Paperviewr

@WatcherGuru Fun fact less than 3% own semiconductor stocks directly globally

FUN FACT: Less than 10% of the global population is invested in crypto.

@SilbergleitJr POOSEE YOU ARE

I don't want to be a party pooper but just keep in mind that Semiconductor ETF $SMH went -80% between July 2000 and October 2002. Those who bought at the peak in 2000, did not break-even until late 2017. So technically anything is possible, just saying.

@StonkChris I read your msg as in “please god let me in semis I don’t want to get left behind”

Feels like semis are setting up for a potential blow-off top, if not today, then very soon. This move is getting a bit too vertical…wouldn’t mind seeing the sector take a breather here. $SMH is getting stretched.

@Tortoisehead @CattardSlim Here’s the problem. You can’t convince millions of hardware 3 owners to take the upgrades. You will end up with class action regardless

@CattardSlim I think they will avoid lawsuits by offering free upgrades

How has Tesla not been sued out of existence? It's like selling millions of boats that don't float. Constantly promising they will soon, & then years later finally admitting they'll never float, but keeping all the money anyway.

@SkipperStagg @MissJilianne Fsd on hw3 stands for Failed Self Driving

@MissJilianne HW3 has FSD... Seems like he did know.

Elon didn’t know what he was talking about.

@spotted_model @diaper It won’t matter because a lot of people won’t accept that offer and it will go to class action regardless

@diaper Yeah I think it should transfer

HW3 FSD owners: How much credit on a new HW4 vehicle trade-in would it take to get you to make the switch?

@Nick86LGM @MissJilianne Guess I’m not selling until I get max profit from a lawsuit

@MissJilianne They will push this upgrade until the vast majority of HW3 owners have sold their cars and moved on.

Tesla no longer offers the Tesla Model S Plaid, and I’m not interested in downgrading to a Tesla Model 3 or Tesla Model Y. I paid for Full Self-Driving over three years ago, and I expect to receive the Unsupervised capability on the car I already own.

Easiest way to reimburse 4 million hw3 owners is just sell 500 billion of stock right now

@EitanPodMD @SawyerMerritt It’s easier to just pay everyone who has hardware 3 out. Upgrading is a waste a time send everyone 50k and be done with it

@SawyerMerritt The discount to upgrade for those who bought FSD should be the price they paid plus the compounding effect over the years. likely 3-5x from the price they paid for the discount - nothing less. what are your thoughts

NEWS: Elon Musk has confirmed that HW3 can't achieve Unsupervised FSD. "For customers that have bought FSD on HW3, we are offing a discounted trade-in for cars that have AI4, and also offering the ability to upgrade the cars computer and cameras. We're going to have to set up micro factories in major metro areas. I do think over time it will make sense for us to convert all HW3 cars to HW4."

All hardware 3 cars should get a trade in for hardware 5 Tesla cars

@RealAndersonBen @JonBbC_TechGeek just like they use to say AI3 can do autonomy too. Lol. do you really believe him?

They just said during the Q1 earnings call that AI4 will run v15 and that v14 or v15 will be unsupervised once regulations allow it. AI5 is coming soon, and it will be a huge upgrade, but it's not needed for unsupervised FSD. AI4 can already do it. I don't understand why people want so badly for Tesla to let their hardware stagnate. The people designing the chips are not the same people working on FSD. Both teams continue to iterate and improve, as they should. AI5 is primarily being developed for Optimus, anyway. It's only going into cars because it makes sense to ship the best chip they have. But AI4 is all they really need.

Let’s be honest. AI4 is cooked. Back in 2019 and 2020, people raved about HW3 in a very similar way, and it was hyped even more than AI4 (by Elon and Tesla). Based on history, the maximum life cycle for a computer like this is 4 years. Guess what? AI4 is approaching the end of its lifecycle. I know this is a hard pill to swallow, and I know this post could have a negative impact on Tesla in the short run due to the Osborne Effect, but it makes literally zero sense to buy an AI4 car right now. The only exception that’s possibly valid is if you can afford TWO new cars in the next 2-3 years. In that case, knock yourself out! You totally deserve it.

the power of semiconductors.

@SpecterAnalyst Wouldn’t be surprised HTX going to bankrupt this year because of this

Tether has just blacklisted two addresses holding a total of $344M—the largest USDT freeze to date. Wallets are hops away from an OFAC-sanctioned entity, the Huione Group, an entity known for laundering stolen funds on behalf of threat actors. TNiq9AXBp9EjUqhDhrwrfvAA8U3GUQZH81 - $212,922,653 TTiDLWE6fZK8okMJv6ijg42yrH6W2pjSr9 - $131,288,800 it looks like LE took over the wallet years ago... look like a seizure

@lookonchain @tether Oh man. Justin Sun is screwed. They going to pin something on him for sure

Tether(@tether) froze 344M $USDT held across 2 wallets 1 hour ago. tronscan.org/#/address/TNiq… tronscan.org/#/address/TTiD…

@WatcherGuru Welcome to crypto

JUST IN: 🇺🇸 Tether freezes $344 million in $USDT following requests by US law enforcement.

@Arronwei3n Now I know your retarded that memory is 2-3 years old. Everyone’s moving to hbm4 already

@Paperviewr No, they are using HBM3e. New Ass Memory.

So we are now expected to see 1:1 = TPU: Axion CPU configurations? _ $GOOGL Debuts TPU 8t and TPU 8i At the Google Cloud Next ’26 event, Google unveiled its next-generation TPU series: TPU 8t and TPU 8i. The key differences between these two TPUs—and versus our prior assumptions—include: Google has split the TPU lineup into dedicated SKUs for training (8t) and inference (8i). It is reported that MediaTek (MTK) is handling the training SKU, while Broadcom (AVGO) is responsible for inference. The HBM capacity for the 8t (Zebrafish) is 216GB, higher than the previously estimated 144GB. The TPU-to-CPU ratio for TPU 8i may shift to 1:1, compared with 2:1 for TPU v7 (Ironwood). If these assumptions are correct, this would be positive for MTK. The conventional view has been that MTK focuses on lower-end TPUs, but it now appears it may be involved in training chips instead. Additionally, Axion 2 CPU shipments could see meaningful upside due to higher CPU-to-accelerator density, while HBM demand should benefit from the 50% increase in memory density for Zebrafish.

@EricJhonsa Hbm3 is not innovating bro they are using 4 year old memory

When it comes to hyperscaler AI ASICs, there's $GOOG (innovating across systems, networking, memory/storage, cooling, etc. in a way that maybe only $NVDA can surpass) and then there's everyone else.

@bdinvestingg It’s HBM3 pretty old ass memory while nvidia going to hbm4

New $GOOG TPU bullish for memory $MU , Increased HBM usage in the new chips: The TPU 8i features 288 GB of HBM per chip — a 50% increase over the prior Ironwood generation (which had 192 GB).

@mweinbach hbm3 tf

If you’re interested, first is TPU 8i and second is TPU 8t These are the block diagrams of the chips. They are very different