Intelligence Goes Physical

The week AI stopped being a software story and started rewriting the physical world

TLDR: Three exponential curves—AI cognition, robotic embodiment, and energy infrastructure—hit their inflection points simultaneously this week. Humanoid robots crossed the $5,000 price barrier and moved into real, unmodified homes. AI models just scored 94% on PhD-level science questions. And America’s electric grid is being asked to absorb a 7x surge in demand to power it all. The world is changing faster than our institutions can absorb…

---

I’ve been saying for years that exponential technologies don’t just advance – they converge. And when they converge, the world doesn’t merely evolve. It transforms.

This week, convergence is advancing along three domains simultaneously: artificial intelligence, robotics, and energy. Each story, taken alone, would be headline-worthy. Taken together, they tell a single story about the decade we’re entering.

Here’s what happened, why it matters, and what you need to do about it.

The Models Are Touching the Ceiling of Human Knowledge

This past week delivered what may be the most significant AI capability cluster in a single seven-day window ever. Let me walk you through it.

First: Google launched Gemini 3.1 Ultra with a 2-million-token context window: the largest in any production model today. To put that in perspective, you could feed this model the entire works of Shakespeare, the complete U.S. tax code, and your company’s last decade of emails, and it would read, reason, and respond across all of it simultaneously. The system uses a proprietary TurboQuant algorithm to compress the memory overhead that would otherwise make this computationally impossible.

Next came Anthropic’s Claude Mythos, released in restricted access to just 50 organizations. The numbers are wild: 93.9% on SWE-bench Verified (the gold standard software engineering benchmark) and 94.6% on GPQA Diamond, which tests PhD-level questions in biology, chemistry, and physics. Think about that. A machine is now answering graduate-level science questions with near-perfect accuracy. Mythos also independently identified thousands of zero-day security vulnerabilities across major operating systems and browsers. We are watching AI systems develop capabilities that were, until very recently, the exclusive domain of the world’s top human experts.

Finally, Meta entered the ring with Muse Spark, their first proprietary large model since the $14 billion acquisition that brought Alexandr Wang in as chief AI officer. The race isn’t just between three companies anymore… it’s between three philosophies of what AGI should look like and who should control it.

So, what does this mean? We’ve crossed from AI as assistant to AI as expert. The question is no longer whether AI can help you think. It’s whether you are changing the way you think… Are you going AI first?Are you allowing yourself to think bigger than ever before in an uncapped fashion with AI as your cognitive partner?

The Robots Are Coming! The Robots Are Coming! And a $4,900 Robot Just Moved Into the House Next Door

My friend Masayoshi Son has been funding the humanoid robotics revolution for years. This week, the revolution showed up on AliExpress.

Unitree’s R1 humanoid robot is now available for pre-order at $4,900: a 99.4% price collapse from where the category stood just two years ago. Unitree shipped 5,500+ units in 2025 and projects 10,000 to 20,000 units in 2026. Meanwhile, Shenzhen launched the world’s first automated humanoid robot production line, capable of producing one robot every 30 minutes at annual capacity exceeding 10,000 units.

And then there’s UniX AI’s Panther, which completed what the company is calling the first real-home deployment of a mass-produced humanoid robot. Not a lab. Not a controlled environment. An actual, unmodified household. The third-generation Panther woke users up, made beds, prepared breakfast, performed whole-home cleaning, and organized objects. We’re not watching a demo reel. We’re watching the first chapter of the home robot era.

This isn’t robotics anymore. It’s the arrival of the physical AI layer, intelligence, finally embodied. The Supersonic Tsunami I wrote about in The Future Is Faster Than You Think isn’t just digital. It’s standing in your kitchen.

Make no mistake, a cute low-cost robot powered by AGI can do a lot. It’s not only the hardware that will enable this physical layer, but the multi-modal AI that underlies its full potential.

What does this mean? Within 36 months, a household robot will be as normalized as a smartphone.

The question for entrepreneurs is: What services, businesses, and entire industries get rebuilt around a world where physical labor is abundant and near-free?

Energy Red Alert: 7x Demand Energy Surge

The infrastructure required to power the AI and robotics revolution is straining the electric grid to its breaking point.

This past week, Texas lawmakers held an emergency hearing after being contacted by ERCOT (the Electric Reliability Council of Texas), which manages the electric grid for about 90% of Texas’s power load. ERCOT warned that data centers in the state have filed plans representing 410,000 megawatts of new electricity demand: a figure that is 7x larger than Texas’s entire 2024 consumption. Nevada’s utility, meanwhile, admitted it cannot meet its own 50% renewable energy target without burning more fossil fuels, because the data centers coming online need power now, not when the wind blows.

The data center energy storage market is projected to surge from $1.2 billion in 2025 to $4.1–$6.0 billion by 2030, growing at 28–38% annually. SoftBank and the U.S. Department of Energy just announced a partnership to build the world’s largest AI data center at the Portsmouth site in Ohio, on land leased directly from DOE.

Think about that. We are building AI infrastructure on former nuclear weapons production land. The energy future and the intelligence future are colliding in real time.

The answer depends entirely on whether we can accelerate clean energy faster than data centers scale. Which brings me to the story that gives me the most optimism this week.

Fusion is Getting Serious Money

All my life, fusion was science fiction, always 50 years away. This week, at the Energy Innovation Summit in San Diego, ARPA-E committed $135 million over 18 months to fusion technology, specifically targeting the barrier removal programs that have stalled commercial fusion from crossing into viability.

At the same time, SHINE Technologies secured a $263 million DOE loan for its Chrysalis facility, and Helion funded 25 external research proposals totaling $4 million in its latest open-science push.

This is an inflection point. Federal money is flowing. Private capital is flowing. The physics has been demonstrated. What remains is engineering. And engineering, unlike physics, is something exponential technologies eat for breakfast, and AGI/ASI will present to us on a silver platter.

If fusion delivers even partial commercial output by 2032 (which multiple credible paths now suggest is plausible) the energy constraint on AI, robotics, and every other exponential technology evaporates. This is the unlock that changes everything.

And Above All This, Rockets Keep Flying

My inner-most child is so happy to watch rockets launching and landing, and our missions to the Moon beginning in earnest.

SpaceX continues to maintain a remarkable launch cadence with Falcon 9 in 2026, ~1 launch every 2.3 days! As of April 11th, SpaceX had already conducted 44 Falcon 9 launches in 2026, following a record-breaking 165 Falcon 9 launches in 2025.

On the Starship front, the much-anticipated Flight 12 (the debut of the new Version 3 hardware) will likely take place in early to mid-May 2026. This V3 configuration features upgraded Raptor 3 engines and increased propellant capacity, and is expected to be the model to reach orbit and demonstrate in-flight upper stage refueling.

This week, SpaceX launched two missions in less than 48 hours. The Cygnus XL cargo ship carried 11,000+ pounds of supplies to the ISS on a Falcon 9 booster, and Crew-12.

Routine space access. That’s what this represents. When rocket launches become as unremarkable as airline departures, and we are nearly there. And don’t forget to fulfill Elon’s vision of a terawatt of compute in orbit. We’re going to need to get to a Starship launch every hour. Imagine that!

The Bigger Picture: We Are Watching Three Exponential Curves Hit Their Inflection Points Simultaneously

Here is the meta-pattern beneath this week’s headlines:

AI is crossing from software capability into physical embodiment, via robots that live in your home and models that surpass PhD-level human expertise. The energy infrastructure to power this transformation is straining the limits of the world’s most advanced electrical grid. And the energy technology to fix that constraint—fusion—is receiving the most serious investment it has ever seen.

These aren’t separate stories. They’re chapters in the same book. The book is titled: The Next Decade of Human History.

This week I’m releasing my latest book, We Are As Gods, which tells this whole story, and predicts where we go next, and what you need to do to survive and thrive in the coming age of Abundance!

The only question that matters now: Are you a participant in this transformation, or a spectator?

To a future of Abundance,

Peter

---

More From Peter

If you’ve enjoyed Metatrends, here are more ways to stay connected: