Epilogue: The Age of Abundance
Epilogue
Leg When people talk about the AI revolution, they obsess over two things: chips and models. They debate whether NVIDIA will maintain its dominance. They argue about whether OpenAI or Anthropic or xAI will win the model race. In this book, we talk deeply about technologies like FSD and Optimus. And yet, we completely miss the third leg of the stool. Energy. The most overlooked aspect of the entire AI story. And the reason I think it gets overlooked is because energy feels boring. It feels like infrastructure. It feels like something your utility company handles while you focus on the exciting stuff. But I want you to understand something. For the next 10 to 20 years, we are going to see a massive demand for energy generation unlike anything we have ever witnessed. Solar. Natural gas. Geothermal. Nuclear. Everything. Because AI needs unbelievable amounts of power. Unbelievable. And this creates an opportunity that almost nobody is paying attention to.
The AI Energy Appetite
Right now, a single data center training a large language model can consume
as much electricity as a small city. I am not exaggerating. When you hear about companies like xAI building out massive GPU clusters - we are talking about hundreds of thousands of chips - each of those chips is pulling power. And then you need to cool them. The cooling alone can account for 40 percent of the total energy consumption. And this is just training. Inference - the actual running of these AI systems at scale - that is going to require even more power over time as billions of people start using AI for everything. Every query to ChatGPT or Claude or Grok requires compute. Every video generated by AI. Every Agent running tasks for you as you personal assistant. Every car and robot pinging the cloud for information, context, knowledge, date… Multiply that by billions of queries per day, per hour, per minute… and you start to see the problem. Most people look at this and think, "That sounds like a problem for AI companies to solve." But energy and AI have become the same conversation. They're inseparable now. A few years ago Elon Musk said that he thinks the energy side of the Tesla business will eventually be as big as the car side of the business. Maybe even bigger. And the AI demand is going to be by far one of its biggest drivers. I think people heard that and sort of filed it away as Elon being Elon. But I actually think he was underselling it.
Why Batteries Change Everything
Here is where I want to introduce an idea that most people do not fully
appreciate. Most of the incremental energy production in the United States is going to come from batteries + solar. Jesse Peltan (@JessePeltan on X) has amazing research on this that has been a cornerstone of this thesis. What do I mean by that? You can essentially double the grid capacity by smartly using battery storage. You do not need to build new power plants
(although we should build more). You do not need new transmission lines
(although we should build more). You just need batteries. The reason this works is because our current grid is wildly inefficient. We build power plants to handle peak demand - those hot summer afternoons when everyone runs their air conditioning at the same time. But most of the 24 hour cycle, those plants are sitting there underutilized. We have all this capacity that just goes to waste because the grid has no memory. Electricity gets generated, and if nobody uses it right then, it is gone. This is insanely inefficient. Now imagine you have massive battery installations throughout the grid. During off-peak hours, when electricity is cheap and plentiful, the batteries charge up. During peak hours, when demand spikes, the batteries discharge. You have effectively added capacity to the grid without building a single new power plant. Tesla's Megapack business is built around exactly this concept. Each Megapack can store enough energy to power around 3,600 homes for an hour. But the real magic is when you deploy them at scale. Tesla Energy deployed 46.7 gigawatt-hours in 2025, which was roughly 50 percent higher than the year before. And this is still the early innings. The gross margin on the energy business is absolutely crushing it compared to automotive. We are talking about 31.4 percent margins on energy versus 16.1 percent on cars in Q4 2025. The energy business is actually more profitable on a percentage basis than selling cars. And it is growing faster. This is a massive signal - the demand is there. Now, it’s all about getting it installed everywhere.
Wright's Law and the Cost Curve
Let me talk about something called Wright's Law, because it applies to
batteries in a huge way.
Wright's Law is this observation that costs decline predictably as cumulative
production increases. It was originally discovered in the aircraft industry in the 1930s - every time cumulative production of airplanes doubled, the cost to produce them fell by about 15 percent. The same pattern shows up in solar panels. In semiconductors. In batteries. For lithium-ion batteries specifically, costs have fallen by around 20 to 30 percent for every doubling of cumulative production. And this pattern has held pretty consistently over the past two decades. Battery pack costs have dropped from over $1,000 per kilowatt-hour in 2010 to around $115 today. Some projections have us hitting $80 or even lower by the end of the decade. Why does this matter? Because as battery costs come down, the economics of grid storage become overwhelming. You can store solar energy generated during the day and use it at night. You can store wind energy from gusty days and deploy it when the air is calm. The intermittency problem that everyone complains about with renewables just goes away. Bye bye. And Tesla - once again, surprise surprise - is in the middle of all of this. They are not just building cars. They are building the energy infrastructure that will power the AI revolution. Every Megapack they deploy makes the economics better for the next one. Every gigawatt-hour of production brings costs down for all the gigawatt-hours that follow. This is the flywheel in action yet again. The Convergence in plain sight. AI needs energy. Energy needs storage. Storage needs batteries. Battery production brings costs down. Lower costs mean more storage. More storage means more renewable energy is viable. More renewable energy means cheaper power for AI. Cheaper AI means more AI can be used. More AI used means better storage technology. And round it goes.
China Is Actually Doing It
Now the elephant in the room - China. China is crushing us in energy deployment. And the reason is simple: they are actually doing it. They have the same technology we do.
The United States has all this potential. We have deserts in the Southwest
that get more solar radiation than almost anywhere on Earth. We have massive wind corridors through the middle of the country. We have the technology to deploy it all. And we just do not. Meanwhile, China installed more solar capacity in 2023 than the entire rest of the world combined. They are deploying batteries at a pace that makes our efforts look like a rounding error. Their electric grid is being transformed while ours is held together with duct tape and prayers. And the reason is politics. It is the oil lobby. It is utility companies that make money by building expensive power plants, not by deploying cheap batteries. It is NIMBYs who fight every solar farm and every wind turbine. It is regulators who move at geological timescales while technology moves at internet speed. I find this incredibly frustrating because it is not a technology problem. We have the technology. Solar panels are cheap. Batteries are getting cheaper by the day. The physics works. The economics work. The only thing that does not work is the political will. Solar is the obvious winner here. Stick solar panels and batteries in deserts and you have essentially free energy. The sun shows up every day. The land is not being used for anything else. The transmission problem is solvable. The whole thing is straightforward - the only obstacle is political will. Why can’t we do this? It’s so simple.
The Nuclear Question
I need to address nuclear because I know some people are wondering why I
am so focused on solar and batteries when nuclear seems like the obvious answer to AI's energy demands. I am not anti-nuclear. Nuclear power has incredible energy density. A single nuclear plant can produce massive amounts of electricity with zero carbon emissions during operation. France gets about 70 percent of its electricity
from nuclear and has some of the cleanest power in Europe as a result. There is clearly a role for nuclear in the energy mix. But building a new nuclear plant in the United States takes 15 to 20 years when you factor in permitting, construction, and regulatory approval. Now to the Trump administration’s credit, they are attempting to massively change this. But the cost overruns are legendary. Vogtle Units 3 and 4 in Georgia came in at more than double the initial budget. And there is the waste problem that nobody has solved to anyone's satisfaction - at least for now. The small modular reactor crowd is excited about new designs that could theoretically be faster and cheaper to build. And maybe they are right. I hope they are right. But we have been hearing about the nuclear renaissance for 20 years and I am still waiting to see meaningful deployment at scale. Meanwhile, solar and battery costs are dropping every single year. The permitting is simpler. The construction is faster. The modularity means you can deploy incrementally rather than betting everything on a massive multi- decade project. That said, things are shifting. The current administration is pushing hard to accelerate nuclear permitting and construction. There's particular interest in co-locating nuclear plants directly with AI data centers - Microsoft, Google, and Amazon have all announced deals to power their AI operations with nuclear energy. When a data center can be built right next to its power source, you eliminate transmission losses and grid bottleneck issues. This makes sense for the near-term. AI data centers need dense, reliable baseload power that doesn't depend on weather. Nuclear provides that in a way solar can't match for a 24/7 operation. If the regulatory environment actually changes - and there are signs it might - we could see nuclear playing a bigger role than I expected in powering the AI buildout. This would be amazing. But my first-principles take: in the long run, the Sun wins. All the energy we use ultimately comes from the Sun anyway, whether that's fossil fuels (ancient solar energy stored in organic matter), wind (driven by solar heating), or
direct solar. The Sun delivers more energy to Earth in one hour than humanity uses in a year. Solar is getting cheaper every year. Nuclear isn't. From a 50-year view, betting on the Sun seems obvious. Elon has been talking about this for how long now? It’s because he’s very obviously right. We should listen to him when it comes to this. So when I think about what is going to actually meet AI's energy demands in the next 10 years, I think it's going to be a combination - nuclear for baseload where it can be permitted quickly, existing fossil fuel plants running at 100%, and solar and batteries for everything else. The near-term bottleneck is going to be solved by whatever energy technologies can actually be deployed at the pace the AI revolution demands. And increasingly, AI companies are taking energy into their own hands rather than waiting for the grid to catch up.
The Home Energy Revolution
I want to zoom out for a second because I have been talking a lot about grid-
scale energy storage. But there is another piece of this puzzle that I think is equally important: home energy. Tesla's Powerwall product has been around for almost a decade now. It started as kind of a novelty - a cool gadget for early adopters who wanted to store solar energy or have backup power during outages. But it has evolved into something much more significant. A Powerwall can store 13.5 kilowatt-hours of energy. That is enough to power most homes for half a day or more during normal usage. When you pair it with solar panels, you can essentially go off-grid during the day and into the evening. You charge up during peak sun hours and draw down through the night. But the really interesting thing is what happens when you connect thousands or millions of these home batteries together into a virtual power plant. Tesla has been doing this in places like California and Texas. When the grid is under stress, Tesla can coordinate all those Powerwalls to discharge
simultaneously, providing emergency power to the grid. The homeowners get paid for participating. The grid gets stabilized. Everyone wins. This is a completely different model than the traditional centralized power plant approach. Instead of a few massive installations, you have millions of small installations distributed throughout the grid. The resilience is built in because there is no single point of failure. And the capital investment is distributed across individual homeowners who are motivated by their own energy bills and backup power needs. I think this model is going to scale dramatically over the next decade. Every new home should have solar panels and a battery as standard equipment. The payback period is already attractive in most markets, and it gets better every year as costs come down and electricity prices go up. The implications for the grid are profound. If every home can store and discharge energy on demand, you fundamentally change the dynamics of supply and demand. Peak demand becomes manageable. Renewable intermittency becomes a non-issue. The entire system becomes more flexible and resilient. And Tesla is positioning itself at the center of all of this. They make the cars that can charge from home solar. They make the home batteries that store the excess. They make the grid-scale batteries that balance everything out. They make the solar panels that generate the power in the first place. It is all connected.
The Space Solar Opportunity: SpaceX
And now we come to the real, long term solution for AI’s energy needs, which is directly correlated to the SpaceX acquisition of xAI on February 2nd, 2026. Shout out to Elon for his amazing timing for closing the deal within 6 days of the final manuscript submittal for this book. You really did me a solid there, pal.
The way I see it, there is increasingly a convergence between SpaceX and
Tesla. If the future is solar powered AI satellites - which I think it pretty
much needs to be if you want to harness a non-trivial amount of the energy of the sun - you have to move to solar powered AI satellites in space. This is a confluence of Tesla expertise in batteries and solar, SpaceX expertise in launching stuff to orbit cheaply, and xAI expertise in building AI systems that can run autonomously. I know that sounds like science fiction. But there is a path to putting 100 gigawatts per year of solar powered AI satellite into orbit and having this actually be the lowest cost way to power and operate AI at a very large scale. Why else do you think SpaceX and xAI merged in one of the biggest acquisitions in history? It’s because the economics make sense. There’s a business case. In space, there is no atmosphere to block the sun. No night. No clouds. No seasons. Solar panels in space collect six to eight times more energy than panels on Earth. And if you are running AI inference in orbit, you do not need to transmit the electricity down to Earth at all. You run the compute right there in space and just send down the results. The challenge has always been that getting stuff to space is expensive. But this is exactly what SpaceX has been solving for two decades. They've already brought launch costs down by an order of magnitude. Starship, their newest rocket set to go into operation in 2026, promises to bring them down by another order of magnitude - potentially to $10 per kilogram to orbit, which is roughly the cost of air freight today. At those economics, the calculus changes completely. Suddenly it becomes cheaper to put your data center in space than to build it on Earth and deal with cooling, grid connections, permitting, and all the rest of it.
Let me walk through why SpaceX's role is so critical to this vision: Launch Cost Reduction: SpaceX has already proven they can achieve cost curves that seemed impossible. Falcon 9 reusability brought costs down
dramatically. Starship - fully reusable, rapidly relaunchable, massive payload capacity - will bring them down again. Every dollar per kilogram reduction makes space-based infrastructure more viable. Operational Proof with Starlink: SpaceX isn't theorizing about large-scale satellite constellations. They've deployed over 9,000 Starlink satellites. They've built the manufacturing, the launch cadence, the ground stations, the customer operations. They know how to deploy and operate massive space infrastructure. This is the "vessel" proving itself before the bigger mission. The xAI Connection: Think about what xAI actually needs. Massive compute for training and inference. Compute needs massive energy. Earth-based data centers are hitting grid constraints. But xAI's parent ecosystem includes a company that can put things in space cheaply and another company (Tesla) that builds solar panels and batteries. The architecture writes itself: SpaceX launches solar-powered AI compute infrastructure. Tesla provides the power systems. xAI provides the AI workloads. Starlink provides the connectivity back to Earth. The entire ecosystem converges on solving the energy constraint for AI. The Cooling Advantage: This is something most people miss. Data centers spend enormous resources on cooling. In space, cooling is free - you radiate heat directly into the void. No air conditioning. No water cooling. No heat management infrastructure. This is a massive efficiency gain that further tilts the economics toward space-based compute. It’s a technology that needs to be solved, but the economics will force the solution. It’s too good to pass up. I am not saying this happens in 2027. But in the four or five year timeframe, the lowest cost way to do AI compute might actually be with solar powered AI satellites as Elon claims. And only one ecosystem has all the pieces: the AI (xAI), the launch capability (SpaceX), and the power systems (Tesla). Talk about The Convergence. If you are not thinking about this, you are not thinking about the full picture. This isn't three separate companies pursuing separate goals. This is one integrated system building toward a future where AI runs on unlimited solar
energy, deployed by rockets that cost less than planes, coordinated by a global satellite network. That's the Musk ecosystem thesis. That's why I think the individual company valuations miss the point entirely.
The Political Obstacles
So why is not everyone piling into energy? Why is this the "forgotten" third
leg of the AI stool? The answer is vested interests. The oil and gas industry has spent decades building a regulatory and political infrastructure designed to protect its position in the US. It’s about as obvious as the sky is blue. Utility companies have business models that reward capital expenditure on big centralized power plants, not distributed battery storage. Politicians on both sides take money from energy incumbents who have no interest in being disrupted. And then there is the environmental movement, which ironically has become one of the biggest obstacles to clean energy deployment. Every solar farm gets fought by someone who thinks it will hurt the desert tortoise. Every wind farm gets fought by someone who thinks it will kill birds. Every transmission line gets fought by someone who does not want it in their viewshed. The result is paralysis. We have the technology to transform our energy system. We have the economics working in our favor. We have the existential need with AI demand exploding. And we cannot get out of our own way. I think this is one of the most important policy questions of the next decade. Not "should we build more solar?" The answer to that is obviously yes. The question is "how do we overcome the political obstacles that prevent us from building the energy infrastructure we need?" And honestly, I do not have a great answer. The forces aligned against change are powerful and entrenched. The permitting process is designed to give everyone a veto. The incentives are all messed up.
But I do know this: the AI revolution is going to demand more energy than
we have ever produced before, by a wide margin. And the countries that figure out how to generate that energy cheaply and cleanly will win the next century. Right now, that looks more like China than the United States. And that should scare us.
The Tesla Energy Opportunity
I want to bring this back to the investment perspective because I think there
is a real opportunity here that the market does not fully appreciate. When most people think about Tesla, they think about cars. The Model 3. The Model Y. The Cybertruck. FSD. Robotaxi. These are the stories that dominate the headlines and the analyst reports. But Tesla Energy is quietly becoming a monster business in its own right. Think about the numbers I mentioned earlier. 31.4 percent gross margin on energy versus 16.1 percent on automotive. More than doubling year over year production. And this is happening while the company is primarily focused on scaling the car business. Now imagine what happens when Tesla decides to really lean into energy. When they start building Megapack factories at the same scale they build car factories. When they leverage their battery expertise and manufacturing know-how to drive costs down even further. When they combine energy storage with their solar business and their vehicle fleet to create an integrated energy ecosystem. A few years ago I would have said the market underestimates Tesla's AI potential. Today I would add that the market also underestimates Tesla's energy potential. And the two are connected because the AI revolution is going to create the demand that makes energy storage the biggest business opportunity of the next two decades. The energy side of the business will eventually be as big as the car side. Maybe even bigger. And the AI demand is going to be by far one of its biggest drivers.
The Grid Infrastructure Problem
I want to spend a moment on something that does not get enough attention: the grid itself. You can have all the solar panels and batteries in the world, but if you cannot move electricity from where it is generated to where it is needed, none of it matters. And our grid infrastructure in the United States is embarrassingly outdated. The American electrical grid was built primarily in the mid-20th century. It was designed for a world where big centralized power plants sent electricity one way to consumers. It was not designed for a world where millions of small generators - rooftop solar panels, home batteries, electric vehicles - are both consuming and producing electricity. It was not designed for the kind of dynamic, bidirectional flows that a modern renewable grid requires. Upgrading this infrastructure is going to require massive investment. We are talking about new transmission lines to move solar power from sunny regions to population centers. We are talking about smart grid technology that can manage variable generation and demand in real time. We are talking about interconnections between regional grids that are currently operated almost like separate systems. The Department of Energy has estimated that we need to expand transmission capacity by 60 percent by 2030 just to meet current clean energy goals. And that was before the AI energy demand story really took off. Factor in the data center buildout we are seeing, and the transmission gap is even larger. But here is where I get frustrated. Building transmission lines is harder than it should be. Every line has to cross multiple jurisdictions. Every jurisdiction has different permitting requirements. Environmental reviews can take years. NIMBY opposition can kill projects entirely. I have seen estimates that a major transmission project in the United States takes 10 to 15 years to complete. In that time, China will have built out entire regional grids. The pace differential is staggering.
And this matters because you can deploy all the solar panels and batteries you
want at the generation point, but if you cannot move that electricity to where AI data centers are being built, you have not solved the problem. Energy storage helps because it gives you temporal flexibility. But you still need the wires. I do not have a good solution to this problem. It is fundamentally a political and regulatory challenge yet again, not a technology challenge. But I think more people need to understand that the grid bottleneck is real and it could significantly constrain how fast we can scale AI in the United States.
The Convergence Continues
I want to tie this back to the broader thesis of this book. In Chapter 1, I
talked about how AI, robotics, and energy are not separate revolutions but one interconnected system that amplifies itself exponentially. Energy is the third leg of that stool, and it is arguably the most overlooked. Without abundant cheap energy, you cannot train massive AI models. Without AI, you cannot optimize energy grids to deploy renewables at scale. Without batteries - manufactured increasingly by robots - you cannot store intermittent energy sources. Each piece enables the others. I hope by now you can see why Tesla is the perfect use case for driving home the concept of The Convergence. They make cars that are essentially batteries on wheels. They make home batteries and grid-scale batteries. They make solar panels. They are building the AI systems that will optimize all of this. And increasingly, they are building the robots that will manufacture all of it. None of this is coincidence. Elon sees the whole picture and is positioning the company to capture value from all of it. Cars and trucks were just the beginning. Energy is the foundation. AI is the brain. And robotics is the labor force that builds it all.
The Coming Energy Shortage
Now, I want to paint a somewhat darker picture for a moment because I
think it is important to be honest about the challenges ahead. The AI buildout is happening faster than our energy infrastructure can keep up. Every major tech company is racing to build data centers. NVIDIA cannot produce chips fast enough. And all of those chips need power. We are already seeing signs of stress. In some regions, utilities are struggling to provide enough power for new data center developments. There have been reports of tech companies being told they cannot get grid connections for years because there simply is not enough capacity. Some companies are building their own power generation facilities - natural gas plants, in many cases - just to guarantee the power they need. We’ve heard reports of electricity prices going up in some areas because of the insane demand from AI Hyperscalers - your OpenAIs, xAIs, etc. This is not how it should work. We should not have trillion-dollar companies building redundant power infrastructure because the grid cannot serve them. It is wasteful. It is inefficient. And it often means falling back on fossil fuels because they can be deployed faster than clean alternatives. I think there is a real risk that the AI revolution gets constrained by energy availability. Not by chip supply. Not by model capability. Just raw electricity. The companies that figure out how to secure reliable, cheap power are going to have a massive advantage. This is why I think Tesla's positioning in energy is so strategic. If you control both the AI workload and the energy to power it, you have a vertically integrated advantage that is very hard to compete with. And the countries that build out energy infrastructure faster are going to capture more of the AI value chain. Right now, that looks like China is winning. They are building power capacity at a pace we cannot match. Their data centers are coming online while we are still arguing about permits.
I do not say this to be pessimistic. I say it because I think we need to wake up to the urgency of the situation. The energy bottleneck is real. It is going to get worse before it gets better. And unless we dramatically accelerate our energy buildout, we risk ceding leadership in AI to countries that move faster.
What This Means For You
If you are reading this book, you probably already understand that the world
is changing fast. And if you didn’t - welcome. Sorry for ruining your day. What I want you to take away from this chapter is that energy is central to everything that is about to happen. Absolutely central. For investors, this means paying attention to the energy transition in a way that goes beyond the usual "green energy" narrative. The opportunity here is recognizing that AI is going to create demand for energy that our current infrastructure cannot meet. The companies that solve this problem will be worth trillions. For policymakers - and I know some of you will read this - the message is even more urgent. We are in a race. China is building the energy infrastructure of the future while we argue about permits. The decisions we make in the next few years about energy policy will determine whether the United States leads the AI revolution or watches it happen from the sidelines. And for everyone else, the message is this: energy abundance is coming. It might take longer than it should because of political obstacles. But the technology is there. The economics work. The demand from AI is going to force the issue whether we like it or not. The question is not whether we get to energy abundance. The question is how we get there - quickly and deliberately, or slowly and chaotically. The answer to that question will shape the next half century. We have the technology. We have the economics. We have the need. What we lack is the will.
And here is what keeps me up at night. The AI revolution will not wait for us
to figure this out. It is happening now. Every day we delay on energy is a day we fall behind China in AI. Every permit that takes three years instead of three months is ground we cede to competitors who move faster. Every solar farm blocked by environmental lawsuits is compute capacity we do not have when we need it. Energy abundance is not some nice-to-have for the future. It is the foundation on which everything else in this book depends. Without it, the AI systems that could cure diseases and solve problems we cannot imagine stay hypothetical. Without it, the abundance I have been describing stays theoretical while scarcity remains our reality. All the while, China will continue its march towards it global dominance, at which point they’ll be free to propagate their world view on the rest of the world, regardless of how any of us feel. Is that what we want? The technology exists. The economics work. The only question is whether we have the courage to build it. I believe we do. But belief is not enough. The next decade will tell us whether we chose abundance or whether we let it slip away.