A Nuclear Household

---

The countdown is on…

Just two days ago, Elon Musk confirmed the launch date of the Tesla Robotaxi fleet. Tesla’s driverless Robotaxis will hit the road on June 12 in Austin, Texas… less than two weeks away.

The question of whether or not we believe Tesla will pull it off is one that’s been popping up in our feedback file…

Hello, I would like to post a question for the Friday AMA.

My question is, what would cause Tesla to miss the mark in June? I read the first role for Robotaxis will start with only 10 vehicles in Austin? Is there anything negative coming from the Tesla camp?

– Maurice R.

Will he miss the mark in terms of timing? Honestly, it’s possible.  Musk and his teams have been known to miss a date or two due to their aggressive development schedules.

But I have to say, not only am I not hearing anything negative, I’m seeing nothing but positive signals about the June 12 Robotaxi launch date.

Not only has Musk confirmed the date, but just yesterday, he posted this on X:

Last month, Tesla received its regulatory approvals to launch the service in June.  And there has already been independent confirmation that Teslas are, in fact, driving around Austin without anyone in the driver’s seat.

Could something happen that might cause a delay?  You bet.  One obvious scenario could be a reckless human driver crashing into a fully autonomous Tesla.  That would prompt a full safety review.  The press would certainly spread fear, uncertainty, and doubt about Tesla and Musk as “they” always do, and it would delay the Robotaxi launch.

But to tell you the truth, whether the Robotaxi fleet launches on June 1, June 12, or June 23… it doesn’t really matter.

The urgency isn’t in the exact date of the launch, but in the technology itself. And there’s no question… the launch is happening, and it’s happening very soon with this limited release in Austin, Texas.

As you said, Maurice, Tesla expects to launch its Robotaxi service in Austin with a small number of Teslas, initially. Notably, of its own cars.  Owners and lessees will not be able to immediately opt their cars into the Robotaxi network. Not just yet, anyway.

The launch starts with maybe a dozen driverless Teslas in the first week, scaling up each week. Within a few months, that number is in the thousands. Opt-in functionality will follow late this year or early next, and Musk already has his eyes on Los Angeles and San Francisco for the next cities to be equipped with Tesla’s autonomous taxis.

Musk narrowed down his timeline and plan of action in Tesla’s recent earnings call. Extensive updates have been made to the app. There has been increased federal scrutiny and local safety protocol training for emergency responders in the area.

We might be on Elon Time, but frankly, whether it’s in the next three days or the next three weeks, the launch is happening. And it will be worth the wait.

Because while he may miss the mark on the exact launch date, will he miss the mark in terms of the technology?

No, not a chance.

No matter how you dice it, Tesla is ready to roll out its Robotaxis. And once the technology has proved itself in the streets of some of these notoriously heavy-traffic U.S. cities, it will quickly scale out across the rest of the country and major cities around the world.

Tesla’s full self-driving technology is a lifesaving technology.  More than a million people in the U.S. have died over the years due to careless driving.  Tens of millions have died on a global level.

All of those deaths could have been avoided with this technology.  And Tesla has already proven with billions of miles of data that its technology is far safer than human drivers.

Austin will be ground zero for a major transformation in the global transportation landscape.

And yet, not shockingly, the skeptics are still out in force, either pointing to Elon’s history of missed deadlines or deciding, untested and unverified, that the technology isn’t road-ready. Forbes has even said the rollout “looks like a disaster waiting to happen.”

I have to assume all the naysayers have never experienced Tesla’s Full Self-Driving (FSD) technology for themselves. It’s hard to dismiss what’s directly in front of you. As I’ve often said, when I’m in my Tesla, I no longer drive it. The car drives me.

It easily and smoothly navigates even the most complex situations on the road, while I can simply sit back and observe. It’s truly incredible.

One way to take advantage of this is, of course, through Tesla directly. It’s a great company that’s about to put driverless taxis in every major city in the world. It’s certainly an option.

But I think folks often forget that it’s the smaller companies connected to Musk’s ventures that tend to see the truly exponential growth when his companies succeed.

That’s why my team and I have focused on the smaller companies – the suppliers and the ones helping build out the technology – that stand to gain much more when Tesla takes off again.

Not to mention that the moment everyone starts to realize just how massive this opportunity is going to be, they’re going to start piling in.

That’s why, all this week, I’ve been re-airing my Robotaxi Briefing. I want you to have the chance to make the most of this opportunity before everyone else catches on and realizes that Tesla, while no doubt a great company, is not the only way to play this. And it’s not going to get them the most explosive results.

It’s not too late. You can still go here to catch the briefing.

And on to the AMA…

AI Connectivity

Hello Brownstone team,

I have been following The Bleeding Edge and have been a Near Future Report subscriber for some time, and have most recently joined Exponential Tech Investor. I am excited about the “picks and shovels” approach to the AI buildout. With that in mind, I was hoping you could comment on a few companies in particular regarding AI connectivity and how you view one versus the other. My understanding of these technologies is limited.

Thanks and keep up the great work!

– Troy R.

Hi Troy,

Thanks for joining us, and thanks for the question.  Data center architectures have become quite complex and expensive in managing these AI-specific computational tasks, regardless of whether they are training or inference.

And you’re right, it’s going to be difficult for me to answer your question with specificity without mentioning specific names.  Obviously, our model portfolios are well-positioned with the most important companies providing all the key components for these high-performance data centers.  So, we’ll have to be more general.

To start, let’s have a look at what an advanced AI data center looks like:

On the surface, these kinds of data centers are row upon row of racks loaded with high-performance servers running GPUs and high bandwidth memory.

Whether it’s NVIDIA (NVDA) or Advanced Micro Devices (AMD) with their own proprietary server architecture and connectivity, or server vendors that incorporate NVIDIA and AMD GPUs like Dell (DELL), Super Micro Computer (SMCI), or Hewlett Packard Enterprise (HPE), they all have to rack’em and stack’em and connect each server unit and rack into a massive supercomputer.

The GPUs are the workhorses of these data centers and make up about 50% of the total hardware costs.  High-bandwidth memory is all solid state and tends to be integrated right onto the printed circuit boards, tightly integrated with the GPUs for speed and performance.  Each of these servers is connected via high-speed interconnect technology.

There are also network interface cards (NICs), which are the interface through which the GPU servers connect to the network.  This is what enables high-speed data transfer across the data center’s network.

Throughout the data center architecture, there are also high-speed switches that facilitate data transfer between servers, storage, and other networked systems.

All of these hardware components are connected via Ethernet cabling (made of copper) and fiber-optic cabling.  Below is a fiber-optic switch stack in Colossus:

And here is an example of the fiber-optic runs in Colossus:

There are also massive storage systems that are usually split between lower-performance storage provided by advanced hard disc drives (HDDs) and high-performance solid-state drives (SSDs).

The distinction lies in which drives are used to store massive amounts of data that doesn’t necessarily need to be accessed regularly – HDDs – and data that needs to be readily available with low latency – stored on SSDs.

In addition to all the computing, storage, and networking technology mentioned above, there are other critical components like optical transceivers that convert electrical signals over Ethernet into light for transmission over fiber-optic cables.

And naturally, there are a lot of cooling and power systems required to keep the data center operational and performing at optimal levels.

There are lots of players in this space. At times, some overlap between one company and another.  Analyzing and understanding the specific details of the best-in-class companies is, of course, what we do in The Near Future Report and Exponential Tech Investor.

Past research issues can be a great source for learning, as well as any new related research that we put out in the months ahead.

Hopefully, this will help you put all the pieces of the puzzle together.

How Feasible Is a Small Modular Reactor for the Home?

I’m looking forward to when some company can create an SMR for household use. Something like 20-30Kw output to let us truly be energy free. With it, I can get rid of my natural gas heating system and my natural gas hot water heater, replacing them with the electric equivalents that are powered by my H-SMR (Household Small Modular Reactor).

When do you think that will happen?

– Thomas W.

Hi Thomas,

That’s quite a bit of power output for a single house.  You’d probably need quite the mansion with a whole lot of electricity needs to require 30 kilowatts. But I do share your vision of being completely off the grid and not dependent upon utilities.  It really creates some great possibilities.

There is hope, but almost certainly not in the form of a small modular reactor (SMR).  The reality is that SMR technology is still too large for a single household.

Nano Nuclear Energy has been working on developing microreactors that are portable and relatively small (i.e., they’ll fit on the back of a semi).  But they’ll generate 1–20 megawatts, which is still way too much.

Radiant Nuclear has also been working on a microreactor designed for microgrids.  Like Nano Nuclear, it is portable and designed for lots of versatility to support military installations, hospitals, and even remote communities.

But even something like Radiant will produce 1 megawatt of power, which is still far too much for the average household.

A far more realistic solution for what both you and I are looking for will be found in nuclear fusion.  One company in particular caught my eye a few years ago, Avalanche Energy, is producing a micro-fusion reactor – small enough to carry by hand.

Avalanche is developing the Orbitron, which is designed for applications anywhere between 1–100 kilowatts of power.

The scale of the Orbitron is best suited for residential applications.  My own analysis suggests that the technology will be ready somewhere between 2027–2029, which is very exciting.  Not far away.

But there is one caveat.  Avalanche is targeting this product for applications like Navy ships, lunar outposts, space travel, military applications, and industrial installations.  This all makes perfect sense.  This is bleeding-edge technology, which requires major breakthroughs and a lot of research and development expenses.

In other words, these fusion reactors won’t be cheap.

But I do believe that for residential customers with the wherewithal to pay an industrial-scale price tag, there’s a chance to get your hands on one of these for an off-the-grid compound sometime before the end of this decade.

It’ll be worth the wait.

A Thank You

Dear Jeff & Team,

Just wanted to say a big THANK YOU for including such a thoughtful and detailed answer to my question in Friday’s mailbag. Your added research really adds extra nuances and angles, and I’m grateful for your breadth of perspective.

*The Risks of Unverifiable Research Data?*

Also, it was my birthday on Friday, so I couldn’t have gotten a better birthday present than this!!

Merci beaucoup,

– Paige D.

Hi, Paige. Thanks again for your interesting question last week.

For those who missed it in last week’s Bleeding Edge – Where Can I Buy a Robotaxi?, Paige wrote in asking about a study done by MIT researcher Aiden Toner-Rodgers on the impact of using artificial intelligence on materials innovation.

The study garnered a lot of attention from major outlets and raised important questions regarding how AI will enable humans to be more productive in scientific discovery…

But it also, in the end, raised concerns about what happens when unverified, non-peer-reviewed research is widely circulated, as it was revealed that Toner-Rodgers’ dataset was apparently manipulated.

Here’s what I wrote…

The MIT economists named by Toner-Rodgers in the paper – Daron Acemoglu (a Nobel Prize-winning MIT economist) and David Autor (an MIT labor economist) – both lauded the paper when it came out.  Acemoglu said, “It’s fantastic,” and Autor said, “I was floored.”

And yet, a few days ago, both “experts” published a request for withdrawal on the basis that they had no confidence in the veracity of the data or the research.  What a black eye for MIT, Acemoglu, and Autor.

What I did appreciate was that someone from the industry contacted Acemoglu and Autor, explaining why the research didn’t make sense.  After all, someone who works in materials science (industry) has a very different perspective than two MIT academics who are economists, nonetheless, not materials scientists.

That’s why they were forced to take the paper down.  And Toner-Rodgers is no longer at MIT, which clearly suggests that there was fraudulent activity regarding the research.

In short, the situation could have been avoided entirely had all parties involved simply followed appropriate research standards, which would have meant having experts in the field review the research before it was disseminated.

To be clear, that means having it reviewed by materials scientists, not economists. Peer revision has long been a tool used to ensure, to the best of our ability, the veracity of research.

But this, of course, brought to mind how academia and the nature of scientific research have devolved. After all, results can be – and are – easily manipulated based on the motivations of whoever is funding the research.

You’re welcome to read my full response to Paige’s question right here. But I need to reiterate here that it is precisely the prevalence of data distortion and the manipulation of information to push a narrative that has seen a new standard being established…

Trillions of dollars are being invested in the private sector in bleeding-edge technology right now.  Academia is moving at a linear pace.  Industry is moving at an exponential pace.

What most will see won’t be academic papers. They’ll see the new product or service that is changing the world.  And most data sets that are used to develop new products by tech companies that give that company a proprietary advantage won’t need peer review.  They’ll remain in-house, under high security.

[…]

The key point is that all the politics, corruption (to get grants), decels, and falsifications in academia mean less and less with each month that passes.

Academia has lost its high perch.  The real turning point was the pandemic, when the most prestigious universities dropped evidence-based scientific research in exchange for pushing a political narrative so that they could receive massive U.S. government grants.  MIT, Harvard, Yale, and so many others are guilty of that.

Industry doesn’t care.  It just accelerates.

This will be the new normal, the new standard.

Thank you again for your question, Paige, and to everyone who wrote in this week.

As always, my team and I can be reached right here for any questions or concerns you’d like addressed in these AMA issues.

Have a great weekend.

Jeff