GM. Welcome back to Milk Road AI, where we track the technologies reshaping how the world works before most people notice.
Here’s what we’ve got for you today:
- ✍️ The robotaxi arms race, explained
- 🎙️ The Milk Road AI Show: Why Physical AI Is the Most Important AI Investment of the Decade w/ Patrick & Duncan
- 🍪 Nvidia cuts a $20B deal to lock down the AI chip stack
Launching The Energy Network on Solana, Fuse Energy has just secured $70M in Series B funding. Discover the future of energy now.

Prices as of 10:20 AM ET.

INSIDE THE WAYMO VS. TESLA ROBOTAXI WAR
Did you know elevators used to have “drivers”?
They wore uniforms, pulled levers, and made awkward small talk about the weather while you waited to get to the 4th floor.
Then, in 1945, New York's elevator operators got confident. They went on strike, and it immediately backfired.

Buildings were forced to install automatic buttons just to function, and the "drivers" were erased from history.
We are currently in the "elevator operator" phase of driving.
We think the human behind the wheel is necessary, but we are one software update away from looking like a relic.
The transition has already started, and the only question is who gets to own the buttons.
Right now, two giants are fighting a blood sport to answer that question: Waymo vs. Tesla.
The hidden cost of Waymo’s perfection
If you look at the scoreboard today, Waymo (Google’s self-driving car) looks like the valedictorian of the robotaxi class.
They are currently operating in about 10 American cities and plan to more than quadruple that number soon.
They have logged over 127 million autonomous miles and completed more than 14 million paid trips.
They have the stats and boast to have a better safety record than anyone else.
In fact, Waymo didn’t just claim it was safer, it showed up with charts, percentages, and enough data to make a regulator blush.

But beneath those shiny stats lies a nightmare.
Their cars look like they drove through a Best Buy and everything stuck to the roof.
They are packed with:
- 5 Lidar (laser scanners that map the world in 3D by bouncing light off objects).
- 6 Radars (radio waves that track distance and speed, especially in rain or fog).
- 29 Cameras (high-resolution vision for signs, lanes, lights, and objects).

All of this hardware makes each vehicle cost roughly 4x more than a Tesla (estimated $150k - $200k per unit).
They basically built a tank to drive you to the grocery store.
Waymo knows this is a problem.
That’s why they’re already testing their sixth-generation build, stripping down hardware and trying to make the system cheaper and more scalable.

Source: neville_medhora/Instagram
The bigger issue is how Waymo’s system thinks.
Waymo relies heavily on ultra-detailed HD maps, what critics call a "geofence prison".
Before a Waymo can drive down a single street, humans have to scan it inch by inch, memorizing every curb, lane marking, pothole, and fire hydrant.
This means Waymo doesn't actually know how to drive, it knows how to recite a map.
We saw proof of this last week when San Francisco suffered a massive power outage.
Traffic lights went dark across the city.
The Waymo "map" indicated there should be working lights, but in reality, there weren't.
The result was absolute chaos.
The cars bricked themselves in the middle of intersections, blinking their hazards in confusion while human drivers honked, yelled, and questioned the future of technology.
Eventually, Waymo suspended its service and announced a fleetwide software update to tackle the problem.

This is the Waymo tradeoff in one sentence:
It's the safest kid in the class, but it needs perfect conditions to look like a genius.
FUSE ENERGY RAISES $70M AT A $5B VALUATION
Is this the most legit energy company to ever enter crypto?
Fuse Energy is a $400M ARR utility powering 200,000+ homes, today announcing a $70M Series B at a blockbuster $5B valuation.
This comes after the recent beta launch of The Energy Network, a new digital layer engineered to scale our grids and save billions in costs.
And now, it’s just building its momentum:
- Today raised $70M in Series B led by Lowercarbon and Balderton
- Now valued at $5B
- Launched beta on Solana
- Received landmark no-action letter from the SEC last month
- Planning listings for early 2026
A new foundation for the grid is coming.
Check out their announcement here and follow Fuse on X for updates.

INSIDE THE WAYMO VS. TESLA ROBOTAXI WAR (P2)
Tesla, on the other hand, has a very different definition of “safe”.
Tesla looked at Waymo’s sensor chandelier and basically said, “No thank you”.
Tesla’s system only uses 9 cameras (AI + Vision).

So how is that even remotely safe?
Because Tesla’s philosophy is simple: If a human can drive using two eyes and a brain, a car should be able to drive using cameras and a neural network.
(aka an AI brain that learns from experience instead of freaking out when something unexpected happens.)
Don’t believe me? Just look at this image:

Those 9 cameras give the car a full 360-degree view (front, sides, rear, wide-angle, long-range), constantly feeding raw visual data into Tesla’s AI.
And here’s the key difference most people miss.
Tesla isn’t trying to detect objects. It’s trying to understand behavior:
- Waymo asks: "What should happen here?" (Rules-based)
- Tesla asks: "What usually happens here?" (Behavior-based)
Why Tesla is playing a different game than Waymo
Tesla has put cameras in millions of consumer cars and decided to solve autonomy like a software problem.
This produces an absurd amount of data - enough to melt your brain.
While Waymo has logged ~127 million autonomous miles, Tesla customers have driven nearly 7 billion miles using supervised Full Self-Driving.

That’s roughly 80,000 years of driving, which explains why Tesla’s AI has seen things you can’t even fathom.
While Waymo is carefully mapping a cul-de-sac in Phoenix, Tesla has millions of cars seeing snow in Norway and roundabouts in London.
They are feeding the neural net every possible edge case on Earth.
Hence, Tesla’s FSD (supervised) data suggests that when cars understand behavior instead of rigid rules, they can dramatically reduce human-error crashes.

I know what you are thinking: "You are comparing apples to oranges".
And you're right. Waymo has thousands of driverless cars on the road today.
Tesla currently only has a tiny pilot program of ~30 robotaxis roaming around Austin, Texas.
But ignoring Tesla because they are "late" is like ignoring the internet because the library has more books.
Tesla is now aggressively moving from supervised to driverless:
- They launched a pilot in Austin with ~30 cars that have already crossed 250,000 miles.
- Musk has promised to eliminate safety monitors by year-end and expand to 8–10 cities.
Analysts like Dan Ives, who recently came on the podcast and told us why the AI trade is not over, expect that number to jump to 30 cities next year.
The verdict: Today vs. Tomorrow
So, who wins the death match? If we’re talking about today, it’s Waymo.
Waymo got there first because it played the regulatory game perfectly: slow, local, and obsessively safe.
They have paid trips, zero driver intervention, and a real product you can use right now across multiple cities.
But when you zoom out, the math starts working in Tesla’s favor.
Tesla can turn millions of existing cars into robotaxis with a simple software update, allowing them to drive prices into the floor.
Waymo, with its $150k vehicles and expensive mapping teams, simply won't be able to compete on cost.
In the long run, this won’t be decided by who looks best today.
It will be decided by who can scale, who can get cheaper, and who ends up controlling the interface.
Waymo might have won the battle for now, but my money is on Tesla to win the war and own all the proverbial elevators when it’s over.
Alright, that’s it for this edition of Milk Road AI, but we’d love to hear from you.
Hit reply and tell us:
Who wins the robotaxi war?
- Waymo, slow, safe, and allergic to chaos.
- Tesla, chaos is the training data.
- I still don’t trust cars with opinions.

THE FUTURE HAS ARMS 🤖
In the Rollup this past Friday, Patrick & Duncan sat down to talk about why physical AI could be the biggest investment opportunity of the decade.
Here’s what you’ll hear:
- Why robotics is moving beyond rigid scripts into flexible, model-driven agents
- How the Physical Intelligence paper cracked cross-domain training at scale
- What makes robot training more like LLMs than you’d think
- Why labor costs, capex cycles, and crypto might collide in this space
It’s a banger of an episode, don’t miss it 👇
YouTube | Spotify | Apple Podcasts

Bridge helps businesses send, store, accept, and launch stablecoins instantly. Learn how Bridge is powering stablecoin-backed money movement.
Chainlink is the industry-standard oracle platform bringing the capital markets onchain, and the market leader powering the majority of decentralized finance. Learn more about Chainlink.
Arc is a new Layer-1 blockchain designed to support real-world economic activity and it’s now live in testnet. Dive into Arc and start experimenting.

BITE-SIZED COOKIES FOR THE ROAD 🍪
Nvidia struck a $20 billion licensing deal with AI chip startup Groq (no relation to Grok) and will hire its CEO. The deal could become Nvidia’s largest ever and further cement its dominance in AI chips.
Nvidia purchased $5 billion worth of Intel shares under a deal announced in September. The investment gives Intel a major financial boost after years of heavy spending and missteps.
Italy ordered Meta to suspend its ban on rival AI chatbots on WhatsApp. Regulators say the policy could harm competition.

MILKY MEMES 🤣


ROADIE REVIEW OF THE DAY 🥛













