Strategic Development Feb 10 2026

Unknown: Thanks, guys.

Unknown: See you next time.

Unknown: relating those down here

Unknown: So, here we are

Unknown: Chancellor

Unknown: Five seconds please, five seconds.

Unknown: Good evening and welcome to the Strategic Development Committee and Special Board Meeting

SPEAKER_09: of February 10, 2026.

SPEAKER_09: This room is equipped with a safety alarm.

SPEAKER_09: If the alarm sounds, please leave in an orderly manner via the exits to the lobby or behind

SPEAKER_09: the dais.

SPEAKER_09: Assemble in front of the building and wait to hear the all clear announcement from security

SPEAKER_09: before reentering.

SPEAKER_09: This meeting is being recorded and can be accessed on SMUD's website.

SPEAKER_09: Please remember to unmute your microphone when speaking in order that our virtual attendees

SPEAKER_09: may hear.

SPEAKER_09: The microphone will display a green indicator light when the mic is on.

SPEAKER_09: For members of the public attending in person that wish to speak at this meeting, please

SPEAKER_09: fill out a speaker's request form located on the table outside the room and hand it

SPEAKER_09: to SMUD's security.

SPEAKER_09: Members of the public attending this meeting virtually that wish to provide verbal comments

SPEAKER_09: during the committee meeting may do so by using the raise hand feature in Zoom or pressing

SPEAKER_09: star 9 while dialed into the telephone toll-free number at the time the public comment is

SPEAKER_09: called.

SPEAKER_09: Technical support staff will enable the audio for you when your name is announced during

SPEAKER_09: the public comment period.

SPEAKER_09: You may also submit written comments by emailing them to publiccommentatsmud.org.

SPEAKER_09: Written comments will not be read into the record but will be provided to the board electronically

SPEAKER_09: and placed into the record of the meeting if received within two hours after the meeting

SPEAKER_09: ends.

SPEAKER_09: Chief Legal Officer, please conduct the roll call.

SPEAKER_01: Chair Herbert.

SPEAKER_01: I apologize.

SPEAKER_01: Director Herbert?

SPEAKER_01: Here.

SPEAKER_01: Director Fishman?

SPEAKER_01: Here.

Unknown: Chair Samborn?

SPEAKER_01: Here.

SPEAKER_01: All committee members are present.

SPEAKER_01: Also present are directors Louie Thompson, Kerth, and President DeMaio.

SPEAKER_09: Thank you so much.

SPEAKER_09: Tonight we have on our agenda the board is going to have an internal and external presentations

SPEAKER_09: on large load grid integration and artificial intelligence.

SPEAKER_09: And we are very excited.

SPEAKER_09: We have our staff presenter is James Frazier, Director of Research and Development Grants

SPEAKER_09: and Partnerships.

Unknown: And we are going to have Ilya, I hope I say this correctly, Terna Kowski, the Senior Researcher

SPEAKER_09: and Group Manager of Grid Planning and Analysis Center at NLR, the National Laboratory of

SPEAKER_09: the Rockies here to present this item.

SPEAKER_09: So go ahead and turn it over to you, James.

Unknown: Thank you.

Unknown: Excellent.

SPEAKER_00: Thank you for the introduction and appreciate the opportunity to speak to the board and

SPEAKER_00: the community tonight.

SPEAKER_00: Before I jump into the presentation, I want to talk just a little briefly about the agenda

SPEAKER_00: and what we're going to cover.

SPEAKER_00: Data centers, large load, AI, these are evolving conversations and rapidly evolving.

SPEAKER_00: The speed of technology in this space is mind boggling.

SPEAKER_00: And that's coming from someone who has spent a majority of their career in research and

SPEAKER_00: development endeavors at the leading edge of technology.

SPEAKER_00: That GPT was initially released at the very end of 2022.

SPEAKER_00: And now when we have conversations around technology, you can't help but have AI data

SPEAKER_00: centers and the growing load from those services be part of all of the conversations that we

SPEAKER_00: see both in the utility base and also in the community on a larger basis.

SPEAKER_00: Given the scope and scale of that dialogue, I just want to call out that tonight's conversation

SPEAKER_00: is really an opening dialogue around what is a very large conversation.

SPEAKER_00: If we don't cover all of the topics or all of the questions that folks are interested

SPEAKER_00: in, don't worry.

SPEAKER_00: There's going to be more opportunities to build this dialogue.

SPEAKER_00: We want to start with the foundational information.

SPEAKER_00: We want to build an understanding.

SPEAKER_00: We've invited our guests here from the National Lab of the Rockies to speak to some of the

SPEAKER_00: national landscape and technologies around this conversation.

SPEAKER_00: We're really hoping to build this dialogue forward.

SPEAKER_00: So to the extent that there's feedback, questions, other topics that we want to cover, we're

SPEAKER_00: very much listening and excited to keep diving into those.

SPEAKER_00: One thing that we can say for certain is that a year from now, two years from now, this

SPEAKER_00: conversation will be very different than it is today and this technology will have moved

SPEAKER_00: a significant distance forward.

SPEAKER_00: We're excited to be part of that.

SPEAKER_00: When we talk about data centers and we talk about Sacramento, it's a natural question

SPEAKER_00: of are data centers actually considering the area?

SPEAKER_00: If so, why Sacramento?

SPEAKER_00: Before we dive into what that conversation looks like on how many folks may be looking

SPEAKER_00: at Sacramento, I wanted to dive in and talk just a little bit about why those data centers

SPEAKER_00: are looking here.

SPEAKER_00: It's important that we start by understanding that not all data centers are the same.

SPEAKER_00: When we go back five, six, seven, eight years, a lot of the data centers that we see are

SPEAKER_00: focused on hosting websites or particular data servers or information exchanges.

SPEAKER_00: That's not the type of data center that we're talking about when we're talking about artificial

SPEAKER_00: intelligence.

SPEAKER_00: With artificial intelligence, there's really two types of data centers.

SPEAKER_00: On one side we have a training data center and on another side we have an inference data

SPEAKER_00: center.

SPEAKER_00: I want to bring that down a little bit to an analogy to kind of simplify that for folks.

SPEAKER_00: The training center is where the AI models are made.

SPEAKER_00: Think about that in the movie industry like Hollywood.

SPEAKER_00: It's where the actual production happens.

SPEAKER_00: It's where the animation happens.

SPEAKER_00: It's where you shoot the scene and then shoot it again and keep tuning it until you get

SPEAKER_00: it just right.

SPEAKER_00: At the end of that effort in that Hollywood studio, you're able to put that film in a

SPEAKER_00: can and then you distribute that can to movie theaters all around the country, all around

SPEAKER_00: the world and they unpack that and then they utilize the film on that can to actually show

SPEAKER_00: that movie to the end user.

SPEAKER_00: That user experience and that interaction that they have at the edge of the grid or at

SPEAKER_00: those local movie theaters, that's the inference.

SPEAKER_00: Our training data centers are massive complexes.

SPEAKER_00: They're taking in huge amounts of information from all over the world, from all of the data

SPEAKER_00: sources that they have access to.

SPEAKER_00: They're processing that.

SPEAKER_00: It's a huge computational resource.

SPEAKER_00: These facilities are often hundreds of megawatts and consume energy nearly 24-7, 365.

SPEAKER_00: Very steady load consumption.

SPEAKER_00: What you'll notice is that there's no user interaction.

SPEAKER_00: The output of that effort is the AI model.

SPEAKER_00: It is the chat GPT tool that is actually used by other data centers when you have that information

SPEAKER_00: exchange.

SPEAKER_00: On the inference side, you'll see that it's now a two-way data communication.

SPEAKER_00: It is a user's computer interacting with a data center to submit a question that a

SPEAKER_00: GPT engine or a large language model is going to provide a response to.

SPEAKER_00: It's a call center where a chat bot picks up the phone on the other side and you have

SPEAKER_00: a conversation.

SPEAKER_00: These are often high speed end use cases where latency matters.

SPEAKER_00: They're not nearly as large as the training centers and the load shapes at these sites

SPEAKER_00: are not nearly as steady.

SPEAKER_00: When folks are interacting with AI models, it has a lot to do with when they're at work

SPEAKER_00: and in the office, when they're at home on their home computer.

SPEAKER_00: Not so much 24-7, 365 overnight.

SPEAKER_00: Why is this difference important?

SPEAKER_00: It's because Sacramento, when we're thinking about data centers, is well suited for those

SPEAKER_00: inference data centers.

SPEAKER_00: We're very strategically located in Northern California with good access to the broader

SPEAKER_00: technology area for low latency communication.

SPEAKER_00: When you have a communication with a chat bot and you ask a question, it's not half

SPEAKER_00: a second before that response comes back which creates a very awkward type of conversation.

SPEAKER_00: It's near instantaneous.

SPEAKER_00: It creates the human type interface that folks are looking for when they're engaging with

SPEAKER_00: AI tools.

SPEAKER_00: We have more land availability than several areas of California, especially highly built

SPEAKER_00: out areas like the Bay Area, for example.

SPEAKER_00: Our low cost of energy is a major driver for this opportunity.

SPEAKER_00: I like to call that out because maintaining that low cost of energy is also a key theme

SPEAKER_00: for ensuring that this becomes a long-term destination for those data centers.

SPEAKER_00: When we think about what that looks like, it often boils down to what is the actual

SPEAKER_00: interest that we're seeing?

SPEAKER_00: How many inquiries are coming into the region and what is this conversation looking like?

SPEAKER_00: The thing there is that it turns out it's not as simple to answer that question of how

SPEAKER_00: many people called and asked if they could build a data center.

SPEAKER_00: How many megawatts are we looking at?

SPEAKER_00: Because there's different scales of those inquiries.

SPEAKER_00: At the national level, if we think about all the inquiries that are coming into our regional

SPEAKER_00: partners, the local governments or the economic development councils that represent the region,

SPEAKER_00: that number might be over 2,000 megawatts, a massive number.

SPEAKER_00: But it's important to understand that in that 2,000 megawatts, there are site selectors

SPEAKER_00: for large hyperscalers that are sending out that same inquiry to hundreds of sites across

SPEAKER_00: the entire country.

SPEAKER_00: They're lurking for preliminary information.

SPEAKER_00: They're kicking the tires.

SPEAKER_00: These aren't what we would consider genuine requests for service.

SPEAKER_00: But it is a conversation that they are looking to start and it is information that they

SPEAKER_00: are looking to collect from the region, whether that's SMUD or our other regional partners

SPEAKER_00: and local governments.

SPEAKER_00: It does create a resource burden.

SPEAKER_00: When we boil that down to regional interest, we start to think about what does the customer

SPEAKER_00: count look like or what does the capacity look like where they maybe aren't reaching

SPEAKER_00: out to the region in general, but they're starting to reach out to SMUD directly.

SPEAKER_00: It's a phone call.

SPEAKER_00: It's an email.

SPEAKER_00: That probably cuts that number at the national level in half, from over 2,000 down closer

SPEAKER_00: to 1,000.

SPEAKER_00: A much smaller number, but still a significant number.

SPEAKER_00: Important at this stage, we're not talking about somebody calling in and saying, I'm

SPEAKER_00: interested in address 1234 and I want to know about your load serving capability.

SPEAKER_00: It's a very high level conversation.

SPEAKER_00: It's not a detailed exchange, often still with these site selectors that are looking

SPEAKER_00: at hundreds of sites across the country.

SPEAKER_00: We then boil it down one level further to where we're talking about actual site interest,

SPEAKER_00: actual service requests.

SPEAKER_00: What we can say is what started as that 2,000 number at that national level, which was not

SPEAKER_00: a realistic number.

SPEAKER_00: When we get down to this specific site interest, we're now talking about something on the order

SPEAKER_00: of a couple hundred megawatts, which feels small because I just prefaced that by starting

SPEAKER_00: this conversation at 2,000 megawatts.

SPEAKER_00: That is a considerable number and something that we would need to be very, very thoughtful

SPEAKER_00: about in terms of how we plan and how we approach that opportunity.

SPEAKER_00: In order to be successful in this space, we need to make sure that we're quickly distinguishing

SPEAKER_00: between those speculative conversations and the real world conversations.

SPEAKER_00: Because from a resource standpoint, that's time that we're spending to get those responses

SPEAKER_00: out.

SPEAKER_00: That's resources that we're spending that could otherwise be committed to enhancing

SPEAKER_00: the service or bringing customers online that are earlier in the queue.

SPEAKER_00: The ability to respond quickly to these load serving questions also becomes critical.

SPEAKER_00: The industry around AI and data centers is moving so quickly.

SPEAKER_00: A lot of that is driven by a land grab type conversation of how quickly can I get my data

SPEAKER_00: center online?

SPEAKER_00: How quickly can I develop the compute needed to launch my service, to launch my business?

SPEAKER_00: There's a lot of interest in moving quickly.

SPEAKER_00: Understanding how we plan for that growth, but in planning for that growth, making sure

SPEAKER_00: that the data center customers or the large load customers that we're bringing online

SPEAKER_00: are sustained load growth.

SPEAKER_00: A customer that we help bring online that's here for two or three years and is tied in

SPEAKER_00: with a merger or an acquisition for a larger data center or doesn't have a sustainable

SPEAKER_00: business model has a significant risk that we need to be thoughtful about in terms of

SPEAKER_00: how we serve that load, how we commit investment into resources to bring that customer online.

SPEAKER_00: Because when we're successful and when these data centers grow, it does translate into

SPEAKER_00: regional economic growth.

SPEAKER_00: It does translate into new jobs.

SPEAKER_00: That's meaningful for us.

SPEAKER_00: It's meaningful for our community and the growth that we're trying to create for the

SPEAKER_00: region.

SPEAKER_00: Inference data centers, of particular note, are very interesting.

SPEAKER_00: Because while training data centers are very large facilities that run all the time, there's

SPEAKER_00: not a huge number of jobs associated with those.

SPEAKER_00: It's often largely construction, but not a lot of follow-on jobs and technical careers

SPEAKER_00: that are tied in with that.

SPEAKER_00: When we start talking about inference, inference also means that there's companies that are

SPEAKER_00: developing those technology solutions that are running on those inference data centers.

SPEAKER_00: There's software developers.

SPEAKER_00: There's coders.

SPEAKER_00: There are individuals that are actually doing the work to utilize those models to create

SPEAKER_00: those services.

SPEAKER_00: Those are jobs.

SPEAKER_00: Those are jobs that universities and colleges in our region are training for and that entrepreneurs

SPEAKER_00: are aspiring to deliver.

SPEAKER_00: To the extent that we are part of bringing these data centers here, especially inference

SPEAKER_00: data centers, we can be part of that economic growth story and we can be part of driving

SPEAKER_00: that job growth for the region.

SPEAKER_00: It's also important that we think about the alignment with other electrification sectors.

SPEAKER_00: If we go back three or four years, a lot of this conversation on capacity and load growth

SPEAKER_00: was centered around DC fast chargers, electric vehicle charging growth, and how we're going

SPEAKER_00: to serve the growing number of electric vehicles.

SPEAKER_00: We've recently passed 80,000 electric vehicles or plug-in vehicles now in territory and a

SPEAKER_00: significant number being adopted every single month in terms of percentage of new car purchases.

SPEAKER_00: That's still a growing load.

SPEAKER_00: That is still a capacity conversation that is happening and we can't have these dialogues

SPEAKER_00: in isolation of each other.

SPEAKER_00: We need to make sure that our electrification initiatives continue to be aligned with our

SPEAKER_00: growth strategies around data centers and how we're addressing both this opportunity

SPEAKER_00: but navigating those challenges of the resources and how we balance that.

SPEAKER_00: Part of that leads to what we're doing on the research and development side in terms

SPEAKER_00: of our R&D roadmap.

SPEAKER_00: The good news on the data center side of this conversation is that while data centers

SPEAKER_00: and AI are daunting, fast-moving technologies, at the core of these data centers are topics

SPEAKER_00: that we're very comfortable with from a research and development standpoint.

SPEAKER_00: Infrastructure utilization, cooling systems, demand flexibility, reliability, load management,

SPEAKER_00: these are all words that have been the centerpiece of pilots and technology work that we have

SPEAKER_00: done for the last 20 years or more at this organization.

SPEAKER_00: Energy efficiency has been a keystone for California, driven by the California Energy

SPEAKER_00: Commission for decades.

SPEAKER_00: Load flexibility and demand response are tools that we're utilizing to make best use of the

SPEAKER_00: infrastructure that we have and drive these solutions forward with limited investment

SPEAKER_00: and maximum affordability.

SPEAKER_00: We're also looking at on-site generation co-benefits.

SPEAKER_00: For data centers, uptime is a critical priority for them, which means that these data centers

SPEAKER_00: will be investing in backup power.

SPEAKER_00: They'll likely be investing in batteries or other energy storage technologies or solutions

SPEAKER_00: to the extent that we can partner with these data centers to utilize that technology to

SPEAKER_00: meet our grid requirements or to support our infrastructure during peak hours.

SPEAKER_00: That co-benefit where they maintain that reliability, that backup resource, but we can leverage

SPEAKER_00: that resource while they don't need it because we're providing reliable power to help support

SPEAKER_00: our market needs and our larger system needs, gives us pathways to collaboration and enhanced

SPEAKER_00: affordability for the overall portfolio.

SPEAKER_00: The bottom line here is that emerging technologies and advancements are something that we're

SPEAKER_00: anticipating in this space.

SPEAKER_00: It's a driver for a significant amount of our collaboration with folks like the National

SPEAKER_00: Lab of the Rockies and others in this space.

SPEAKER_00: Keeping our eye on the horizon from a research and development standpoint is a critical job

SPEAKER_00: that we have at hand and one that, while I use the word horizon, is only just barely

SPEAKER_00: on the horizon given the speed of the conversation that we're having.

SPEAKER_00: The last item that I want to touch here before I hand it off to our friends here from the

SPEAKER_00: lab is that it's not just the data center growth that we're having a conversation about.

SPEAKER_00: We're also very excited about the growth of the use of AI tools here at SMUD to the extent

SPEAKER_00: that these data centers are getting built, that more applications are being supported,

SPEAKER_00: that more innovation is happening in this space.

SPEAKER_00: It's driving the creation of new tools and we are embracing those tools here at SMUD.

SPEAKER_00: The tip of that spear is really our policies and governance as being at the heart of everything

SPEAKER_00: that we do with AI.

SPEAKER_00: We make sure that we're looking into all of the use cases and understanding those use

SPEAKER_00: cases and have appropriate controls in places.

SPEAKER_00: We bring those systems online.

SPEAKER_00: We're providing all of our employees training on how to utilize AI as a priority.

SPEAKER_00: It's not just enough to offer the tools, but making sure that individuals know how to use

SPEAKER_00: those tools.

SPEAKER_00: We have general purpose chatbots for our staff to engage with and our teams to engage with

SPEAKER_00: to help us brainstorm, draft papers, draft memos, reference material from the organization.

SPEAKER_00: For example, today I typed in Rule 16 because I wanted to make sure that I understood how

SPEAKER_00: Rule 16 would apply to data centers and making sure that I understood the cost correlation

SPEAKER_00: between bringing a new customer online and what their cost obligation was to pay for

SPEAKER_00: those services and facilities.

SPEAKER_00: That chat GPT tool or that SMUD GPT tool gave me the opportunity to quickly access

SPEAKER_00: that information, ask me follow-on questions about the different policies and procedures

SPEAKER_00: that I may also be interested in.

SPEAKER_00: SD2 about our competitive rates priority and how we have an underscored absolute priority

SPEAKER_00: of affordability and equity across our customer rates.

SPEAKER_00: Those are things that come to mind for me.

SPEAKER_00: With their GPT tools, it's able to actually be brought to the front with minimal effort

SPEAKER_00: and time and maximum efficiency to drive this conversation forward.

SPEAKER_00: We continue to customize these tools with our own material.

SPEAKER_00: All of the data that we're using within these tools starts here at SMUD, stays here at SMUD.

SPEAKER_00: This is not information that's going out to these large centers.

SPEAKER_00: We're keeping that here.

SPEAKER_00: We have a strong data custodianship over all of that information.

SPEAKER_00: Further beyond just chat bots, looking at applications and how we can enhance those

SPEAKER_00: services to not just be language but also image and workflows and overall operational

SPEAKER_00: efficiency.

SPEAKER_00: It's an area that we're very excited about.

SPEAKER_00: It's just the start of the conversation, but it's something that's moving very, very quickly.

SPEAKER_00: As an organization, we're very much embracing.

SPEAKER_00: We're excited for where this conversation is going and the opportunity to bring these

SPEAKER_00: types of customers and loads online and utilize these tools in a responsible way.

Unknown: Thank you, James.

SPEAKER_03: Real quickly, is the load associated with data centers, what percentage of it, I guess,

SPEAKER_03: is associated with the actual running the computers themselves, the servers versus the

SPEAKER_03: cooling?

SPEAKER_03: Is that different for a training data center versus an inference data center?

SPEAKER_03: Are we likely to see a difference in load for those facilities here seasonally, winter

SPEAKER_03: to summer?

SPEAKER_00: I can provide a general answer and then I will defer to the more technology experts

SPEAKER_00: and encourage you to re-ask that question.

SPEAKER_00: Cooling is a major load factor in the conversation.

SPEAKER_00: With cooling, the ramps and ebbs and flow in the load play a large part in the amount

SPEAKER_00: of cooling that is needed.

SPEAKER_00: For a learning data center, there is no ramp.

SPEAKER_00: It is maximum all the time.

SPEAKER_00: That becomes a significant load component.

SPEAKER_00: For an inference data center, it is not as significant because that ramp up only happens

SPEAKER_00: for a smaller period of time.

SPEAKER_00: There is a cool down window.

SPEAKER_03: Are these considered clean room facilities where you can't just bring in outside air

SPEAKER_00: when it's cool outside and cool them out?

SPEAKER_00: That's correct.

SPEAKER_00: A significant number of these facilities are also water cooled.

SPEAKER_00: It's not all forced air cooling.

SPEAKER_00: There are a number of different cooling solutions.

SPEAKER_00: In fact, some of the more newer technology is looking at potentially fully submerged

SPEAKER_00: cooling.

SPEAKER_00: The cooling space is an area where there is significant innovation going on.

SPEAKER_03: Thank you.

Unknown: Any other questions?

SPEAKER_09: I have a question.

Unknown: Sure.

SPEAKER_10: Say, James, help me remember here because I don't know this off the top of my head.

SPEAKER_10: Do we offer some kind of economic development rate to companies who come to locate here?

SPEAKER_10: At one point when I worked here, it was the number of jobs you created.

SPEAKER_10: I think we got rid of that.

SPEAKER_10: Can you bring me up to speed?

SPEAKER_00: We do have an economic development rate.

SPEAKER_00: That is to help drive new businesses and business growth and commercial development to the region.

SPEAKER_00: It is tied in with job growth.

SPEAKER_00: I don't know the specific percentages off the top of my head, but we could get you that

SPEAKER_00: additional information.

SPEAKER_00: What I do want to say is that the things that we got rid of or what you may be remembering

SPEAKER_00: back to is custom rate agreements.

SPEAKER_00: Custom rate agreements are not a standard offering.

SPEAKER_00: Any approval for a custom rate agreement would need to come back through the board for approval

SPEAKER_00: and governance.

Unknown: Thank you.

Unknown: Is that for any size contractor?

SPEAKER_00: Any custom agreement, yes.

Unknown: Great.

SPEAKER_09: Thank you so much.

SPEAKER_09: I think we are ready to move on to the rest.

SPEAKER_09: I'm sure more questions will come.

SPEAKER_09: Thanks for coming all the way from Colorado.

SPEAKER_05: Thank you.

SPEAKER_05: Great to be here.

SPEAKER_05: Thanks for having me.

SPEAKER_05: My name is Elliot Rynakowski.

SPEAKER_05: I'm from the National Laboratory of the Rockies.

SPEAKER_05: Here from Golden, Colorado.

SPEAKER_05: Really great to be with you today.

SPEAKER_05: I'll be talking about large load planning from a grid planning and analysis point of

SPEAKER_05: view.

SPEAKER_05: I come from the Grid Planning Analysis Center.

SPEAKER_05: I manage a group of researchers that develop models, tools, data sets for use within the

SPEAKER_05: National Laboratory Complex of the Department of Energy.

SPEAKER_05: Also we ensure that our tools and data are useful for regional planning authorities and

SPEAKER_05: utilities and municipal governments.

SPEAKER_05: Part of our mission is to get these capabilities out into the world and into industry.

SPEAKER_05: What we know and as Jay mentioned is that we've been working on the challenges of an

SPEAKER_05: evolving power sector for many years now.

SPEAKER_05: We know that the power system is changing.

SPEAKER_05: We know that technology does not stand still.

SPEAKER_05: We've been managing a changing supply side infrastructure.

SPEAKER_05: We've been managing extreme weather events on the power grid and planning for adoption

SPEAKER_05: of distributed energy resources, electric vehicles, a new suite of novel technologies

SPEAKER_05: that can provide generation on the supply side.

SPEAKER_05: At the same time, what we've seen over the last few years is that electricity demand

SPEAKER_05: is starting to grow in a rate that we haven't seen since the 90s.

SPEAKER_05: In different pockets of the United States, what we're seeing is growth both in magnitude

SPEAKER_05: and year over year percentage that we haven't seen since the 40s, 50s, and 60s in some cases

SPEAKER_05: when we have massive electrification of industry.

SPEAKER_05: That growth is in large part driven by data centers.

SPEAKER_05: Just in...

SPEAKER_05: Sorry, I'll skip ahead here for a second.

SPEAKER_05: Just in the end of 2023, we had about 4.5% of the U.S. electricity demand was attributed

SPEAKER_05: to data centers.

SPEAKER_05: There's a lot of uncertainty with that number, so I want to caveat all of these numbers and

SPEAKER_05: say that there are error bounds around them.

SPEAKER_05: The fact is that since the end of 2023, and if you look at the end of 2025, so in the

SPEAKER_05: course of just two years, we've had an explosion of data center capacity.

SPEAKER_05: We're now at about 47,000 megawatts across the U.S.

SPEAKER_05: Like all the numbers on this slide, take that with a grain of salt because there are big

SPEAKER_05: error bounds around them.

SPEAKER_05: What we do know is that the applications for new data centers, or the planned capacity

SPEAKER_05: and the capacity under construction, could see that number double by 2030.

SPEAKER_05: If we're going from 4.5% at the end of 2023, where we are today could be upwards of 7%

SPEAKER_05: across the U.S. and by 2030, between 6.7% and 12% of total U.S. energy consumption from

SPEAKER_05: data centers.

SPEAKER_05: That's really the challenge that we're dealing with is managing this growth within the bounds

SPEAKER_05: of the constraints that we put on the power system and on the consumer.

SPEAKER_05: To back up a second, a little bit about our laboratory's vision and the role that we play

SPEAKER_05: as a national laboratory.

SPEAKER_05: We focus on bringing best in class cutting edge grid planning and analysis capabilities,

SPEAKER_05: both in terms of technology and hardware testing, working with industry to test novel technologies

SPEAKER_05: in our laboratories and develop new capabilities, and to do analysis and applied analysis with

SPEAKER_05: utilities and with our industry partners.

SPEAKER_05: Across this space, we focus on a few pillars.

SPEAKER_05: We have abundance as one of the key visions of the Department of Energy is enabling load

SPEAKER_05: growth within the bounds of affordability and reliability.

SPEAKER_05: We ensure that when we are planning the system, we're not unduly burdening customers, and

SPEAKER_05: that we're ensuring a secure and reliable grid.

SPEAKER_05: Data centers are currently the largest source of growth for many utilities, but we see that,

SPEAKER_05: as James mentioned, other sources of growth are also continuing.

SPEAKER_05: In many places, electric vehicles are coming onto the grid, electrified industries as well.

SPEAKER_05: We do see technologies coming on that will further require the development of the system

SPEAKER_05: to require higher amounts of electricity.

SPEAKER_05: For example, hydrogen production from electrolyzers.

SPEAKER_05: I mentioned that data centers are growing, but they're also growing in size.

SPEAKER_05: What we're seeing is that the trend has been towards these very large training data centers

SPEAKER_05: that could be upwards of 100 megawatts per facility.

SPEAKER_05: If we're looking at inference data centers, those are smaller in size and closer to the

SPEAKER_05: grid edge, so closer to consumers, as James mentioned, to minimize that latency.

SPEAKER_05: One of the pathways to solve this challenge is to really understand the siting and how

SPEAKER_05: developers are thinking about their siting decisions.

SPEAKER_05: One of the ways we do that is we gather and understand data around the infrastructure

SPEAKER_05: that supports data centers.

SPEAKER_05: This map is from the Speed to Power Initiative that the National Lab of the Rockies implemented

SPEAKER_05: with the Department of Energy.

SPEAKER_05: It brings together a collection of data that is both nationwide and can provide insights

SPEAKER_05: for local governments and local utilities.

SPEAKER_05: It looks at transmission lines, high voltage transmission lines that can bring power to

SPEAKER_05: load centers.

SPEAKER_05: It also incorporates the data infrastructure and the fiber optic cables that move data

SPEAKER_05: between places and load centers across the country.

SPEAKER_05: The gray areas here are those data centers that are currently planned.

SPEAKER_05: Those are announced projects.

SPEAKER_05: You can see a lot of them are at such scales that we just haven't seen before added to

SPEAKER_05: the grid.

SPEAKER_09: I'm sorry, I'm having a really hard time with this slide in understanding what gray is.

SPEAKER_09: Are you talking about the white dots?

Unknown: Yes, so the gray bubbles.

SPEAKER_05: The gray bubbles are those data centers and those are the sizes of the data centers in

SPEAKER_05: terms of megawatts of capacity.

SPEAKER_05: Those are just the planned ones.

SPEAKER_05: All of the yellow and the orange dots, those are the capacity of the data centers that

SPEAKER_05: are currently existing.

SPEAKER_05: You can see the difference between the sizes of the historic builds and those builds that

SPEAKER_05: are coming on that are currently under construction.

SPEAKER_05: In many cases, they're double or triple the biggest ones that we've seen historically.

SPEAKER_09: It doesn't look like there's one really in California.

SPEAKER_09: It looks like there's a big one in Reno.

SPEAKER_05: That's what we see is under construction.

SPEAKER_05: These data are not complete of all applications.

SPEAKER_05: We go through a process of screening those data sources to make sure that we're in this

SPEAKER_05: map only displaying those that are high confidence that are going to come online.

SPEAKER_05: In many cases, you don't see the smaller data centers that are inferenced because they're

SPEAKER_05: too small to see at the scale of this map.

Unknown: Thank you.

Unknown: Pills.

Unknown: Pills.

Unknown: As James mentioned, not all data centers are made the same.

SPEAKER_05: We talked about inference.

SPEAKER_05: We talked about training.

SPEAKER_05: We also know that we need to understand how those data centers will operate from an infrastructure

SPEAKER_05: point of view.

SPEAKER_05: Building out the infrastructure for a data center that's going to operate at 40% of its

SPEAKER_05: planned capacity will be very different from the infrastructure needed to power a data

SPEAKER_05: center that's going to be running full on in both cooling and compute demand.

SPEAKER_05: Getting that information from developers is a key challenge and something that we are

SPEAKER_05: seeing is more and more important from utilities as they manage this application process.

SPEAKER_05: We're seeing utilities explore different options for actually incentivizing large load developers

SPEAKER_05: to provide that data and to provide it accurately, both in terms of technical design and the

SPEAKER_05: likelihood of a project reaching maturity.

SPEAKER_05: One number that we can cite is that in ERCOT, they calculated that the percentage of large

SPEAKER_05: loads that were expected to be in service at the end of 2024, about half of them were

SPEAKER_05: actually in service shortly thereafter.

SPEAKER_05: The project realization stage can be highly uncertain.

Unknown: I'm sorry, just to clarify.

SPEAKER_09: You're saying that almost half didn't actually get online on time?

SPEAKER_05: That's right.

SPEAKER_09: What was that due to?

SPEAKER_05: There could be various reasons.

SPEAKER_05: It depends.

SPEAKER_05: That's some of these interconnection requests for speculative.

SPEAKER_05: It could be that they have energized behind the meter before they have an interconnection

SPEAKER_05: with the utility.

SPEAKER_05: Another trend that we're seeing is that utilities are increasingly being challenged to be proactive

SPEAKER_05: in communicating to data center developers where they should be looking to connect so

SPEAKER_05: that we can more efficiently utilize the latent grid capacity that exists on the network and

SPEAKER_05: more proactively screen projects so that we can efficiently manage that interconnection

SPEAKER_05: process.

SPEAKER_05: Part of providing this information to data centers is understanding what they're looking

SPEAKER_05: for.

SPEAKER_05: What we've seen from the industry is that there are three pillars of the data center

SPEAKER_05: developers problem is they're looking for speed.

SPEAKER_05: They want to interconnect and energize as quickly as possible.

SPEAKER_05: They're looking for high levels of reliability so they're trying to understand where is the

SPEAKER_05: infrastructure that will provide them with the reliability metrics that they're looking

SPEAKER_05: for.

SPEAKER_05: They're looking for market opportunities in many cases that is translated into low rates.

SPEAKER_05: They're understanding where are the opportunities for both low electricity rates but also their

SPEAKER_05: water and land costs.

SPEAKER_05: We can take that data center developer problem and bridge the gap with what utilities are

SPEAKER_05: concerned about into a framework around data center siting.

SPEAKER_05: We have a lot of experience in doing this across the industry.

SPEAKER_05: We have experience in doing this on the supply side with renewable energy technologies that

SPEAKER_05: also have this interconnection queue process and siting challenges.

SPEAKER_05: We can apply that to the data center siting problem.

SPEAKER_05: In applying this framework, we can narrow down where are those locations where data

SPEAKER_05: centers in the short and medium term would be most efficient to integrate into the network.

SPEAKER_05: If we apply this to a national scale, we can identify where are those locations that might

SPEAKER_05: have those low hanging fruit opportunities.

SPEAKER_05: We can then take this framework and apply it to each individual municipality or state

SPEAKER_05: government so that they can tune the metrics that are most impactful to their policies

SPEAKER_05: and regulations and goals.

SPEAKER_09: So prioritizing long haul fiber connectivity, we have that here.

SPEAKER_09: And then access to natural gas, why that?

SPEAKER_05: In many cases, data center developers are interested in building natural gas generation

SPEAKER_05: behind the meter or inside the fence line.

Unknown: Good to know.

SPEAKER_05: Part of that is a bridging power solution.

SPEAKER_05: One of their goals is to speak to power.

SPEAKER_05: If they have an interconnection request that doesn't fill their full capacity that they

SPEAKER_05: want to bring online in the short term, they'll build some bridging power behind the meter

SPEAKER_05: and power it up either through mobile gas generators or through generators that they

SPEAKER_05: then use as backup power once the full grid connection is available.

Unknown: And I heard these diesel generators is backup a lot too?

Unknown: Yes.

SPEAKER_05: Yes.

Unknown: Thank you.

Unknown: So once we take that siting question, we can then apply our system reliability metrics

SPEAKER_05: and really think about evaluating how we maintain reliability and what opportunities there are

SPEAKER_05: for data centers to contribute to the reliability of the network and ensure that they are participating

SPEAKER_05: as what I would call a good grid citizen in upholding their reliability.

SPEAKER_05: So we think of reliability in four categories or four main pillars.

SPEAKER_05: One of them is capacity, so ensuring that you have adequate resources to meet the demands

SPEAKER_05: that are coming online.

SPEAKER_05: Another one is flexibility.

SPEAKER_05: So when in operations, once the data centers are energized and operating, do they have

SPEAKER_05: the ability to uphold the flexibility that the system needs to stay reliable on an hourly

SPEAKER_05: and weekly and monthly basis?

SPEAKER_05: And then stability questions.

SPEAKER_05: So on a more electrical engineering dimension, can they maintain the frequency of the system

SPEAKER_05: and the local voltage so that power quality remains high and then you have high reliability

SPEAKER_05: for all of the customers on the network?

SPEAKER_05: So one of the opportunities that we see is really evaluating what is the existing transmission

SPEAKER_05: hosting capacity and using novel technologies, like grid-enhancing technologies, one of

SPEAKER_05: those is dynamic line ratings.

SPEAKER_05: So that's a way of really evaluating what is the existing capacity of the transmission

SPEAKER_05: grid and using data to dynamically understand the latent hosting capacity that is there

SPEAKER_05: in the network so that you don't have to build new infrastructure in all cases.

SPEAKER_05: And one of the ways we do that is we apply our weather modeling and dynamic modeling

SPEAKER_05: of the heat of the network to understand where are there opportunities to increase the rating

SPEAKER_05: of existing lines.

SPEAKER_05: What we also see is that many data centers are connecting to the distribution system

SPEAKER_05: and at some level there will be a distribution system component to all data center interconnections.

SPEAKER_05: And distribution system reliability is often the source of reliability challenges.

SPEAKER_05: So what this figure shows is that the metric of reliability in terms of major interruptions

SPEAKER_05: on average for a year is actually a bigger challenge on the distribution system than

SPEAKER_05: it is at the high voltage bulk power system.

SPEAKER_05: And those are the blue bars there.

SPEAKER_05: The dark blue bars are interruptions in terms of minutes for an average number of minutes

SPEAKER_05: for a customer without major events and the light blue bars is if you add major events.

SPEAKER_05: So that distribution system reliability metric kind of pales the bulk power system.

Unknown: And so what we see is that there are different mechanisms for designing distribution systems

SPEAKER_05: to provide higher levels of reliability.

SPEAKER_05: This is an active area of work is bolstering the reliability of the distribution system

SPEAKER_05: that serves the data centers.

SPEAKER_05: But even with higher levels of reliability we're seeing that data centers are targeting

SPEAKER_05: reliability levels that are even higher than what we can do in the best designs that are

SPEAKER_05: currently in place.

SPEAKER_05: So that kind of four nines of reliability is what many data centers are targeting.

SPEAKER_05: And even higher levels, right, five nines or six nines.

SPEAKER_05: And that could be achieved with more behind the meter generation but of course that comes

SPEAKER_05: with tradeoffs.

SPEAKER_05: And then on the data center interconnection study process when we're evaluating how these

SPEAKER_05: data centers will impact reliability there's a real opportunity to use AI and incorporate

SPEAKER_05: AI into that process to speed that interconnection study timeline which is often a bottleneck

SPEAKER_05: and causes extra costs.

SPEAKER_05: And one novel application that we're seeing is injecting sort of AI in screening interconnection

SPEAKER_05: applications so that we can more quickly move through them and save time both for developers

SPEAKER_05: and for utilities in doing that study process.

SPEAKER_05: And then finally a really active area of work in the industry is to understand how we can

SPEAKER_05: evaluate large load customers in terms of their impacts on the wider customer base.

SPEAKER_05: And do understand their impacts on utility revenue requirements and utility rate design.

SPEAKER_05: Now utilities can design rates to ensure that customers can maintain affordable rates across

SPEAKER_05: customer classes.

SPEAKER_05: So in summary, you know, we've covered a lot of bases.

SPEAKER_05: We've kind of covered a large scope of the types of analysis that is currently being

SPEAKER_05: applied to this data center challenge.

SPEAKER_05: And what this requires is really solutions from the grid edge so understanding what the

SPEAKER_05: needs are at the customer interconnection point for data centers up to the regional

SPEAKER_05: and national scale planning that happens to understand what are the infrastructure requirements

SPEAKER_05: to meet these data center demands while maintaining those guardrails of affordability and reliability.

SPEAKER_05: So with that, I'm happy to take any questions.

Unknown: Any questions?

SPEAKER_03: We hear all the time that I mean, I forget which organization it is that's restarting

SPEAKER_03: Three Mile Island or one of the units of Three Mile Island.

SPEAKER_03: Everybody else is investing in fusion technology in Virginia.

SPEAKER_03: I mean, these are huge investments.

SPEAKER_03: And they're, I mean, there's still three, four, five years down the line at best and

SPEAKER_03: probably more like 10.

SPEAKER_03: I guess I'm a little boggled by that.

SPEAKER_03: And where else are you seeing that kind of interest in that large scale of investment?

SPEAKER_05: Yeah.

SPEAKER_05: Yeah.

SPEAKER_05: Well, so what we're seeing is that developers are now still very interested in that area

SPEAKER_05: around Northern Virginia, but now starting to look at other parts of the country and

SPEAKER_05: very actively looking at other parts of the country where they can get faster speed to

SPEAKER_05: power and maintain low rates.

SPEAKER_05: And those are what are driving some of the big announcements of generator contracts

SPEAKER_05: and new generator procurement.

SPEAKER_05: What I'll say about the Three Mile Island example is that that is capacity that has

SPEAKER_05: already existed in the grid.

SPEAKER_05: And so that is some latent capacity that data center can look at to utilize.

SPEAKER_05: But what we'll need as this capacity grows is new capacity on the grid, right?

SPEAKER_05: Because at some point we'll exhaust and we're, I think, in the foreseeable future seeing

SPEAKER_05: that exhaustion of what is already built in the grid and we'll need new capacity.

SPEAKER_03: This is really more a question for my colleagues, I guess.

SPEAKER_03: Something we've made this zero carbon promise to our customers by 2030.

SPEAKER_03: What if somebody came to us and said, we'll give you $250 million for one of your power

SPEAKER_03: plants in South Sacramento.

SPEAKER_03: It's no longer yours.

SPEAKER_03: You're not producing any carbon.

SPEAKER_03: We need it.

SPEAKER_03: We're going to put in a 200 megawatt data center.

SPEAKER_03: And you can use the $250 million to keep your rates low or to invest in other zero carbon

SPEAKER_03: technology someplace.

Unknown: I'm really glad you brought that up because I will be bringing up an idea that I had about

SPEAKER_09: this that I think we need to actually do in ad hoc because there's so much around this

SPEAKER_09: that I think we need to dig into with the staff about the siting of these large facilities.

SPEAKER_09: I have like 100 questions by myself.

SPEAKER_09: So I just feel like there's no way we can get through all the detail.

Unknown: I mean, there's so many questions here.

SPEAKER_09: We could go on for hours.

SPEAKER_09: So that's my suggestion is that we do not have...

SPEAKER_09: I didn't expect an answer.

SPEAKER_03: But that is that.

SPEAKER_09: I'm glad you brought it up because I was going to bring it up anyway.

Unknown: But we can talk about that at the end.

SPEAKER_10: I'm just wondering what is it that an ad hoc committee would do?

SPEAKER_10: We've already heard about what's going on in the country.

SPEAKER_10: We've heard from our James telling us that we're getting some requests but not huge.

SPEAKER_10: I'm just...

SPEAKER_10: I'm confused about what it is that you want to study.

SPEAKER_09: Well, I'll ask this question.

SPEAKER_09: How many jurisdictions in this country have a policy about siting of data centers?

SPEAKER_05: It depends on what you mean by policy.

SPEAKER_05: But I would say that this is an active area of work across all utilities and state governments

SPEAKER_05: that we have worked with.

Unknown: Okay.

Unknown: Well, Greg and I both heard, Director Fishman and I, that we were at an event yesterday

SPEAKER_09: where we met with Sud, who is a gentleman we've known from other events.

SPEAKER_09: And he was saying that I think that San Jose has their own in their general plan siting

SPEAKER_09: policy for data centers.

SPEAKER_09: He's a city council member for Santa Clara.

SPEAKER_03: Santa Clara, sorry.

SPEAKER_03: And so he's on the Santa Clara utility.

SPEAKER_09: So I guess one of the things I would like to do in an ad hoc is gather up the policies

SPEAKER_09: that we've seen other jurisdictions do and contact them and investigate what policies

SPEAKER_09: do they have.

SPEAKER_09: Do they have them?

SPEAKER_09: Was something useful in there?

SPEAKER_09: This is a big deal.

SPEAKER_09: And if we're going to, we've already had our customers come and say, I still live in Texas,

SPEAKER_09: our rates went way up because our jurisdiction enticed these companies to come here.

Unknown: They didn't create a lot of jobs.

Unknown: They created a huge amount of demand.

SPEAKER_09: And then they sent the rates up throughout the jurisdiction.

SPEAKER_09: I'm not interested in a cost shift.

SPEAKER_09: And I'm certainly not interested in making natural gas or turning on diesel generators

SPEAKER_09: when we've promised zero carbon by 2030 either.

SPEAKER_09: So I just think there's a lot of things that we could look at as in an ad hoc committee

Unknown: around whether we think there should be a large power request policy that the board

SPEAKER_09: gives to the staff to give them some parameters on where the board's at.

SPEAKER_10: I don't disagree with wanting to set some parameters.

SPEAKER_10: I absolutely do think we should have a policy.

SPEAKER_10: And I'll look forward to working on that in the policy committee.

SPEAKER_10: But I don't think that our staff doesn't have their arms around this.

SPEAKER_10: Again, I'm more than happy to have the board set a policy if there are those that feel

SPEAKER_10: that we need to do this.

SPEAKER_10: But we're not doing all of the fearful things.

SPEAKER_10: We're not running into agreements and making our customers pay for other infrastructure.

SPEAKER_10: We're not saying come to Sacramento County no matter who you are.

SPEAKER_10: We're actually setting guidelines.

SPEAKER_10: We have our staff that are talking to folks.

SPEAKER_10: So I feel like we are on top of this.

Unknown: However, I think it's always good to have policies.

SPEAKER_10: So I don't object to bringing a policy forward to the policy committee.

SPEAKER_09: Well, and again, ad hoc may not even suggest a policy, but it's worth investigating at least.

SPEAKER_09: Yes, Director Kurt.

Unknown: Actually, I speak to the policy question.

SPEAKER_02: I had a question for our guest here.

SPEAKER_02: So we're very proud of the fact that we have some of the lowest rates in California.

SPEAKER_02: But I'm mindful of the fact that they're much higher than many other states.

SPEAKER_02: And so just how attractive are local lower in their equations that these site selection

SPEAKER_02: committees are making for the data centers, how big a deal is that to them, the lower rates?

Unknown: Yeah.

SPEAKER_05: So I'm not an expert on Sacramento's rates relative to nearby jurisdictions.

SPEAKER_05: But what I will say is that rates are only one of the components or one of the metrics

SPEAKER_05: that data centers are interested in looking at.

SPEAKER_05: And in the prioritization order, they're not on top.

SPEAKER_05: So it's speed to power and getting energized is their primary concern, is what we're seeing.

Unknown: Okay, thanks.

SPEAKER_02: And to the policy question, we have 47 policies now.

SPEAKER_02: I think within those is all the guidance the staff already needs.

SPEAKER_02: And so I'm not sure that we need to add a 48th policy just to talk about data centers.

SPEAKER_02: But see what we can come up with.

SPEAKER_02: I could be convinced otherwise.

SPEAKER_02: Thanks.

Unknown: It sounded like one of the functions of the ad hoc committee was gathering data and information

SPEAKER_06: wrong with it.

SPEAKER_06: That sounds more like a staff function to me.

SPEAKER_06: I was hoping that we would start talking here about what kinds of things should we be thinking

SPEAKER_06: about and giving staff some direction of hey, can you get us some information about

SPEAKER_06: this question, about this question, about this question.

SPEAKER_06: I don't see that there's a need to have an ad hoc committee.

SPEAKER_06: I do think it's something that we need to explore more.

SPEAKER_06: And once we get more information, maybe a policy would come out of it or maybe not.

Unknown: What I hope is if people have questions, doubts, things like that, express them here.

SPEAKER_06: That's what I was hoping to get out of this particular meeting.

SPEAKER_06: Let's get those out on the table.

SPEAKER_06: We might not necessarily get answers at this point.

SPEAKER_06: But express our concerns.

SPEAKER_06: So, no.

SPEAKER_09: That's fine, too.

SPEAKER_09: I just have so many questions.

SPEAKER_09: I don't know where to begin, honestly.

SPEAKER_09: I would have to write them all down.

SPEAKER_09: We heard one of the Santa Clara conditions for a data center was they had to be carbon neutral.

SPEAKER_03: He also talked about the big difference is he's on the city council and they control

SPEAKER_03: the utility so they also have land use authority.

SPEAKER_03: So they were doing things like requiring, I forget exactly what, something about how

SPEAKER_03: much water they use and reducing water use.

SPEAKER_03: They had aesthetic aspects to the building.

SPEAKER_03: So it wasn't just another big white box somewhere.

SPEAKER_03: It had to look decent.

SPEAKER_03: We don't have that kind of authority.

SPEAKER_03: But I mean, I think, and I don't know if we even have the authority to require them to

SPEAKER_03: be carbon neutral.

SPEAKER_03: If somebody wants to build a business here and we can serve the load, I'm not sure that

SPEAKER_03: we can say, and you must be carbon neutral or you must be environmentally sensitive in

SPEAKER_03: other ways.

SPEAKER_03: I don't know if we have that authority.

SPEAKER_03: But I guess it's my question for Paul and maybe other staff who's here.

SPEAKER_03: How much more direction do you want from us in regards to our decision making about how,

SPEAKER_03: where, when, if we cite large loads, regardless of what kind they are?

SPEAKER_04: So I'll actually try to attempt to answer that.

SPEAKER_04: I think what it is is that when you talk about data centers, that is what we call general

SPEAKER_04: load increases, like just no different than industrial load.

SPEAKER_04: I think we have policies now that kind of give us, govern us how to work with those

SPEAKER_04: developers.

SPEAKER_04: I think the question actually came in because right now that you are beginning to see and

SPEAKER_04: maybe possibly in the future to see what happened if you have load coming at 100 megawatts,

SPEAKER_04: which is like five of our neighborhood substations.

SPEAKER_04: And the other piece too is that actually the load gets bigger than the whole conversation

SPEAKER_04: I think that we're thinking about is that we may be established really thinking about

SPEAKER_04: when you look at other policies as we're collecting right now, what is the best way for us to

SPEAKER_04: protect our ratepayers, our customers from stranding investments?

SPEAKER_04: And there's different ways to do that.

SPEAKER_04: I mean, some are out requiring, like certain jurisdictions are requiring the new load to

SPEAKER_04: come in and bring everything in upfront.

SPEAKER_04: And then as the load materializes through the years, then there was still, because you're

SPEAKER_04: collecting the revenue through the rates, then you slowly pay them back.

SPEAKER_04: Some is actually requiring them to put a big deposit in the escrow account and then

SPEAKER_04: say, well, you know what, after five years or 10 years, load is going to be there.

SPEAKER_04: We're covering our money, it's not stranding investment.

SPEAKER_04: You get them back.

SPEAKER_04: So I think what staff is looking at right now is when the large load is that we're looking

SPEAKER_04: at, well, first we as staff is actually reviewing the different policies, like, for example,

SPEAKER_04: people from Washington, Oregon, Virginia, Oklahoma, that right now are seeing the big

SPEAKER_04: and also a comment in the Chicago area.

SPEAKER_04: So we as staff is actually looking at those and say, you know, this is really what other

SPEAKER_04: jurisdictions and other cities are putting together.

SPEAKER_04: And we are looking at it to say, you know, what they're putting together, do we need

SPEAKER_04: to bring it back to the board?

SPEAKER_04: If we don't think the existing policy is not enough to giving us, or maybe we want the

SPEAKER_04: board's input, what is the best way to protect it?

SPEAKER_04: Is it collecting a bonds upfront?

SPEAKER_04: Is it, you know, do we just say, you know what, for this size, we're not going to take

SPEAKER_04: the risk at all.

SPEAKER_04: We're not going to spend $500 million, $300 million and expect it to happen.

SPEAKER_04: So we're thinking about from a staff perspective, maybe the best way for us to do it is let

SPEAKER_04: staff maybe do the analysis, see what policies out there compared to what we have.

SPEAKER_04: And then if we think there's a gap, there may be something that we will actually bring

SPEAKER_04: it back to the, you know, have a discussion with the board, maybe with the policy chair

SPEAKER_04: and say, hey, this is something that I think would be a really good board discussion to

SPEAKER_04: give staff directions, because sometime when you look at the different policies, there's

SPEAKER_04: just one way to do it.

SPEAKER_04: And maybe this is a good time, instead of staff making the decision and say, well, we're

SPEAKER_04: just going to make you pay everything upfront.

SPEAKER_04: Maybe we want to solicit the board's input and create a policy that we want the developer

SPEAKER_04: to pay it upfront and then pay it back through the years.

SPEAKER_04: And then here's some of the conditions that we've got to put into it.

SPEAKER_04: So this is what staff is thinking right now.

SPEAKER_04: So I mean, you can, I agree with, you know, President Tamayo that, you know, collecting

SPEAKER_04: some of the data right now is not, it wouldn't be useful for the ad hoc.

SPEAKER_04: But I think what would be useful is once we find all those information to bring it back

SPEAKER_04: to the policy committee and then have a conversation with the board.

SPEAKER_04: And then as part of the pre-brief, I mean, now the board actually have a chance, look

SPEAKER_04: at it and ask the questions to see how we do it.

SPEAKER_04: And just, you know, if look, hearing the conversation, this is what I recommend is

SPEAKER_04: maybe have staff do the collections of different policies out there compared to what we have.

SPEAKER_04: And then let staff make a determination and say, do we think we have the policy that we

SPEAKER_04: need to address it?

SPEAKER_04: Or maybe right now what we have is not enough to cover the big data center that's coming

SPEAKER_04: through.

SPEAKER_04: That is bringing their own, you know, their own, basically their own generation behind

SPEAKER_04: the meter.

SPEAKER_04: And do we want to set a policy on that?

SPEAKER_04: Because right now we don't have a policy, you know, we have an obligation to serve,

SPEAKER_04: and we don't really have a right to say, well, you're going to put 100 megawatts here, I'm

SPEAKER_04: going to bring my own backup generator with diesel.

SPEAKER_04: We don't really have, we don't really have any policy or anything that we can't serve

SPEAKER_04: you because it's not legal.

SPEAKER_04: So there's a lot of work to be done there.

SPEAKER_04: So I think I would suggest to the board is maybe we'll put this on the parking lot, let

SPEAKER_04: staff do the data collection and summarize is what we found, summarize compared to what

SPEAKER_04: we, you know, with existing policy and then maybe if it triggers some policy question

SPEAKER_04: that we need the board to have, we'll bring it back to the policy chair and say let's

SPEAKER_04: have a conversation.

SPEAKER_03: Paul, when you collect all that data, you can feed it into chat GPT and let it do the

SPEAKER_03: analysis.

SPEAKER_12: You know, Director Fisherman, it's funny that you say that.

SPEAKER_04: There's a lot of folks actually doing that and actually using that to say, you know,

SPEAKER_04: based on all those policies, give me a summary or take the best out of those and create a

SPEAKER_04: new policy and then with the different prompts, you can actually get it to write it for you.

SPEAKER_04: So that's possible.

SPEAKER_04: But we haven't got that, we haven't got that really.

SPEAKER_04: I see James back there smiling right now.

SPEAKER_04: Let staff decide the best way to do the analysis.

SPEAKER_03: Why do I suspect I know the answer?

SPEAKER_02: It'll be good.

SPEAKER_04: So would that satisfy the board?

SPEAKER_09: That's fine with me.

SPEAKER_09: I just want to get a lot more information about this and get it in the public sphere

SPEAKER_09: about where having a more robust discussion, I thought we might get more into details about

SPEAKER_09: like what kind of conditions might we want to have.

SPEAKER_09: If they're going to come in with these big load demands, are we going to say, you know,

SPEAKER_09: we need you to produce some of your power.

SPEAKER_09: We need you to do backup battery, not diesel.

SPEAKER_09: I mean, here we are promising our customers zero carbon and we're going to power the thing

SPEAKER_09: that's going to maybe start burning a bunch of diesel.

Unknown: I think the question is a little more complicated than that, right?

SPEAKER_04: Because a lot of data centers actually have operation requirements that they will last

SPEAKER_04: not more than what the battery can handle.

SPEAKER_04: So a lot of right now, when you actually look at data center operations, I don't know all

SPEAKER_04: the, because I don't run the data center, but I know there's actually rules and regulations

SPEAKER_04: about what they need to put in and what's allowable.

SPEAKER_04: But this gets to my point.

SPEAKER_09: Is that undermining our entire policy on zero carbon by 2030?

Unknown: I think these are just big questions we have to discuss.

SPEAKER_09: That's all.

SPEAKER_09: It's just not a one hour presentation.

SPEAKER_09: We're done.

SPEAKER_09: Like we have a lot more work to do on this.

SPEAKER_09: And at least in my book, I look forward to hearing from the staff and doing a lot more

SPEAKER_09: digging, but I would hope it'd be a mutually beneficial agreement if we do get these big

SPEAKER_09: guys coming in and it's not just a one way ticket and we just, you know, I just want

SPEAKER_09: to protect the repairs.

SPEAKER_04: I think we all do.

SPEAKER_04: So, but I guess fortunate for us right now that we are not getting those, I need a 500

SPEAKER_04: megawatt in 18 months.

SPEAKER_04: Well, our system can't handle it anyway because we just don't have such a massive system and

SPEAKER_04: we don't have a lot of spare capacity that we can actually do.

SPEAKER_04: And a lot of those I think, you know, you see, you actually put on there a lot of actually

SPEAKER_04: requires transmission, anything more than that, you know, to create it, you need to

SPEAKER_04: create transmission and transmission at least five years out.

SPEAKER_04: So I think this is really what I think is actually brought up a very good point.

SPEAKER_04: Right now, when we talk to the data centers, it is the speed to power.

SPEAKER_04: That's why they're bringing in and they're bypassing the utility saying that, well, we

SPEAKER_04: can't make it.

SPEAKER_04: Well, we're going to put our generation in and then when the grid gets ready, we'll switch

SPEAKER_04: it over.

SPEAKER_04: So whether we like that answer or not, that's actually happening in the industry right now.

Unknown: Can I ask how long do you think it will be in the parking lot before we can bring it

SPEAKER_04: back?

SPEAKER_04: I would say probably collecting the data probably a month or two and then we'll bring back because

SPEAKER_04: I don't think it's that hard to get the policy.

SPEAKER_04: The policy will just, it is what they have out there right now.

SPEAKER_04: And we know and we have access to those policies available.

SPEAKER_04: I think what's going to take time is for us to balance that against existing policies,

SPEAKER_04: everything that we have and say, is it sufficient or not sufficient or what are some of the

SPEAKER_04: questions that we need to bring back to the board for the board to actually give directions.

SPEAKER_04: So I would imagine no more than three months.

Unknown: That'd be great.

Unknown: I'm looking at staff right now before I speak.

SPEAKER_04: I see James nodding his head all over there.

SPEAKER_04: Yeah, because I don't think compiling the policy is going to take that long because

SPEAKER_04: they already exist.

SPEAKER_04: And then I think we know what kind of mechanisms available to not to have cost shift or protect

SPEAKER_04: cost shift.

SPEAKER_04: I think that's actually very straightforward.

SPEAKER_04: It is now bringing back and say, okay, well, if there's some of the things that you just

SPEAKER_04: asked in terms of what we bring on new load, what is the impact to our zero carbon plan?

SPEAKER_04: I think we can answer that because that actually happens on new load that we bring on now and

SPEAKER_04: we have policies on that now.

SPEAKER_04: So I think we can actually get a lot of, at least within three months, you actually get

SPEAKER_04: the first round of discussion.

SPEAKER_04: I don't know.

SPEAKER_04: We can answer every single question even on the next round, but I think we're definitely,

SPEAKER_04: within six months, I think we have a policy.

SPEAKER_07: Okay, that'd be great.

SPEAKER_07: I guess I was going to say, I think we have existing policies that can support what our

SPEAKER_07: goals are, but we also need to be careful because saying things like you can't put a

SPEAKER_07: gas generator because it goes against a zero carbon, you have to be very careful of that

SPEAKER_07: because not just with data centers, but any large facilities that are emergency care centers

SPEAKER_07: like hospitals, you can't say they can't put a gas generator because we have requirements

SPEAKER_07: to be up.

SPEAKER_07: You can't over dictate things like that because then it becomes a safety issue.

SPEAKER_07: I know in this room, we care about zero carbon by 2030, but people care about being alive

Unknown: and that trumps, I would say, most people's concerns.

SPEAKER_07: So we just need to tread lightly when we are over prescribing policy because you have to

SPEAKER_07: look at human life, safety, predictability and reliability, whether it's a data center

SPEAKER_07: or a hospital or what have you, we can't be reaching into or hearsay regulations and joint

SPEAKER_07: commission regulations which regulate many industries, especially healthcare, because

SPEAKER_07: we're here to serve and keep people alive generally today.

SPEAKER_07: So we need to be careful when dictating policies as to what we're allowing people to build

SPEAKER_07: on their sites to protect their business and the people that they serve.

SPEAKER_07: You have to remember that it's not just, it's a goal, remember, it's a zero carbon by 2030

SPEAKER_07: goal and goals of healthcare facilities and other facilities that serve humans, their

SPEAKER_07: goal is to keep people healthy and alive.

SPEAKER_07: So you just need to be careful about that.

SPEAKER_03: So absolutely, I think there's a difference between what kind of backup generator a facility

SPEAKER_03: puts in place and somebody who's coming in and saying we're going to do our own generation

SPEAKER_03: behind the meter to run 24 seven.

SPEAKER_03: So Laura, I guess I have a question for you and if it takes whatever amount of time to

SPEAKER_03: research it, we don't have the ability to say, sorry, you can't come in behind and build

SPEAKER_03: something behind the meter like that.

SPEAKER_03: Does a local land use authority or the air quality management district have the ability

SPEAKER_03: to say, no, you cannot build a data center with a 50 megawatt diesel generator that you're

SPEAKER_03: going to run 24 seven.

SPEAKER_03: I suspect somebody has the ability to say no, but I'm not sure who.

SPEAKER_08: It's Laura Lewis, Chief Legal and Government Affairs Officer.

SPEAKER_08: So SMUD has limited authority and we have an obligation to serve other entities like

SPEAKER_08: the California Energy Commission, the Air Resources Board.

SPEAKER_08: There's also federal permits that have to be acquired, for example, for the large data

SPEAKER_08: centers that have diesel generation, they have to have a permit to run those.

SPEAKER_08: So there's other regulatory agencies that ensure that if those things are built that

SPEAKER_08: they're properly permitted, but that's not within SMUD's purview.

SPEAKER_06: So I like the idea of identifying where we might have gaps.

SPEAKER_06: It strikes me that there's – the other thing I'd like to find out other than just policies,

SPEAKER_06: I'd like to see what sort of business models are out there of how – what the arrangements

SPEAKER_06: are between utilities and these data centers.

SPEAKER_06: I'm actually a little bit more interested in the – it sounds to me like from what

SPEAKER_06: James put forward, it sounds to me that we're a lot more likely to get the – and I forget

SPEAKER_06: the term for the smaller ones, the inference data centers.

SPEAKER_06: It sounds like that's a lot more likely based on our proximity to Silicon Valley and the

SPEAKER_06: state of California.

SPEAKER_06: And also because our rates are so much higher, it's hard to imagine that somebody's going

SPEAKER_06: to want to locate – buy our electricity when we might be twice as expensive as in

SPEAKER_06: Nevada or Nebraska or wherever they can get cheaper electricity.

SPEAKER_06: And when they don't – they don't – those big ones don't need to worry about the latency.

SPEAKER_06: They're just creating a product, right?

Unknown: But I would like to learn more about what those potential business models – and I'd

SPEAKER_06: like to learn more about what are we thinking, what are the different pathways that we might

SPEAKER_06: be thinking, and how do those help our system and then how – what are we thinking is the

SPEAKER_06: best way to protect our ratepayers from the stranded assets.

SPEAKER_06: So not just a compilation of policies, but what are we thinking about how we're going

SPEAKER_06: to do that?

SPEAKER_06: I think I'm kind of comforted that we're unlikely to attract those big data centers,

SPEAKER_06: but – because then we won't face those – at least not the same magnitude of the

SPEAKER_06: issues as some other places apparently are.

Unknown: So thanks.

Unknown: Thanks.

Unknown: And I just want to say I'm really glad that you brought this up, Chair Sanborn, because

SPEAKER_10: it is a topic that a lot of people are talking about.

SPEAKER_10: I really like the idea of having staff collect the different policies that are out there.

SPEAKER_10: I'd like to know a little bit about how other communities are handling this.

SPEAKER_10: And then a discussion back here would be good, because I think we all share the same

SPEAKER_10: goal that we don't want our ratepayers to have to pick up the tab for a new AI business.

SPEAKER_10: So I think this is good that we're talking about it, and I'm glad that it will come

SPEAKER_10: back.

Unknown: Yeah.

Unknown: Me too.

SPEAKER_09: And I do think that – I mean, I don't know if you saw that Europe is now really cracking

SPEAKER_09: down on the tech folks, and I think we've gone fairly unregulated, and I think that

SPEAKER_09: might change in the future as information gets out with what they're identifying in

SPEAKER_09: Europe.

Unknown: So a lot of things could change soon on the demand side too.

SPEAKER_09: So thank you.

SPEAKER_09: This was very helpful.

SPEAKER_09: I know I haven't gotten any comment cards.

SPEAKER_09: Do we have any public comments?

SPEAKER_09: We do have one comment from John.

SPEAKER_08: And John, are you there?

Unknown: John, are you there?

SPEAKER_09: You can please go ahead and speak.

Unknown: Yeah, this is John Weber.

SPEAKER_11: Can you hear me?

SPEAKER_09: We can, John.

SPEAKER_09: Great to hear you.

Unknown: Just so you know, there's a pretty significant delay when you say I can speak until I get

SPEAKER_11: the button that says unmute.

SPEAKER_11: So just so you know, we'll cover that in the future.

SPEAKER_11: Yeah, a couple of things.

SPEAKER_11: So good evening, Chair, Board and committee members.

SPEAKER_11: Thank you for the presentations from the staff and from the person from Colorado.

SPEAKER_11: So I would really like to know what are SMUD's plans for a new rate class for large data

SPEAKER_11: centers to ensure other classes aren't impacted with increasing bills.

SPEAKER_11: It seems like that's the best solution from the utilities have already dealt with this

SPEAKER_11: before is they set up a specific new class for large data centers.

SPEAKER_11: And that protects the rest of the data centers from increasing bills.

SPEAKER_11: And then also, what will the impact be for SMUD's commitment to zero carbon in 2030 if

SPEAKER_11: two or three 50 megawatt centers come in?

SPEAKER_11: Is SMUD still going to be able to meet the zero carbon commitment?

SPEAKER_11: So those are my things.

SPEAKER_11: I thought you had a very good discussion.

SPEAKER_11: Coming from Texas, there is no community that welcomes data centers.

SPEAKER_11: Every community in Texas has been fighting them, fighting them because their rates go

SPEAKER_11: up and mostly because of water.

SPEAKER_11: Water is very scarce now in Texas and they use a ton of water for cooling.

SPEAKER_11: So I don't think we should incentivize any of these.

SPEAKER_11: I think we should disincentivize them through a new rate class because I do think there

SPEAKER_11: will be huge massive public outcry if these data centers are built here because I've seen

SPEAKER_11: it already in Texas.

SPEAKER_11: I know Virginia is dealing with it and other places.

SPEAKER_11: Thank you and have a good evening.

SPEAKER_09: Thanks John.

SPEAKER_09: And comment noted.

SPEAKER_09: Yeah, I have, again, a lot of questions about the environmental stuff outside of the energy

SPEAKER_09: side.

SPEAKER_09: There's a lot more questions that are coming up from data centers and cancer clusters and

SPEAKER_09: a whole bunch of stuff from the water.

SPEAKER_09: And that's you can all read online.

SPEAKER_09: It's all there.

SPEAKER_09: Are there any other comments?

SPEAKER_09: No, they're not.

SPEAKER_09: Okay.

SPEAKER_09: And none in the room?

Unknown: Not now.

SPEAKER_09: Then I think we will go ahead.

SPEAKER_09: The next item on the agenda is public, well, I guess we'll wrap that one up.

SPEAKER_09: So I want to say thank you to James and to our friends from Colorado for coming all the

SPEAKER_09: way out and presenting to us.

SPEAKER_09: Thank you so much.

SPEAKER_09: So now we will go on to the public comment for items, not on the agenda if we have any.

SPEAKER_09: I do not see any hands raised.

SPEAKER_09: Okay.

SPEAKER_09: So if we don't have any more comments, written comments that are received on items not on

SPEAKER_09: the agenda will be included in the record if received within two hours at the end of

SPEAKER_09: the meeting.

SPEAKER_09: And this last item on the agenda is to provide a summary of committee direction.

SPEAKER_08: Staff will collect information about large load policies from other utilities, how utilities

SPEAKER_08: are partnering with data centers including potential business models as well as potential

SPEAKER_08: rates and agreements to protect existing customers and bring that information back to the board

SPEAKER_08: for discussion at a future policy committee meeting.

SPEAKER_08: And does that include what John just suggested, the rate class?

SPEAKER_09: Yeah, that would be the...

SPEAKER_09: Okay, good.

SPEAKER_09: Yeah.

SPEAKER_09: That's what I thought.

SPEAKER_09: Thank you.

SPEAKER_09: Okay.

SPEAKER_09: Fantastic.

Unknown: So with that, I think we can wrap up the meeting.

SPEAKER_09: And we will now enter into a closed session to discuss the recruitment process of our

SPEAKER_09: new CEO since Paul's decided to leave us pursuant to Section 54957B1 of the Government Code.

SPEAKER_09: Thank you so much.

SPEAKER_09: Have a good night.