AI company leads in autonomous vehicle production

Investment Analysis Clubs Macro Economic Trends and Risks
1

Alphabet is not a manufacturer, but their company Waymo is the leading US producer of autonomous vehicles.

Today, scaling autonomous robotaxis is

  • first about developing the AI software and hardware that meets performance requirements (can drive passengers to a destination) with sufficient safety
  • second about deploying the AI driver in new cities, each with unique streets, highways, intersections, addresses and governance
  • third about implementing an efficient operation in fleet management (refueling/charging, storage, maintenance, cleaning, customer support, etc)
  • fourth about efficiently manufacturing a large number of AI-driving vehicles

Waymo has about 4,000 autonomous taxis in commercial service across 15 markets (and another 15 markets in testing), so manufacturing vehicles at large scale is not a need - their needs over the next couple of years are likely in the 10s of thousands.

Waymo is transitioning from Jaguars to vehicles manufactured by Zeekr and Hyundai.

The Zeekrs are being deployed into commercial service throughout 2026, with about 5,000 expected to be imported through the year.

The longer term plan for Zeekrs is not clear because they are impacted by US-China trade policy.

The Hyundais started on-road testing earlier this year and are now beginning larger scale production at a Georgia plant in Q4 2026.

Vehicle No. In Service Expected Total Timeline
Jaguar I-Pace ~3,500 ~3,500 In service, phasing out
Waymo Ojai (Zeekr RT) ~400 to 500 5,100 through 2026 In service since Aug 2026
Hyundai Ioniq 5 0 10,000s Road testing since ~Jul 2026, mass production Q4 2026

At Hyundai Motor Company’s 2026 CEO Investor Day in Seoul, the automaker said deliveries of IONIQ5 vehicles to Waymo are scheduled to begin in the fourth quarter of 2026.

The Ioniq5 manufacturing is done by Hyundai at their Georgia plant and Waymo completes the build, particularly the final sensor installations.

Another concern with AI driving vehicles is the cost of the Lidar sensors, but according to Rivian, these costs have fallen dramatically, like hardware costs often do.

”You’re talking on the order of a few hundred dollars versus if you went back several years ago where it was thousands or even tens of thousands of dollars within the vehicle,”

7 Likes
2

Why SpaceX and Tesla Are Moving Into POLE POSITION

I hadn’t heard from Jo Bhakdi for quite some time. It’s a long hour and a half video but Cybercab starts at merely 1:40. After 11:15 it’s all about SpaceX

In case you have not met Jo Bhakdi:

GoogleAI:

Jo Bhakdi (Johannes Bhakdi) is an economist, investor, and tech entrepreneur best known as the founder and CEO of Quantgene, a precision genomics and AI biotechnology company. Over the last several years, his work has expanded into macroeconomics, AI-driven capital markets, and community building. [1, 2]

:dna: Biotech & Precision Medicine

Bhakdi founded UC Berkeley-born Quantgene in 2015. The company combines advanced machine learning, cloud systems, and single-molecule DNA sequencing to detect the deadliest cancers at early stages (Stages 1 and 2) using standard blood samples (liquid biopsy). [1, 2, 3]

  • The Mission: To extend the healthy human lifespan by a decade within a decade.

  • Core Product: Developed Serenity, a subscription-based direct-to-consumer genomics service combining hereditary risk assessment and liquid biopsy diagnostics.

    [1, 2]

:robot: AI, Investing, & Community

Beyond biotechnology, Bhakdi is heavily involved in navigating the economic transitions triggered by artificial intelligence:

  • Pioneerlands: He is the founder of Pioneerlands, a community focused on the age of Artificial General Intelligence (AGI), macro trends, and tech stocks like Tesla. [1, 2]
  • PioneerAlpha: An investment fund built around the networking and insight of the Pioneerlands ecosystem. [1]
  • Content Creation: He actively publishes research and video essays via his Jo Bhakdi YouTube Channeland offers sharp market commentary on his X profile (@JOBhakdi). [1, 2]

:graduation_cap: Background & Early Career

  • Education: Holds a Master’s degree in Economics and Psychology from Tübingen University in Germany, specializing in financial theory and statistics. [1, 2]
  • Y Combinator: Bhakdi is an alumnus of the prestigious Y Combinator startup accelerator. [1]
  • Prior Ventures: Before Quantgene, he created i2X, an investment framework utilized to build risk-mitigating venture capital portfolios, which laid the analytics foundations for his later biotech machine learning platforms. He also initially worked in strategic executive roles for major global marketing networks like WPP and Omnicom. [1, 2]

The Captain

4 Likes
3

I asked AI to summarize this video with respect to Tesla autonomy.

The summary had basically zero new factual data or quantitative, time-specific forecasts.

Tesla does not yet have an autonomous vehicle that it can scale.

For example, the video says Tesla has registered over 100 Cybercabs in Austin.

Well, registered vehicles are not a functioning, scaling, autonomous taxi service.

If you want to have a speculative discussion on all of the great things Tesla is going to do in the future, then please create a new thread for purely speculative discussion.

Or this thread is available for endless opinions and speculation:

3 Likes
4

Which AI did you use? The ten minutes on this in the video were beyond those assumptions. Across the EU FSD has been permitted as of now for the cyber taxis. New factual data as of this week? Or two days ago? Tesla is scaling this time. Zero new factual data according to AI does not mean you know what is in the video.

5

You guys told me that AI only uses old data. That means “zero new factual data” to be expected. :cat_with_wry_smile:

The Captain

2 Likes
6

Not how it works. Summarizing a video means it translates the audio and summarizes from that. It doesn’t use its “deep knowledge” except in the translation process.

3 Likes
7

And in addition to what Goofy correctly pointed out, no one told you that AI only uses old data. We told you that free public AI relies primarily on its training data (which is anywhere from six months to two years old, depending on the model) along with very constrained web searches - a handful of searches and a dozen or so snippets from them, in the case of Gemini. AI is unreliable for looking at questions about the current world because it looks at such a tiny amount of updated information, but it’s not zero. That’s how AI is able to create the illusion of a well-considered response on the current state of things.

8 Likes
8

That clearly did not happen. Listen to the video.

Is that all you heard?

Misspoke here, small countries are allowing the app for the cybercab.

Something like 2/3s of the EU has approved FSD for their roads. Might be more because there are major nations in that group.

Production is around 2000 units in the last couple of months. I am forgetting the exact number, but it is ramping up this time. There are Tesla Robots in the factories.

9

Yes.

I’m quite certain Alphabet Google Gemini knows how to summarize Alphabet YouTube content.

Curious how no one has bothered to state the new and factual and very specific statements made about Tesla autonomy from the video except me.

But anyway, let’s continue to debate about what AI can and cannot do versus actual facts and quantitative, time-specific predictions about vehicle autonomy.

Waiting for Godot:
actual facts and quantitative, time-specific predictions about vehicle autonomy

1 Like
10

@albaby1 You are in over your head.

No, Gemini cannot natively transcribe a YouTube video from scratch using just the URL if the video completely lacks a text dialog (like speech, auto-generated captions, or a text layer). [1, 2]

When you paste a YouTube link directly into the Google Gemini chat interface, the AI does not “watch” the video pixels or listen to the raw audio track in real-time. Instead, it relies entirely on the existing text layer—meaning it pulls YouTube’s auto-generated closed captions (CC) or user-uploaded transcripts. [2, 3, 4]

Here is how Gemini handles different types of videos without a “text dialog,” and how you can work around it:

1. Videos with Spoken Words but No Manual Subtitles

If the video has spoken words but the creator didn’t upload a script, Gemini can still summarize it. It achieves this by reading YouTube’s automatic speech-to-text captions. As long as YouTube’s system can generate an automated transcript behind the scenes, Gemini can grab that text and use it. [3, 4, 5]

2. Silent Videos, Instrumental Music, or Purely Visual Content

If a video contains absolutely no speech (e.g., a silent vlog with background music, a timelapse, or a visual compilation) and has no captions enabled, Gemini will fail to process the URL or will give you mixed/hallucinated results. It will usually tell you that it cannot find a transcript to summarize. [1, 2]

How to Bypass This Limitation :hammer_and_wrench:

If you have a video that completely lacks a transcript, or you want Gemini to analyze the actual visual frames, you have to change your workflow:

  • Upload the raw file directly: If you have the video file saved to your device, you can upload it directly into Gemini (or Google AI Studio). Because Gemini is multimodal, it can analyze visual frames and raw audio sequences when a file is directly attached—it just cannot do this via a standard YouTube link. [3, 6]
  • Use Gemini 3.5 Transcribe (for creators/developers): If you are processing video files externally, Google’s specialized audio model, Gemini 3.5 Transcribe, is built specifically to extract raw audio from files and turn messy speech, background noise, or uncaptioned dialogue into formatted text. [7, 8]

Are you trying to summarize a specific video that is completely silent / visual-only, or does it have spoken dialogue that just doesn’t have subtitles yet? Let me know, and I can give you the exact steps to get the summary you need!

[1] https://www.transkripe.com

[2] https://lynote.ai

[3] https://geminiprompt.id

[4] https://otio.ai

[5] https://www.youtube.com

[6] https://www.youtube.com

[7] https://www.youtube.com

[8] https://bibigpt.co

As long as YouTube’s system can generate an automated transcript behind the scenes, Gemini can grab that text and use it.

AI Mode Conversation: If a YouTube video does not have a text dialog will gemini still transribe it to summarize it?

If a YouTube video does not have a text dialog will gemini still transribe it to summarize it?

No, Gemini cannot natively transcribe a YouTube video from scratch using just the URL if the video completely lacks a text dialog (like speech, auto-generated captions, or a text layer). [1, 2]

When you paste a YouTube link directly into the Google Gemini chat interface, the AI does not “watch” the video pixels or listen to the raw audio track in real-time. Instead, it relies entirely on the existing text layer—meaning it pulls YouTube’s auto-generated closed captions (CC) or user-uploaded transcripts. [1, 2, 3]

Here is how Gemini handles different types of videos without a “text dialog,” and how you can work around it:

1. Videos with Spoken Words but No Manual Subtitles

If the video has spoken words but the creator didn’t upload a script, Gemini can still summarize it. It achieves this by reading YouTube’s automatic speech-to-text captions. As long as YouTube’s system can generate an automated transcript behind the scenes, Gemini can grab that text and use it. [1, 2, 3]

2. Silent Videos, Instrumental Music, or Purely Visual Content

If a video contains absolutely no speech (e.g., a silent vlog with background music, a timelapse, or a visual compilation) and has no captions enabled, Gemini will fail to process the URL or will give you mixed/hallucinated results. It will usually tell you that it cannot find a transcript to summarize. [1, 2]

How to Bypass This Limitation :hammer_and_wrench:

If you have a video that completely lacks a transcript, or you want Gemini to analyze the actual visual frames, you have to change your workflow:

  • Upload the raw file directly: If you have the video file saved to your device, you can upload it directly into Gemini (or Google AI Studio). Because Gemini is multimodal, it can analyze visual frames and raw audio sequences when a file is directly attached—it just cannot do this via a standard YouTube link. [1, 2]
  • Use Gemini 3.5 Transcribe (for creators/developers): If you are processing video files externally, Google’s specialized audio model, Gemini 3.5 Transcribe, is built specifically to extract raw audio from files and turn messy speech, background noise, or uncaptioned dialogue into formatted text. [1, 2]

Are you trying to summarize a specific video that is completely silent / visual-only, or does it have spoken dialogue that just doesn’t have subtitles yet? Let me know, and I can give you the exact steps to get the summary you need!

  • [image]

    Lynote

    How to Use Gemini to Summarize YouTube Videos - Lynote

    Jun 20, 2026 — Can Gemini summarize YouTube videos without transcripts? Generally, no. Most AI models, including the standard version of Gemini, …

    [image]

  • [image]

    GEMINIPROMPT.id

    Using Google Gemini to Summarize Long YouTube Videos …

    Jan 16, 2026 — Gemini isn’t analyzing the pixel data of the video frames in real-time. It is parsing the auto-generated captions or the uploaded …

    [image]

  • [image]

    Otio AI

    How to Write a YouTube Summary AI With Gemini — Otio Blog

    Jan 9, 2026 — Gemini can summarize YouTube videos by extracting spoken content from captions or transcripts and turning that text into concise, …

    [image]

Show all

As long as YouTube’s system can generate an automated transcript behind the scenes, Gemini can grab that text and use it. Otio AI +2

Is this often not available

No, it is not “often” unavailable, but it does fail under specific, common conditions.

For the vast majority of standard spoken videos (like vlogs, tutorials, or news clips), YouTube successfully generates automatic captions within a few hours of uploading. However, the system is not perfect. [1, 2]

YouTube’s Official Support Documentation outlines several common reasons why automatic transcripts fail to generate entirely, leaving Gemini with nothing to read: [1]

11

This thread is not about YouTube and AI.

Can you contribute any NEW,
actual facts and quantitative, time-specific predictions about vehicle autonomy, from Waymo, Tesla or anyone else?

If yes, please post.

(Note: Tesla FSD is not vehicle autonomy, it’s driver assist where the human is liable for all driving.)

2 Likes
12

Listen to the first eleven minutes.

This was posted very recently on YouTube. Your summary was too soon for YouTube to transcribe it.

What AI bot did you use?

adding

At 5:30 the app being allowed for download across the EU.

13

I asked Perplexity to summarize the video (I find Perplexity is good at this task)

Topic What Bhakdi argues Numbers and timing he gives
Cybercab scaling The purpose-built vehicle is scaling materially faster than the initial Model Y robotaxi rollout. He expects deployment to become a catalyst for Tesla’s valuation. He cites 169 registrations, 55 photographed vehicles, and estimates roughly 80 actual vehicles after the first month, versus about 20 in the comparable initial robotaxi period. He forecasts 4,000 manufactured by October’s end and at least 10,000 by year-end.
International expansion Wider availability of the Robotaxi app and supervised FSD approvals signal preparation for broader expansion. He names multiple European markets, plus Singapore and Malaysia, where he says the app has become downloadable.
Optimus and AI5 Internal robot deployment and preparations for large-scale AI chip production suggest Tesla is moving toward commercialization. He estimates 5,000 Optimus V3 robots deployed internally by year-end and discusses AI5 memory being reduced to 72 GB per chip.

Let’s take these one at a time. Cybercab is not scaling. According to robotaxi.com the number of Cybercabs is use on any given day is tiny. There have been only 20 total Cybercabs reported in the last three days, and most of these only made one trip. Cybercab service is almost non-existent.

His claims of 4000 Cybercabs by the end of the month and 10,000 by the end of the year could be correct, however this is not a positive outcome for Tesla. We have all seen the pictures of Cybercabs being shipped to parking lots around the country where the vehicles are simply left to rust and the tires go flat. Tesla is producing a product they have no market for. This is not good.

International expansion: Wider availability of the Robotaxi app and and $6 will get you a latte’. You need regulatory approval and demonstrated technical ability. App downloads have nothing to do with anything. It is straw grasping.

Optimus and AI5: There is zero evidence a single Optimus robot will be deployed internally in any meaningful task. Reality check: Have you seen any video of Optimus doing anything better than a regular robot? Have you seen any video of Optimus doing any task that could replace a human?

One day Optimus might replace human labor. But building 10,000 of them by the end of the year when there is no use case would be a huge business blunder. Not yet as big as the Cybertruck blunder, but give it time.

Remember, we were told there would be 10,000 Optimus by this time last year. These claims are starting to get old. At some point, an Optimus needs to show up to work.

4 Likes
14

The summary is good. I would like to see evidence…

The domain robotaxi.com is owned and registered by Robotaxi Ltd, an autonomous vehicle service company directed by Ian Hoy.

No market today. How accurate is your crystal ball?

Par for the course, all competitors in the market have this hurdle.

Lack of evidence is not proof. (ask AI about it).

One Fool’s opinion. Elon has accomplished several impossible tasks."

Yes, Elon Time is an issue.

The Captain

15

Isn’t that rather the point, though? If someone claiming great things are afoot for Tesla, but doesn’t provide any evidence, then it’s not much proof that the points claimed in the video are correct.

“Elon Time” is the issue for Optimus and Cybercab. Tesla is pursuing “physical AI.” It’s making the containers for the AI to go into today, but they don’t have the AI capable of using those containers developed yet. Because even though the cars are coming off the line, and the robots are being manufactured, the AI is on Elon Time. Which means Tesla is accumulating few hundred millions of dollars’ worth of robots and vehicles that it can’t actually use. Sure, some portion of those robots and cars are going to end up being useful for testing purposes - but not too many more than they have already.

Tesla has been persistently incorrect in assessing how ready their driving program is to serve as a driverless vehicle. Musk in particular is consistently too optimistic about how soon AI will be ready for that task. And they might also be too optimistic about how close they are to having an AI that can use a robot body to do useful work. If the AI is, in fact, on Elon Time then we won’t necessarily see much “leading in autonomous vehicle production” any time soon, because the vehicles won’t actually be autonomous.

5 Likes
16

Opps, my bad. Correct link is below. The claim is the Cybercab program is scaling based on vehicle registrations and app downloads. This claim is not supported by the only metric that actually matters: The number of trips. In the last 24 hours, a total of eight unique Cybercabs have been spotted, each making only one trip. The Cybercab program is almost non-operational.

While all operators face regulatory hurdles, Tesla’s competitors are measurably farther along in the process. I like permitting as a metric because it is honest. Permits are a matter of public record and are a fixed requirement for robotaxi operations. The number of app downloads tell us nothing. Permits tell us everything.

Right! He’s making bold claims but unable to provide evidence for these claims. He’s just saying stuff based on no proof.

3 Likes
17

Yes, they started sooner. As I see it, the most important aspect is getting autonomous taxis legalized and regulated, that helps all providers and consumers. Where we disagree is on who will lead the pack once legalized and regulated.

I suppose you mean Musk, not the guy whose website you cited. Everything Musk says is burdened by Elon Time. The very first thing I heard from Elon was about Tesla able to do battery swapping. It was just an ego trip about and not something that would make me buy TSLA stock. Some ten years later Tesla had proved itself “investable.” Ranking Tesla by ElonTalk is a waste of time. Rank Tesla by what the business is doing. You don’t get to become the leading EV seller by talking or boasting but by executing. Tesla’s execution has been stellar with a few duds on the way. How come no one in this board talks Tesla semis? But talk, talk, talk about the famous truck that has no market but is loved by their users.

BTW, the Robotaxi Tracker page is stuff that I entirely ignore. Cash flow is what makes the word go around.

The Captain

18

No, I’m talking about the claims in the video you posted. The number of app downloads is meaningless. Regulatory approval is critical. I have zero problem with Tesla being deliberate and methodical. Applaud them even. But I find no benefit in using meaningless numbers to advance the false narrative Cybercab is scaling. The number of cars on the road and permission to operate them are the things the matter. Everything else is just hand waving.

Feel free to start a thread! The video wasn’t about the Tesla semi, hence no discussion.

Tesla Q2 2026 free cash flow: -$1.09 billion

4 Likes
19

A meaningless number in isolation. Business does not accumulate cash a la

images
images738×414 108 KB

When you have nothing better to do with cash you pay dividends or buy back shares. Tesla is in transition from an EV pure play to embodied AI in robots and self driving taxis. Tesla and SpaceX are building the massive Terafab facility in Grimes County, Texas with an initial $16.8 billion investment to produce custom AI inference chips for Cybercabs and Optimus robots.

The Captain

LOOK what I found!