Farzad: Tesla Says 75% of New North American Deliveries Include FSD + His "F-" for Tesla Earnings Production + Why Neoclouds "Have to Go to Spot Pricing"

Tesla's Q3 numbers landed Wednesday afternoon, and Farzad went straight at garyblack00 over a 75% FSD attach rate. Later that afternoon he gave Tesla's earnings production an F-. Earlier Wednesday he argued other neoclouds need spot pricing to make sense next to SpaceX, and Thursday morning he framed Starlink Mobile as bandwidth for agents.

"Please get on your knees and apologize to me"

SawyerMerritt, Oct 21 at 3:08pm CT: WOW, Tesla says that they achieved record FSD attach rates in North America in Q3 2026, with a whopping 75% of all new deliveries including FSD subscriptions, up from 55% in Q2 2026

The post quotes Tesla: "We believe that FSD is driving demand in markets where it is available." The text in this pull cuts off after that.

Farzad quoted it at 3:12pm CT: Hey @garyblack00 please get on your knees and apologize to me right now. A minute later he replied to SawyerMerritt: wait so an incredible feature sells cars what

stevenmarkryan, 3:16pm CT: 75%. Imagine when it goes unsupervised... Farzad RTd it at 3:20pm CT. Elon, 3:54pm CT: People are buying FSD with a car attached, not the other way around. Farzad RTd it at 4:24pm CT.

SawyerMerritt also posted at 3:09pm CT: Tesla says they ended Q3 2026 with their largest order backlog for passenger vehicles in four years. The Tesla quote in that post says "battery availability continued to be the limiting factor in increasing auto volumes in the quarter" and that "we have made additional progress in securing future supply." Farzad RTd it at 3:21pm CT.

The 75%, the 55% and the four-year backlog are Tesla's claims as SawyerMerritt posted them. The "car attached" line is Elon's.

At 7:50pm CT, Farzad posted a video titled Analysts Expected Tesla to Burn $4.5B. Tesla Reported +$1.5B. At 8:20pm CT came another: Tesla Says Over 75% of New North American Deliveries Now Include FSD. The $4.5B and +$1.5B figures appear only in that video title in this pull.

At 3:07pm CT, a minute before SawyerMerritt's FSD post, Farzad had posted: As a $TSLA, $SPCX, and $META investor, today is a good day! On Thursday at 7:05am CT he quoted it: I take this back. He didn't say why.

Check SawyerMerritt here. Farzad's quote here and reply here. stevenmarkryan here. Farzad's RT here. Elon here. Farzad's RT here. The backlog post here. Farzad's RT here. The videos here and here. The good day post here. "I take this back" here.

"Tesla earnings production: F-"

Farzad, Oct 21 at 4:31pm CT: "SpaceX earnings production: A+" and, on the next line, "Tesla earnings production: F-".

He was grading the production, not the results. At 3:21pm CT he'd asked: Are we getting a video call for $TSLA ER? ER is short for earnings report. The SpaceX call, the A+ half of that post, is covered in our full breakdown of the SpaceX earnings call.

stevenmarkryan replied at 4:40pm CT: It's almost like having an existing extremely skilled and experienced live stream production team means that live streaming your earnings call is possible. Farzad, 4:48pm CT: Yeah just like Palantir and Robinhood, right?

Elon, 6:01pm CT, replying to TeslaBoomerMama and travisraxelrod: It will be next time. Their post isn't in this pull. Farzad replied to Elon at 9:04pm CT: YAY

Elon, 10:36pm CT, replying to Farzad: Fair enough, we will make it more entertaining next time. This pull doesn't include the Farzad post Elon was answering. Farzad's response, Thursday at 5:39am CT, has no words, just a link.

Check the grades post here. The video call question here. stevenmarkryan here. Farzad's reply here. Elon at 6:01pm here. Farzad's YAY here. Elon at 10:36pm here. Farzad's response here.

"Big positive signal"

aelluswamy, Oct 21 at 1:58pm CT: We are launching the Tesla Robotaxi safety report today.

The post says Robotaxi "has driven a total of 3.3 million miles autonomously, and has had zero injuries nor airbag deployments." It then says that if the same miles were driven manually, "we should expect roughly 9 injuries." The text in this pull cuts off there.

Farzad quoted it at 2:30pm CT: Tesla releasing this means that they are highly confident on their path for increased safety, which means more Robotaxis. Big positive signal $TSLA

That's Farzad's read. The 3.3 million miles and zero injuries are aelluswamy's figures.

jamesdouma, 7:39pm CT: 7 years ago it looked like Tesla Robotaxi would have competition, but it would be 5 years behind. Now it looks like there just isn't going to be any. Farzad RTd it at 8:24pm CT. That call is jamesdouma's.

At 8:50pm CT, Farzad posted a video titled Tesla Says Robotaxi Hit Nearly 2 Million Unsupervised Miles. The title counts unsupervised miles. aelluswamy's post counts miles driven autonomously. Tesla's Q3 update puts cumulative unsupervised Robotaxi miles at nearly 2 million, the figure in the title. The 3.3 million is aelluswamy's total to date for the new safety report, a separate count that his post doesn't break down.

Check aelluswamy here. Farzad's quote here. jamesdouma here. Farzad's RT here. The video here.

"SpaceX is effectively the best neocloud right now"

A neocloud is a company that rents out AI compute.

jukan05, Oct 21 at 5:03am CT: SpaceX is effectively the best neocloud right now...

The post continues: "SpaceX insists on short term leases only, so as long as compute keeps getting scarcer and the compute shortage curve keeps sloping up and to the right, SpaceX can keep raising prices." It cuts off at "On top of that, SpaceX has the best". Farzad RTd it at 10:31am CT.

SawyerMerritt, 11:57am CT, quoted Morgan Stanley's Adam Jonas in a new SpaceX note: "Getting other people to pay for your own compute infrastructure that you will ultimately use to train your own robots is a good idea." The note goes on: "We see the neocloud build as more of a 'funding strategy'". The post cuts off soon after. Farzad RTd it at 12:00pm CT and replied: SpaceX is laying out the foundation of the next generation of the economy.

The lease and pricing claims are jukan05's. The funding strategy read is Jonas's, via SawyerMerritt. The foundation line is Farzad's.

Earlier, at 8:35am CT, munster_gene posted: SPCX should be up 30% today. His case: "analysts cut the low end of his guide in half." The guide is Elon's line that SpaceX would exit 2027 at a $500B-$1T revenue run rate. Farzad RTd it at 9:12am CT. The 30% is Munster's call.

Check jukan05 here. Farzad's RT here. SawyerMerritt here. Farzad's RT here and reply here. munster_gene here. Farzad's RT here.

"Neoclouds have to go to spot pricing"

amitisinvesting, Oct 21 at 11:28am CT: so $SPCX obviously blew the doors out with those earnings

The post says "the question now seems to be what this means for every other neocloud" and that "SpaceX is up today, the rest are flat to red."

Farzad replied at 11:36am CT: Neoclouds have to go to spot pricing for them to make sense I think like the All In guys said.

Spot pricing means selling compute at the going market rate instead of under fixed long contracts. His reasoning: "Borrowing at higher rates vs SpaceX, deploying slower than SpaceX, and taking in lower revenue than SpaceX per watt doesn't feel smart." Then: "You can at least fix one of the three." The post cuts off after that.

amitisinvesting followed up at 11:37am CT, asking whether other neoclouds look attractive or "it converges into the company with the cheapest debt and best spot pricing". Farzad answered at 11:42am CT that he thinks it converges on whichever company handles the power shortfall best. In theory, he wrote, that holds "as long as AI demand is insane, as long as you can borrow at REASONABLE rates, and can throw up compute relatively quickly" while getting past "weird NIMBY-ism." NIMBY stands for "not in my backyard," meaning local pushback against new projects. The post cuts off mid-sentence.

Couch_Investor, 9:24am CT: "The problem is the money." The post asks: "Where do you get the money to build 20 GW, or even 10 GW, of compute capacity in a year?" Farzad replied at 9:53am CT: I don't think Elon is gonna have a problem raising capital.

At 8:34am CT, Linahuaa wrote that others "have as much compute as Elon but need it for themselves" and that "nobody is using his AI (because it's inferior)". Farzad replied at 10:30am CT: This makes 0 sense. His argument: "If SpaceX has a compute advantage, then by default they have a model advantage long term because they can do training runs and serve the models at cost, whereas everyone else who doesn't do this has to pay a premium to SpaceX."

These are Farzad's opinions. None of the posts give actual borrowing rates or revenue per watt for SpaceX or any other neocloud.

Check amitisinvesting here. Farzad's reply here. The follow-up here. Farzad's answer here. Couch_Investor here. Farzad's reply here. Linahuaa here. Farzad's reply here.

"Bandwidth creation for agents"

SawyerMerritt, Oct 21 at 11:52pm CT: SpaceX has just officially received FCC approval to acquire three additional spectrum licenses from Tristar to expand its upcoming next-gen @Starlink Mobile network.

Spectrum is the slice of radio frequencies a wireless network is licensed to use. SawyerMerritt wrote that the licenses cover 15 MHz of spectrum in the 1695 to 1710 MHz band. The post also says the FCC granted waivers, but the text in this pull cuts off before saying what they allow. Farzad RTd it Thursday at 5:39am CT.

Three minutes after the FCC post, at 11:55pm CT, SawyerMerritt posted: I don't think people appreciate yet just how disruptive @Starlink Mobile will be. Farzad replied Thursday at 6:40am CT that it's "less about" Starlink Mobile itself and "more about bandwidth creation for agents IMO."

He'd laid out the idea at 5:52am CT: Elon's secret is that he keeps thinking about the next bottleneck.

In his telling, a typical CEO would look at the bandwidth bottleneck with agents and say "that's probably decades out no sense to worry about it." Elon, he wrote, says "man that seems like a huge problem in a couple years as agents" and the post cuts off there.

That's Farzad's read on how Elon thinks, and the agents angle on Starlink Mobile is his. SawyerMerritt's FCC post, as far as this pull shows, doesn't mention agents.

Check SawyerMerritt here. Farzad's RT here. SawyerMerritt's follow-up here. Farzad's reply here. The bottleneck post here.

"Do you care if it's real me or not real me?"

Farzad, Oct 21 at 11:19am CT: I had a very interesting off-line discussion with Dave Lee @davelee on my use of an AI voice.

He added that he had his Grok Bot video process make a video about it.

davelee, 11:37am CT: The future is getting weird. Had a discussion with @farzyness where I was critical of him using his own cloned voice in videos. Later in the post: "And his cloned voice isn't bad? Or is it? I'm torn." Farzad RTd it at 11:43am CT.

At 11:49am CT, Farzad summed up the disagreement for HamiTahm: His main pushback was that there's a break of trust if you know the voice is an AI clone. My pushback is that as long as the content is good, the voice is a signature of the content that comes from me (aka my ideas).

His reason for the clone, in an 11:47am CT reply to rodamn: "I have about 10-20 ideas for content per day but only the energy to record 1." SulfuricDrake, 12:24pm CT: "I genuinely doubt all 10 of your daily ideas are worth broadcasting". Farzad, 12:40pm CT: "it means that I can at least make 10 different videos per day and I can publish the best one."

nerdalert, 2:13pm CT, wrote: "I'm much less interested in it than his actual original content that was a bit more unscripted and sloppy." Farzad, 2:35pm CT: Real me is not going fully away. But AI-me can do 100x more than real me, especially if it's educational/informational/analytical, which means that as a % of total content, it'll be far less real me.

ICannot_Enough, 2:52pm CT: "Doesn't bother me much, but I can tell it isn't you." Farzad, 2:54pm CT: If the content is valuable, do you care if it's real me or not real me?

The 100x is Farzad's estimate.

Check Farzad's post here. davelee here. Farzad's RT here. His HamiTahm reply here. His rodamn reply here. SulfuricDrake here and Farzad's reply here. nerdalert here and Farzad's reply here. ICannot_Enough here and Farzad's reply here.

More from the feed

ALScyborg, Oct 21 at 10:55am CT, posted Neuralink This Week, episode 1. The first item: "@elonmusk is now officially listed as Neuralink CEO". The episode also lists DJ Seo on the "all green light schedule". Those are ALScyborg's items. Farzad RTd it at 11:32am CT.

TheEVuniverse, 2:16am CT Wednesday: "honestly, I think I've never seen more people rave about a product here than what I'm seeing about Grok @Bot". Farzad quoted it at 12:14pm CT: AI agents are probably one of the stickiest products of all time. His projection: "Humanoids will be that in the physical world." His example: "Hey, do all the laundry and clean the house thanks."

Thursday at 7:04am CT, Farzad posted: Doing a deep dive on turbine blades and my mind is melting beyond belief.

At 7:46am CT Thursday, stevenfiorillo posted "my new investment thesis on $SPCX" and added "btw I am bullish." Farzad RTd it at 8:29am CT. At 8:35am CT, stevenfiorillo asked Elon: "What if the rest of us helped make it two?" The "it" is the one SpaceX share stevenfiorillo says Elon has pledged to every kid in America next year. The post asks whether there could be "a fund where shareholders donate $SPCX" before the text cuts off. Farzad RTd it a minute later. The pledge is as stevenfiorillo describes it.

Check ALScyborg here. Farzad's RT here. TheEVuniverse here. Farzad's quote here. The turbine blades post here. stevenfiorillo's thesis here and Farzad's RT here. stevenfiorillo's fund post here and Farzad's RT here.

Check SpaceX Is Very Undervalued, Wednesday at 4:50pm CT, here. Tesla Left Whole Parts Off Its AI5 Chip on Purpose, Thursday at 8:32am CT, here. Google & SpaceX Just Took the AI Race Into Orbit, Thursday at 8:40am CT, here.