A Revised Colorado Energy Plan 2040
With the recent changes, what a sensible plan will entail
I’m guessing the Colorado Energy Office and it’s various consultants and 3rd parties put in thousands of hours coming up with the Colorado Clean by 2040 and Colorado Greenhouse Gas Pollution Reduction Roadmap 2.0. And in a democracy that process is important.1
Until recently if a blogger or other individual wanted to rework those plans based on substantial recent changes, you’d be out of luck. You could postulate a thing or two but coming up with a competing report would be very expensive.
No more. This is a place where A.I. shines. As doing this will be new to a lot of you, I’m going to walk through the process I followed so you can understand how this works and see the results from each step.
And yes A.I. can be wrong. But so can a committee that invests thousands of hours. You do want to look at this with a critical eye. But I’m willing to bet Will Toor a quarter2 that if they rework the Colorado Energy Plan, it will come out the same as the plans generated by the A.I.
Step 1 - The Prompt
Using A.I. takes careful thought. The critical one is the prompt you write. You are giving the A.I. guidance with your question. You want to keep this as loose as possible so you don’t restrict it from telling you about areas you had not considered, but can be of interest. At the same time too loose and you get something that is not targeted enough.
I started with one question:
You are an expert on the power grid.Your goal is to write a report that specifies in a moderate level of detail what the state of Colorado should use for power generation and transmission lines going forward. Write this as a new Colorado Energy Plan.
The goal of the state of Colorado is to have power that is reliable, abundant, inexpensive, and has minimal carbon emissions.
Make the following assumptions:
There will be no federal support (grants, loans, subsidies, tax credits, etc.) for wind, solar, or batteries.
The State of Colorado will add nuclear as an approved green energy solution.
Do not assume any future technology will become available.
There are presently no SMRs shipping. Therefore if nuclear is a part of the plan, that means using either the Westinghouse AP1000 or the Korean APR14003
Where possible make decisions backed by logical reasoning rather than listing options. Only provide options where there is no clear choice. Explain why you come to each decision, providing citations where appropriate.
Where there are multiple options, this report should leave the reader with an understanding of the costs and trade-offs of each approach.
Perform deep research as needed. Take your time as needed.
Write the report for an audience that has a college degree, but no specialized knowledge of the electrical grid. Maintain the highest standards of accuracy and objectivity.
Your statements must match reality.
And I then fed it to the following A.I.s in order:
First issue was that Gemini just said build nuclear power without specifying what designs. That’s ok, but while that ambiguity leaves the question of SMRs out, it also doesn’t get specific about how much to build.
Perplexity then specified SMRs. So saying “Do not assume any future technology will become available“ was not sufficient to preclude recommending SMRs. Qwen also recommended SMRs. I did not re-run the question on Perplexity4 but I did add assumption 4 to re-run Qwen and used it on DeepSeek too.
The New Improved Colorado Energy Plan 2040
All 4 A.I.s are fundamentally in agreement. Read the documents for details, which includes their citing references as appropriate - here are the conclusions.
Gemini
This Colorado Energy Plan provides a roadmap for a future where Colorado has access to reliable, abundant, inexpensive, and low-carbon electricity. By prioritizing nuclear power, strategically using natural gas, and maintaining existing hydropower resources, while acknowledging the economic limitations of unsubsidized wind and solar, this plan ensures a stable and affordable energy supply for all Coloradans. Investing in transmission infrastructure and demand-side management further enhances the efficiency and reliability of the grid. Finally, supporting research and development ensures that Colorado remains at the forefront of energy innovation. This plan represents a pragmatic and responsible approach to meeting Colorado's energy needs for generations to come.
Perplexity (using SMRs)
This plan achieves 65% emissions reduction by 2035 through nuclear baseload (35%), wind (28%), and limited gas (22%). Transmission upgrades ($4.5B) and SMRs ($8.4B) dominate costs but provide long-term savings versus renewable overbuild strategies. Immediate actions:
Pass HB25-10405 to enable nuclear financing
Fast-track Eastern Plains reconductoring under SB23-016
Launch workforce retraining at CSU-Pueblo for SMR technicians
Colorado’s energy future hinges on marrying nuclear’s reliability with wind’s cost leadership—a blueprint replicable across Western states.
Qwen
Colorado’s path forward hinges on leveraging nuclear energy as the primary baseload provider, supplemented by carefully integrated renewables and retained natural gas assets during the transition period. Simultaneously, modernizing transmission infrastructure ensures seamless integration and enhanced reliability. By adhering to this plan, Colorado can achieve its goals of reliable, abundant, inexpensive, and low-carbon power without relying on uncertain federal assistance or unproven technologies.
DeepSeek
Colorado’s optimal path combines nuclear baseload power, strategic gas retrofits, and targeted renewables, supported by expanded transmission and microgrids. While costs rise initially (~$20–30/month by 2030), this plan ensures long-term affordability and reliability while cutting emissions 75% by 2040. Immediate action on nuclear permitting and transmission permitting is critical to avoid delays.
The big difference between the four is some give wind a larger role while others minimize the role of renewables. All four unequivocally see nuclear as the baseload provider.
Sticking with the Existing Plan
I then followed up asking each A.I. the following:
Write an addendum to the report you just created. In the addendum discuss what will occur if Colorado continues on its present path to have 95% of electricity generated by renewables, with no nuclear, by 2040.
Speak to the cost and reliability impacts. Make the same assumptions as in the just created report.
Perform deep research as needed. Take your time as needed.
Write the report for an audience that has a college degree, but no specialized knowledge of the electrical grid. Maintain the highest standards of accuracy and objectivity.
Your statements must match reality.
What was super interesting in the response in these addendums is they first answered the question, then strongly suggested going with nuclear. This shows the value in leaving your prompt open enough, the A.I. will tell you if your plan is stupid.
I will call out two conclusions from the reports. Please do read all four reports in full to get a more complete picture of the clusterfuck6 our present plan would entail.
Gemini
Maintaining grid reliability requires constant balancing of electricity supply and demand. With a 95% renewable portfolio, this becomes extremely difficult.
Blackouts: Extended periods of low wind or solar output, especially during peak demand periods, could lead to widespread blackouts. Even with accurate weather forecasting, predicting the exact output of renewables is challenging. A sudden drop in generation could leave the grid vulnerable.
Frequency Instability: The grid operates at a specific frequency (60 Hz in the US). Fluctuations in generation can cause frequency instability, which can damage equipment and lead to blackouts. Maintaining frequency stability becomes much more complex with a high penetration of intermittent renewables.
Voltage Instability: Similar to frequency, voltage must also be kept within a narrow range. Intermittent generation can cause voltage fluctuations, impacting the performance of electrical devices and potentially leading to damage.
Qwen
Pursuing a 95% renewable electricity grid by 2040 without nuclear power presents significant challenges related to reliability, cost, and practical implementation. While wind and solar are essential components of a low-carbon future, their intermittency and storage limitations make them unsuitable as standalone solutions for baseload power. Without federal subsidies or advanced technologies, the financial burden of overbuilding renewables and deploying massive energy storage systems would fall squarely on Colorado taxpayers and ratepayers. Additionally, the environmental and social trade-offs associated with large-scale renewable development could undermine public support for the transition.
Why Two Nuclear Plants
All four of the A.I.s suggested starting with two nuclear plants. So I then asked them (yes I know the baseload is greater than 2.8GW - wanted to check if the A.I. knew):
What is Colorado's baseload power? If it is greater than 2.8GW why do you suggest just 2 nuclear plants?
All four had similar answers so here’s one of them. If you read the reply in full you will see it is taking into account political and economic considerations.
Qwen
The recommendation of two nuclear plants is a pragmatic starting point that balances Colorado’s immediate baseload needs with practical considerations such as cost, construction timelines, and grid integration. These plants would provide a reliable, low-carbon foundation for the state’s energy system, allowing for a gradual phase-out of fossil fuels and the integration of renewables. Over time, additional nuclear capacity or complementary technologies could be added to fully meet baseload demands.
In Conclusion
Using A.I.
I hope you took the time to read at least one of the full reports and see how the A.I. writes really well thought out and accurate researched reports. This is a game changer for governmental reports.7 And as more people see the quality of the work produced, we’ll see the government using this more and more.
Even where they need buy-off from all the interest groups, they can start with the A.I. generated report. And that will lead to major discussion for the prompts fed to the A.I. We may get to the point where all the discussion is over the prompt, it’s run, and then we use that report. Interesting times…
And whenever I write something like this I am asked which A.I. is best so people can use just one. The models are improving so fast, the “best” keeps changing. And it also depends on what you use it for. For programming I use Cursor and Copilot. For this kind of work I use the above four. 8
Perplexity is generally among the best and it can run both its own model and DeepSeek. So if you need to limit yourself to one as they are expensive, get Perplexity and run a question using Perplexity’s model, from the answer improve your prompt. The run that new prompt on the DeepSeek model on Perplexity. Keep ping ponging back and forth until you have your prompt optimized.
A Sensible Energy Plan
It’s clear that Colorado needs to rethink its Energy Plan 2040. Going with the present “renewables for 95% of our power” is a clear recipe for increased prices, lower reliability, blackouts, and a clear message to industry to locate elsewhere.
It’s not just that we’ll pay more for less power, the present plan also delivers fewer jobs at lower pay scales. Building and running a nuclear plant has much better jobs than installing wind & solar. Bringing manufacturing & datacenters to Colorado means lots of quality well paying jobs.
In short, the Colorado Energy Plan needs to be reworked taking reality into account.9
This means the Colorado Energy Plan needs to change from “any green energy” to “nuclear for full baseload and everything else, including gas, for peak load.” I was surprised that wind was listed as a very useful part of the peak generation. But the arguments for it in the reports are compelling so yes, wind has a role.
As Colorado’s baseload is roughly 10 - 15GW and we’re looking at a 30% increase10 in base over the next 10 years which means build 13 - 20GW of nuclear. That’s 14 APR1400 plants.
Following the unanimous conclusion in the four reports to start with 2 plants, then if 18 months in they are on time and under budget, start 4 more. And then 18 months further in, if the first 6 are on time and under budget start the last 8. That could have us at 14 plants in 10 years.
Read The Power Broker
The most I ever bet.
This item was added for Qwen & DeepSeek.
So you can see the result without the specific exclusion.
This shows how up to date the A.I.s are.
This description understates how bad it will be.
Business too.
I also have Copilot via Microsoft Office. But I found it tended to match Perplexity and in the past it did not have recent data (no longer a problem) so I settled first with Gemini & Perplexity. Then as DeepSeek and Qwen came along, I tried each and they had different takes on the questions I first tried them on. So sticking with them too.
Ella Rose has a great quote - “The Colorado Energy Plan is composed of wind + solar + pixie dust. And we can’t find the pixie dust.”
My guess is more than 50% between electrify everything and A.I. usage.
I wondered what it would look like with if the prompt didn't include an assumption there will be nuclear power and that it wouldn't be subsidised either.
https://www.perplexity.ai/search/you-are-an-expert-on-the-power-PaJjX_R1TBC9szayKeJlcQ
Thank you, I was unsurprised by the analysis presented, but completely taken aback by the 'competency' and succinctness of the AI output. It really is a game changer.