Latest / Elon Musk Podcast / Google AI Powers new 100-Hour Rust Battery

Transcript

Google AI Powers new 100-Hour Rust Battery · Elon Musk Podcast

• 0:00
  30 GW hours. That is the number on the table
• 0:03
  today. We're looking at AA300 MW
• 0:07
  battery system going up in Minnesota.
• 0:10
  But the power output isn't even the headline here, it's the
• 0:13
  duration. This system provides 100 hours
• 0:17
  of storage. Which is.
• 0:19
  Just total capacity, 30 GW hours.
• 0:21
  That makes it the largest battery project by energy
• 0:24
  capacity ever announced in the world.
• 0:26
  It is really hard to overstate the sheer scale of 30 GW hours.
• 0:29
  I mean, usually when we talk about grid batteries, we're
• 0:32
  talking about, you know, a lithium ion container sitting at
• 0:34
  a substation somewhere, and that's just designed to handle a
• 0:38
  quick spike in air conditioning use between maybe 5:00 and 9:00
• 0:41
  PM. This is an entirely different
• 0:43
  animal. And it's not for general grid
• 0:44
  storage either. No, exactly.
• 0:46
  This massive battery is specifically being built to
• 0:49
  power a new Google data center. They're partnering with Xcel
• 0:52
  Energy to handle the massive, just continuous electrical load
• 0:57
  coming from artificial intelligence.
• 0:58
  It really feels like a shift from simply, you know,
• 1:02
  supporting the public grid to building an entirely new kind of
• 1:05
  infrastructure for just one specific customer.
• 1:08
  It is. This represents a very physical
• 1:11
  answer to that AI energy crunch we keep hearing about.
• 1:13
  But what makes this specific case study so compelling is that
• 1:18
  it touches on three really distinct layers.
• 1:21
  First, you have the chemistry, which is radically different
• 1:23
  from what is sitting in your phone right now.
• 1:25
  Then you have the financial structure which involves Google
• 1:28
  paying for essentially everything.
• 1:31
  And then finally you have the local reality on the ground in
• 1:35
  Pine Island, MN where this the sci-fi technology literally
• 1:39
  meets small town zoning politics.
• 1:42
  So the question hanging over all of this for you listening is
• 1:45
  pretty straightforward. If we can actually build
• 1:47
  batteries that last for days instead of hours, and using
• 1:51
  materials as cheap as iron and air, does that finally solve the
• 1:54
  reliability problem for renewables?
• 1:56
  Or does it just create a grid where only the wealthiest
• 1:58
  companies in the world can afford for the lights to stay
• 2:01
  on? And to answer that, you really
• 2:03
  have to look at the pressure driving the deal in the 1st
• 2:05
  place. It isn't just that Google wants
• 2:07
  to be green for PR reasons. Right.
• 2:09
  It's operational. Highly operational data centers,
• 2:12
  especially the ones running large language models.
• 2:15
  Your Geminis, your TAT GPTS, they have a totally flat
• 2:19
  relentless demand profile. They run 24/7.
• 2:23
  And wind and solar obviously do not, right?
• 2:25
  The sunsets, the wind dies down. That mismatch is the fundamental
• 2:30
  problem of the whole energy transition, and until now the
• 2:34
  primary tool to fix it has just been lithium ion batteries.
• 2:37
  Which are literally everywhere. I mean they run our phones or
• 2:40
  EVs a huge chunk of our short term grid storage.
• 2:43
  And lithium ion is an incredible technology.
• 2:46
  It is highly power dense and it reacts instantly.
• 2:49
  If a cloud passes over a solar farm, a lithium battery creates
• 2:53
  a bridge in milliseconds to keep the grid frequency stable.
• 2:56
  Yeah, but financially and chemically, lithium ion just
• 2:59
  hits a hard wall right around 4 hours of duration.
• 3:02
  Which four hours gets you through the evening peak?
• 3:04
  Yeah, but it doesn't help you if a storm system just parks itself
• 3:08
  over the Midwest for three days. Exactly.
• 3:11
  And in the energy industry, there is a very specific German
• 3:15
  word for that exact scenario, dunkle float.
• 3:17
  Dunkle Float. Yeah, dunkle float.
• 3:19
  It translates roughly to dark doldrums.
• 3:23
  Dark doldrums. Right.
• 3:24
  These are periods where there is absolutely no sun and no wind
• 3:28
  for days at a time. So if you are a data center
• 3:31
  manager and your mandate is to run on 100% carbon free
• 3:37
  electricity. Which, by the way, is
• 3:38
  Minnesota's official state goal by 2040.
• 3:41
  So dunkle flout is the thing that keeps you up at night.
• 3:44
  Because if your battery dies after four hours and the
• 3:47
  forecast says there's no wind for another 40 hours, you have
• 3:50
  to fire up a natural gas plant. And that totally destroys your
• 3:53
  carbon free goal. So that is why the 100 hour
• 3:56
  duration is the specific answer to the dunkle flout problem.
• 3:59
  It's a bridge. Right.
• 4:00
  This battery system isn't designed for speed, it's
• 4:03
  designed for endurance. It bridges those multi day gaps
• 4:07
  without burning fossil fuels. OK, so let's actually look at
• 4:10
  the machine itself. The company making these is Form
• 4:13
  Energy. Yeah.
• 4:15
  And I was looking at their spec sheet and under chemistry it
• 4:18
  literally just lists iron air. It sounds almost too simple,
• 4:22
  because the fundamental principle is something we
• 4:24
  usually spend a lot of time and money trying to prevent.
• 4:28
  It relies on rusting. Hold on, we are trusting the
• 4:32
  reliability of the grid to Rust. In a highly controlled way, yes.
• 4:36
  The process is officially called reversible rusting.
• 4:39
  The battery uses iron, water and air.
• 4:42
  That's it. OK.
• 4:43
  When the battery is discharging, meaning it's sending power out
• 4:46
  to the data center, it takes in oxygen from the outside air and
• 4:49
  exposes the iron inside the cell to it.
• 4:52
  The iron turns into iron oxide. Rust and that releases energy.
• 4:55
  Exactly. That chemical reaction releases
• 4:58
  electrons. So the active rusting creates
• 5:00
  the electricity wait back up. How do you recharge a box of
• 5:03
  Rust? You reverse the current.
• 5:05
  When the wind is blowing hard and you have excess electricity
• 5:07
  on the grid, you push that power back into the battery.
• 5:10
  That incoming energy forces the oxygen to detach from the rust,
• 5:14
  turning it back into metallic iron, and then it just expels
• 5:17
  the oxygen back out into the air.
• 5:19
  It's essentially breathing. That is actually the best way to
• 5:22
  visualize it. It breathes in oxygen to
• 5:25
  discharge and it breathes out oxygen to recharge.
• 5:28
  I'm trying to picture the physical unit here.
• 5:30
  Is this just a giant tank of metal?
• 5:33
  It's modular. Each individual module is about
• 5:36
  the size of a side by side washer and dryer set OK, and
• 5:39
  inside that box you have stacks of about 50 cells, each 1A meter
• 5:44
  tall. They contain the iron anode, the
• 5:47
  air electrode and a water based electrolyte.
• 5:50
  That water based part seems really significant for safety
• 5:54
  because we constantly hear about lithium batteries Catching Fire
• 5:57
  or going at the thermal runaway. It is a major differentiator
• 6:00
  because the electrolyte is water based.
• 6:02
  It is completely non flammable. There is 0 risk of thermal
• 6:06
  runaway. Wow.
• 6:07
  You could technically shoot a hole in the side of this thing
• 6:09
  and it wouldn't explode. I mean it might leak water, but
• 6:11
  it will not burn. That has to make the permitting
• 6:13
  process so much easier compared to putting in a massive lithium
• 6:17
  installation 0. Absolutely.
• 6:19
  But if it's safer and it holds power longer, why isn't this
• 6:23
  exact chemistry in our electric cars right now?
• 6:26
  Wait, iron is incredibly heavy. You do not want an iron air
• 6:30
  battery in your car. It would weigh as much as the
• 6:33
  vehicle itself, right? But for a stationary grid
• 6:36
  battery just sitting in a field in Minnesota, weight is totally
• 6:40
  irrelevant. Cost is what matters.
• 6:42
  And Form Energy claims they can store energy at less than 110th
• 6:45
  the cost of lithium ion. 110th That is the economic trigger
• 6:49
  here. Lithium ion packs cost somewhere
• 6:51
  around 130 to $150.00 per kWh. OK, Form is targeting closer to
• 6:57
  $20. That is a massive.
• 6:59
  Gap. It entirely comes down to the
• 7:00
  materials. Iron is the most abundant metal
• 7:03
  on earth. We know exactly how to work with
• 7:04
  it and it's extremely cheap. We aren't mining cobalt in
• 7:08
  unstable regions. Yeah, the entire supply chain is
• 7:11
  domestic. In fact, these specific
• 7:13
  batteries are being manufactured right now in Weirton, WV.
• 7:17
  That's form factory 1. Right, right.
• 7:18
  And the location really tells the story.
• 7:21
  Weirton used to be a massive steel town.
• 7:24
  Form Energy actually built their factory right on the side of a
• 7:27
  defunct steel mill, so they're using the legacy workforce and
• 7:31
  the existing infrastructure of the old steel industry to build
• 7:35
  the storage medium for the renewable age.
• 7:37
  That is a fascinating transition.
• 7:39
  So we have a battery that is cheap, safe, and last for 100
• 7:42
  hours, but 30 GW hours of storage is still a massive
• 7:47
  infrastructure project. Usually when a utility builds
• 7:50
  something this big, the cost just gets passed right on to the
• 7:53
  ratepayers. And this is where the so-called
• 7:56
  Minnesota model comes into play. The regulatory framework here is
• 7:59
  honestly just as innovative as the chemistry.
• 8:02
  It's being called the Clean Energy Accelerator Charge, or
• 8:05
  CEAC. Clean energy accelerator charge.
• 8:09
  How does that actually differ from a normal utility rate hike?
• 8:12
  Well, in a traditional utility model, if Xcel Energy builds a
• 8:15
  new power plant, they take that cost, they add a Guaranteed Rate
• 8:18
  of return, and they just spread it out across everyone's
• 8:21
  electric bill. You, me, the local bakery, we
• 8:24
  all pay a slice. Which is standard since we all
• 8:26
  ostensibly use the grid, but this project is explicitly for
• 8:30
  Google. Exactly, and that is why the
• 8:33
  Cesc is so different. Google is paying all the costs
• 8:37
  for this service. All of them.
• 8:38
  All of them. They're paying for the wind
• 8:40
  farms, the solar panels, the form energy batteries, and this
• 8:44
  is critical, the grid infrastructure upgrades required
• 8:48
  to actually hook it all up. So Excel Energy is claiming that
• 8:51
  residential bills will not go up 1 cent to support this growth.
• 8:54
  That is the promise. Excel has been very clear that
• 8:57
  the residential customer pays nothing extra for this specific
• 9:00
  project, and this connects directly to the broader
• 9:03
  ratepayer protection debate we are seeing nationally right now.
• 9:06
  Yeah, we heard about this recently in the State of the
• 9:08
  Union. President Trump announced A
• 9:10
  ratepayer protection pledge, basically demanding that data
• 9:13
  centers pay their own way so they don't drive up prices for
• 9:16
  regular families. Right.
• 9:18
  And it is a very politically popular stance.
• 9:20
  You have these trillion dollar tech companies coming into town
• 9:24
  and demanding literal gigawatts of power.
• 9:28
  It is completely natural to worry that regular people are
• 9:31
  going to foot the bill. But there is some disagreement
• 9:34
  on whether a Plex like that is even necessary.
• 9:37
  Yes, Brian Janis, who used to lead energy at Microsoft, argued
• 9:41
  recently that the pledge is, quote, meaningless.
• 9:43
  Meaningless because the data centers are already paying.
• 9:47
  His argument is that industrial customers have always had
• 9:50
  entirely different rate structures than residential
• 9:53
  ones. Data centers have generally
• 9:55
  always paid for their own load. Janus is basically saying we
• 9:58
  were already going to pay for this anyway.
• 10:00
  So Google might look at the pledge and say it's completely
• 10:02
  redundant. Right.
• 10:03
  But the Minnesota deal formalizes it in a new way.
• 10:07
  It creates a highly specific tariff that totally isolates the
• 10:11
  cost. So just to reset here, Google
• 10:13
  gets the clean power and the reliability they need for AI,
• 10:17
  Excel gets a massive upgrade to their grid paid for by a
• 10:20
  corporate client, and the residential customer
• 10:23
  theoretically sees no change in their monthly bill.
• 10:26
  That's the deal on paper. I want to look at the sheer
• 10:29
  scope of what Google is actually buying here.
• 10:31
  We talked about the battery, but what is charging it?
• 10:33
  The agreement involves 1.9 gigawatts of new clean energy
• 10:38
  total. So you have the 300 megawatts of
• 10:41
  storage we discussed, but that is paired with 1400 megawatts of
• 10:45
  wind and 200 megawatts of solar. 1400 That is a lot of wind.
• 10:50
  Minnesota is a very wind rich state, but there is another
• 10:54
  detail in the deal that is really worth highlighting.
• 10:56
  It's a $50 million investment in something Excel calls Capacity
• 11:01
  Connect. Capacity Connect.
• 11:03
  What does that money actually go toward?
• 11:05
  It's a distributed energy program.
• 11:07
  Instead of putting all the batteries in one giant field
• 11:09
  strictly for Google, this money goes toward putting utility
• 11:12
  owned batteries at local businesses and industrial sites
• 11:15
  throughout the grid. So local factory just gets a
• 11:17
  battery on site. Yes, the idea is to help grid
• 11:20
  reliability at the edges, not just at the central data center.
• 11:24
  It's essentially A sweetener for the broader grid.
• 11:27
  Let's zoom in on where this giant battery and data center
• 11:30
  are actually going. Pine Island, MN.
• 11:32
  Yeah, a small city population is less than 4000.
• 11:36
  It's roughly an hour South of the Twin Cities.
• 11:39
  And before this, their biggest claim to fame was holding a
• 11:42
  world record for a 6000 LB block of cheese on a railroad flat
• 11:47
  car. Which is a very, very different
• 11:49
  kind of heavy industry than artificial intelligence.
• 11:52
  Just a little bit. But clearly, not everyone in
• 11:55
  Pine Island is happy about this transition.
• 11:57
  No, they aren't. There is a Facebook group called
• 11:59
  Stop the Pine Island Data Center that currently has over 600
• 12:02
  members and in a town of 4000 that is a significant percentage
• 12:06
  of the local population. What are their specific
• 12:08
  concerns? It's the usual friction you see
• 12:11
  with data centers, mostly water usage and noise.
• 12:14
  Data centers need massive amounts of water for cooling.
• 12:17
  Even with closed loop systems, the overall consumption is high.
• 12:21
  Locals are worried about drawing down the local aquifer.
• 12:24
  And then there is the noise from the cooling fans.
• 12:26
  Yeah, we've seen reports from Ohio, where residents living
• 12:29
  near data centers complain about just a constant, inescapable
• 12:32
  hum. Right.
• 12:34
  And while this specific project avoids the air pollution
• 12:37
  concerns you normally get with diesel backup generators, since
• 12:41
  they are using the iron air batteries instead, the physical
• 12:44
  industrial footprint is still massive.
• 12:46
  You are turning quiet farmland into a high tech industrial
• 12:50
  park. It's that classic tension
• 12:52
  between wanting all these digital services to work
• 12:55
  instantly but not wanting the physical infrastructure next
• 12:58
  door. Exactly.
• 12:59
  However, from a political standpoint, this project is
• 13:02
  being held up nationally as the model.
• 13:05
  Former Energy Secretary Jennifer Granholm specifically praised
• 13:09
  this exact deal. Really.
• 13:10
  Yeah. She called it the prime example
• 13:12
  for how sighting should be done. Just because of the Google
• 13:14
  funding structure. That and the legislative
• 13:17
  groundwork that made it possible.
• 13:18
  Minnesota passed laws in 2025 that set very strict
• 13:22
  environmental guardrails for these types of projects.
• 13:25
  Granholm's point is that you can't just let a massive build
• 13:28
  out like this happen in a vacuum.
• 13:30
  You need state legislation that protects the environment 1st,
• 13:33
  and then you need a utility willing to actually create a
• 13:35
  contract structure like the CEAC to protect the race.
• 13:40
  So let's pull back and look at what all of this means for the
• 13:43
  energy transition as a whole. We have the world's largest
• 13:46
  battery by energy capacity, made of iron and air, funded entirely
• 13:51
  by Google to basically bypass the limitations of the public
• 13:55
  grid and weather intermittency. It essentially proves that you
• 13:58
  can get firm 24/7 carbon free power if you are willing to move
• 14:02
  beyond lithium ion. That is the key take away here.
• 14:06
  It feels like the Minnesota model shows that Big Tech can
• 14:09
  actually act as a catalyst. They aren't just buying paper
• 14:12
  credits to offset their carbon footprint anymore.
• 14:15
  They're funding the hard physical hardware that makes the
• 14:18
  grid reliable. That is the major shift.
• 14:20
  For years, companies would just buy renewable energy
• 14:23
  certificates. They'd pay for a wind farm in
• 14:25
  Texas to technically offset set the emissions of a data center
• 14:28
  in Virginia. It was purely accounting.
• 14:31
  Yeah, spreadsheet exercise. Exactly.
• 14:33
  But this deal is about physics. It's about Google saying I need
• 14:37
  actual power here in this spot right now, regardless of the
• 14:40
  weather, and I will pay for the infrastructure to make that
• 14:43
  happen. But if this technology works at
• 14:46
  this scale, if these iron air batteries do exactly what they
• 14:49
  say they can do, does that finally kill the argument that
• 14:52
  renewables are too unreliable? It weakens it significantly.
• 14:56
  The main argument against a purely renewable glid is always
• 14:59
  intermittency. But if you can cost effectively
• 15:02
  store 100 hours of energy, you simply don't need a gas peaker
• 15:06
  plant. You just have four days of
• 15:07
  backup sitting in a box. Right.
• 15:09
  The catch, of course, is the price tag.
• 15:12
  Google can afford to be the first mover and pay the premium
• 15:14
  to build this out. Can a municipal utility in a
• 15:17
  small town afford to do the same thing?
• 15:19
  That is the risk. Does this technology get cheap
• 15:22
  enough for everyone or do we create a 2 tiered grid, 1 tier
• 15:27
  for the googles of the world with these private bulletproof
• 15:29
  green energy islands, and another tier for the rest of us
• 15:33
  relying on an aging grid that flickers every time the wind
• 15:36
  dies down? That is the real test of the
• 15:38
  Minnesota model. We'll be watching to see if the
• 15:40
  technology actually trickles down or just stays at the top.
• 15:44
  If you're not subscribed yet, take a second and hit follow on
• 15:46
  whatever app you're using. It helps us keep making this.
• 15:49
  We appreciate you being here.