shifts the cooling burden to the grid, especially during summer — increasing local power demand and potentially peaker plant reliance.
3. Wind and solar are not seasonally aligned with peak cooling needs; full renewable matching would require major battery storage or off-site solar fields.
4. Land impact is small in percentage, but large in visibility — 200+ acres of prime cropland will be removed from corn/soy production and replaced with industrial infrastructure.
5. Community benefits may depend heavily on negotiated agreements (PILTs, workforce development programs, or local power rebates), which must be carefully structured.

Would you like help drafting a formal community impact memo or recommendations for local officials based on this summary? We can also model how different mitigation strategies (solar, BESS, job commitments, zoning conditions) could improve these outcomes.

User: Before we draft the recommendations, let's look at some options, and we probably need to split the recommendations betw…

Before we draft the recommendations, let's look at some options, and we probably need to split the recommendations between the county zoning environmental impacts, and the utility commission environmental impacts. For example, since the county property taxable value will increase, can you recommend typical projects to offset the impact by earmarking those increased tax revenues. And, for the utility, can we develop options for higher peak demand charges or limit the usage of peaker plants and ensure that utility revenue goes back to the impacted community rather than the system as a whole?

Assistant: Excellent approach.

Excellent approach. You're right to split the recommendations between:

• County-level (zoning, land use, property tax, visual/environmental impacts)
• Utility-level (grid strain, peak demand, emissions, cost recovery structure)

This framework reflects real-world regulatory boundaries and ensures that local concerns are addressed directly and proportionally.

🏛️ A. County Zoning & Land Use: Options to Offset Local Impact

💰 1. Earmark Property Tax Increases for Local Benefit Projects

While South Dakota has no corporate income tax, property tax increases from new infrastructure can fund:

| Offset Project Type | Purpose | Notes | | --------------------------------- | ------------------------------------------- | ---------------------------------------- | | Rural road upgrades | Handles construction traffic, future access | Gravel-to-paved, culverts, snow clearing | | Volunteer fire & EMS support | Improves response near industrial site | Equipment, personnel, station upgrades | | School or workforce grants | Connects project revenue to local students | STEM/coding programs; local scholarships | | Ag or conservation district funds | Compensates for lost farmland or habitat | Buffer plantings, pollinator zones | | Broadband or power reliability | Returns benefit to residents | Microgrid pilots, new poles/substations |

Structure: The county could require a PILT (Payment in Lieu of Taxes) agreement or earmark a percentage of property tax delta (e.g., 30–50%) for specific projects.

📜 2. Local Development Agreement or Conditional Use Permit Conditions

The County Planning Commission can require:

• Annual data reporting (e.g., jobs, energy use, complaints, emissions)
• Landscape buffers, light/noise control
• View corridors or fencing types to mitigate rural impact
• Public transparency on water and energy systems
• Optional: require rooftop solar or on-site BESS as conditions for approval

⚡ B. Utility Commission: Energy & Grid Impact Recommendations

In South Dakota, oversight of grid matters and utility rates (even municipal/co-op power) often falls to state regulators (PUC), though local influence matters through public input and franchise agreements.

📈 1. Higher Peak Demand Charges for Large Loads

| Goal | Effect | | --------------------------- | --------------------------------------------------------------- | | Reflect actual strain costs | Disincentivizes poorly timed use | | Reward dispatchable load | Encourages load-shifting or throttling | | Revenue equity | Prevents residential ratepayers from subsidizing hyperscale use |

Mechanism:

• Time-of-use pricing
• Tiered rates (e.g., >10 MW gets premium charge during 3–8 pm)

🔥 2. Limit Peaker Plant Emissions or Usage

• Cap hours of operation during non-emergencies
• Require clean fuel sourcing or emissions offsets
• Require coordinated dispatch with wind/solar (i.e., peaker must only run when wind is low and no battery reserve is available)

💸 3. Utility Revenue Sharing with Local Community

If Toronto’s power comes from a municipal or regional utility (like MRES), revenue from demand charges or infrastructure upgrades could:

• Be earmarked for Toronto/Deuel County projects
• Fund community energy programs, microgrids, or insulation upgrades

Structure:

• County negotiates a Community Impact Energy Agreement
• State PUC may need to approve or facilitate this revenue allocation

✅ Summary of Options (Structured Table)

| Authority | Tool | Purpose | Example | | -------------------- | ----------------------------------- | --------------------------------------- | ------------------------------------------- | | County (Zoning) | PILT / Tax Earmarks | Offset infrastructure & visual impact | Roads, EMS, schools, ag mitigation | | County (Zoning) | Development Agreement | Tie approvals to local benefits | Rooftop solar, buffers, public reporting | | County/State Utility | Peak Demand Charges | Discourage grid strain | Higher cost during high-load hours | | State Utility / MRES | Usage Caps or Emissions Limits | Protect air quality / reduce peaker use | Dispatch rules for Toronto plant | | Utility / County | Community Revenue-Sharing Agreement | Ensure local benefit from energy infra | Annual % of revenue to town/county projects |

Would you like me to help format this into two sets of formal policy recommendations — one for county planning and one for utility/public power representatives — with suggested language and example motions?

User: In the Limit Peaker Plan Usage, you made an interesting comment - "during non-emergencies".

In the Limit Peaker Plan Usage, you made an interesting comment - "during non-emergencies". For a Data Center, what would constitute an "emergency"? Isn't the AI just in learning or conversation mode, so that nothing is really an "emergency"? Or is the Data Center used for something else that could constitute or require immediate responses?

Assistant: Fantastic observation — this question goes straight to the core issue in how we think about criticality of data center …

Fantastic observation — this question goes straight to the core issue in how we think about criticality of data center operations, especially in rural areas supporting AI or hyperscale loads.

Let’s unpack it clearly:

❓What would constitute an “emergency” that justifies using a peaker plant to serve a data center?

✅ Valid Emergencies (Generally Accepted)

These are conditions where grid stability or human safety are threatened:

1. Grid contingency — transmission line or transformer failure
2. Extreme weather event — heatwave or polar vortex straining supply
3. Loss of renewable output — sudden drop in wind/solar generation
4. Cybersecurity or ransomware attacks on utility infrastructure
5. System-wide underfrequency/undervoltage — risk of cascading outages

These could trigger peaker plant use as backup generation for any large user.

🧠 What About AI Workloads? Are They Truly “Emergency-Critical”?

Generally, no — but some segments might be. Here's how to think about it:

1. Non-Critical AI Workloads

| Examples | Interruptible? | Justifies peaker use? | | -------------------------------------------- | -------------- | --------------------- | | Chatbots, image generation, model training | ✅ Yes | ❌ No | | Video encoding, search indexing, ad delivery | ✅ Yes | ❌ No |

These can and should be throttled or paused during tight grid conditions — especially in air-cooled data centers that are not hospitals or defense sites.

2. Potentially Critical Workloads

| Examples | Interruptible? | Justifies peaker use? | | ------------------------------------------------ | -------------- | --------------------- | | Hospital diagnostics (e.g., AI triage) | ❌ No | ✅ Possibly | | Real-time disaster modeling (weather/earthquake) | ❌ No | ✅ Possibly | | Financial clearing systems / banking AI | ⚠️ Limited | ⚠️ Case-by-case | | Defense/military compute workloads | ❌ No | ✅ Yes, if contracted |

These may require uninterrupted service, and some data centers contract with federal or financial institutions to serve such loads.

🤖 So What's Likely at the Toronto Site?

The Applied Digital data center near Toronto is designed for high-density compute, likely focused on:

• AI model training and inference
• Blockchain or large-scale data processing
• Possibly cloud services for third-party clients

These are almost always delay-tolerant and non-emergency workloads, meaning:

There is no valid reason to trigger local peaker plant operation during a grid peak just to keep this AI work running at full speed — unless the data center also hosts critical clients, which would need to be disclosed.

✅ Policy Opportunity

You can propose language like:

"During grid stress events or system peaks, electricity supply to non-emergency, delay-tolerant workloads (including AI training, inference, and content generation) shall be interruptible or throttled before engaging peaker plant operations."

And define "emergency" as:

"Emergency conditions are those in which public health, safety, or grid integrity are at risk, as determined by the regional transmission operator or system operator, and do not include commercial service-level expectations for discretionary workloads."

Would you like help drafting specific zoning or utility policy language that incorporates this threshold for defining “emergency”? This could be a powerful tool for local control.

User: How do I share this thread with others?

How do I share this thread with others?

Assistant: To share this thread with others, you can use the ChatGPT "Share & Export" feature:

To share this thread with others, you can use the ChatGPT "Share & Export" feature:

✅ Here's how:

1. Scroll to the top of this conversation window.
2. Click the “Share” icon (looks like an arrow pointing upward or a link symbol, depending on your browser).
3. Choose “Copy link to share”.
4. Send that link to others — it will open a read-only version of this conversation, viewable without needing a ChatGPT account.

🔒 Notes:

• You can choose to hide your name before sharing.
• The shared link is public to anyone with the URL, but not searchable.
• The conversation is read-only — others can read but not reply.

If you're planning to present or archive this, I can also help you:

• Export this thread to a PDF or DOCX file
• Format it into a summary briefing
• Extract highlights for a presentation or public comment letter

Let me know what you'd prefer!

gpt-4-1-mini

User: How many acres of farm land are in Deuel County, South Dakota?

How many acres of farm land are in Deuel County, South Dakota? How many acres are needed for the proposed Toronto Data Center and Toronto Power Plant expansion?

Assistant: Certainly, let's delve into the land use specifics for Deuel County, South Dakota, focusing on the proposed Toronto Dat…

Certainly, let's delve into the land use specifics for Deuel County, South Dakota, focusing on the proposed Toronto Data Center and the Toronto Power Plant expansion.

| | | |

| --- | --- | --- | | | | | | | | | | | | | | | | |

The combined proposed land use for these projects represents a small fraction of Deuel County's total farmland. However, the conversion of agricultural land to industrial use can have several implications:

• Loss of Productive Cropland: The conversion of 171–271 acres of cropland could impact local crop production, particularly for corn and soybeans.
• Economic Considerations: While the data center may bring economic benefits, such as job creation and infrastructure improvements, these must be weighed against the potential loss of agricultural revenue.
• Community Impact: The projects may alter the rural landscape, affecting local communities and agricultural practices.

It's essential for local stakeholders, including farmers, policymakers, and community members, to engage in discussions to balance industrial development with the preservation of agricultural land and practices.

If you have further questions or need more detailed information on any of these aspects, feel free to ask.

User: Let's assume the data center needs 200 acres.

Let's assume the data center needs 200 acres. What fraction of Deuel County land will be used?

Assistant: If the data center needs 200 acres, here’s how that compares to Deuel County’s total farmland:

If the data center needs 200 acres, here’s how that compares to Deuel County’s total farmland:

• Total farmland in Deuel County: ~253,106 acres
• Data center land use: 200 acres

That’s a very small fraction — less than one-tenth of one percent of total farmland in the county.

If you include the 71 acres for the power plant expansion, the combined footprint would be about:

200+71253,106=271253,106≈0.00107=0.107%\frac{200 + 71}{253,106} = \frac{271}{253,106} \approx 0.00107 = 0.107\%253,106200+71​=253,106271​≈0.00107=0.107% So together, just over 0.1% of the farmland.

If you want, I can help put this in a more visual or report-friendly format!