Click betterhelp.com/tsyou for 10% off your first month of therapy with my sponsor BetterHelp. Join over 4 million people who’ve met with a therapist on BetterHelp and started living a healthier, happier life.

Notwithstanding investment in this particular AI chip Breakthrough and its potential Bullish, I still feel investors should be focusing on under-the-radar stocks, and considering the current rollercoaster nature of the stock market, Because 35% of my $270k portfolio comprises of plummeting stocks which were once revered and i don't know where to go here out of devastation.

Man it gets so infuriating how much misinformation is out there about markets and investing. Thank you for shedding some light on how ridiculous KZfaq finance really has become. I followed these advises and ended up with bunch of stocks in my $205,000 portfolio, not knowing when to take profits.

I agree that many people are considering NVDA as the "Stock of the year." However, I'm curious about which stocks could potentially become the next META in terms of growth over the next decade. I've allocated $200k for investment, aiming to retire comfortably.

No, it's not a big threat. The market for AI is so huge that AMD will do just fine. It's amazing how so many continue to underestimate what Lisa Su brings to the table. What she's accomplished at AMD since becoming CEO in 2014 is incredible. And she now has a team of over 25,000 employees, we haven't seen anything yet. IMO she's just getting started.

Is it just me or is anybody else mind boggled by the fact that 1.4 exaflops is in a single rack cabinet?!? One of the biggest supercomputers is literally 1.2 exaflops. Now they probably achieved this metric by scaling down neural networks to 4 bit floating point, but that just shows how they optimize software to reduce compute time and power to make better neural networks. 25x less power for way more compute... Juggernaut.

Connecting two dice into one device was done previously by Apple in its M1 and M2 ultra machines (and soon M3 ultra). This is essentially the same thing. Not to diminish it, but it's not the first. And, DPUs are not "more parallel" than these GPUs. In San Francisco, there are tens of thousands of homes that have Google fiber, 1Gb/s up and down, symmetric. That's 125 MB/s, up and down, or 250 MB/s per connection. If there are 100,000 such homes, that's 25TB/s aggregate bandwidth. If there are 10 such cities--there are far more than that in the West--that's 250 TB/s. So, no, the bandwidth in the system that Jensen talked about does not have "more bandwidth than the Internet." Now, he might have meant something like backbone bandwidth or some other qualification, but that's not what he said.

1st vid of yours I've seen Great stuff!

One thing that I found interesting is that Nvidia named their chips after famous mathematicians and physicists. For example, Grace is named after Grace Hopper, who invented the first compiler and coined the term “bug” for a computer error. Hopper is named after William Hopper, who developed the Hopper matrix, a tool for solving linear equations. And Blackwell is named after David Blackwell, who made contributions to game theory, probability theory, and Bayesian statistics. I wonder who they will name their next chip after. Maybe Turing, Ramanujan, or Einstein? Anyway, great video and keep up the good work! 👍

Great content as always bro, I remember when you first came out, knew u was gone be here to stay, I do miss your OG thumbnails though

Agreed. Bought Mellanox for $7 billion seemed so cheap today. Without its own high-speed high-throughput networking, NVIDIA margin would be much lower.

thanks for your work great video....

One of a few people out there doing quality work, appreciate the good investigation.

Well done - good vid. The mood of the staff working there is also so important for teamwork and success. An independent visit to the works could be revealing. Google is known for its staff friendly approach and the enterprise of their execs too.

Great video, thanks for sharing 👍🏽

I was part of that Mellanox acquisition. It has paid off handsomely.

Hvala!

You have a great channel man, thanks so much for your time.

Your list of top stocks looks similar to the typical go to tech portfolio. However a few exceptions. First, although overvalued by now, what about the companies supplying all of these server farms? SuperMicro, Dell. Taiwan semi makes Nvidia chips, where's TSM? The AI chip design is also dependent on ASML and Lam Research. Fortinet was a 2023 play for cyber security. Palo Alto is fine, but replace Fortinet with Crowdstrike. Memory ? How do you build server farms without DRAM or storage? Add Micron and Western Digital. Although early keep an eye ARM. Palintir? Oh god! The sleeper AI supposed powerhouse that has more PR that stock stability. This stock cannot get traction and is just a short term option play. Hopefully their military contracts can pan out cause we all need AGI powered automated military weapons, right? not!

"Wall Street does not understand NVidia" - this was clearly demonstrated when the stock price basically shuffled sideways instead of exploding to $1500+ after the announcements.

Yes comparing in terms of times or fractions is easier to picture than just raw data speed.

Hi all Medium to long term targets? For Nvidia share price

Thanks bro for the reassurance. One thing tho, what about groq and other potential specialized chips that are to come?

Great info. Thanks

amazing vid man ! i appreciate it

Found you after clicking randomly on a KZfaq suggestion. Liked and subscribed after the first 30 seconds precisely because you're respectful of my time. Good job, keep it up!

Nailed it, Bro !

Hopefully we get a stock split!

Would compare LLY and Nvidia then look into your crystal ball for short and long term growth? Thank you

My three favorite channels: Ticker Symbol: YOU, Stock Brotha, & Everything Money. Love them all! ❤ ❤ ❤

Most of the performance improvements is because they went from an 8bit architecture to a 4bit architecture and also glued 2 chips together. 4 bit architecture will be less accurate for AI training. This may be ok or maybe not. So overall not much of an improvement (only a few percentage).

AMD and Intel will definitely have their share of the market. TSMC is at max capacity and investing in other semiconductor companies will be an absolute power move. Different chips are good at different things and Nvidia has been very specialised, which leaves other aspects of AI open.

great job Alex, putting it all together for the non professionals- Thank you

Awesome Bud! 🥳🥳🥳❤️❤️❤️

Dear Alex , please could you share your knowledge on which apps/trading platforms are the safest yet also easy to use / beginner level. ( into Nvidia ) Thank You . from Australia

your videos are always optimistic. Can you ever do one where you highlight risks potential bubbles

What do you think of aisp? A good stock or another lcid?

I think in one year Tesla stock price $ 120 Nvidia $ 2,200

Great video! Thanks Alex!

Thanks for providing the great content as always!

great video dude

Excellent video! Consider buying NVDL which is a 2x leveraged ETF of NVDA. The way this ETF works is that for every percentage NVDA goes up or down, NVDL will move 2x on the upside and downside respectively. NVDL closed at 41.50 on 29-Mar-24 while NVDA closed 903.56.

another great video! well done!

How’s much chips would they need and how long would they last to buy new

THANKS

Amazing video! Thanks Alex

Ya, ya ya, but can these DPU's do color graphics?

We gonna have to pedal really fast to power theses things

Can we have a more rounded view ? Competition, markets size, risks etc.

Keep bangin' Alex! Thanks for bringing NVIDIA's exploits to my attention, I responded immediately. NVIDIA's scale reminds me of Apocalypse in the X-Men series "I am as far above Mutants as Mutants are above Humans" "I am the Rock on the Stormy Shore... CRASH INTO ME AND BE BROKEN!!!" --Apocalypse I remember the 486 DX Days of Math Co-Processors, PSPICE Circuitry computations were entered in CLI format and a simple Class A Amplifier simulation took HOURS to compute. Today, creating the circuit is GUI based, intuitive and the simulation will run in a few nanoseconds. Wall St. can't even flowchart a simple Y2K fix. Let "Y" equal Year/ increment Year/ IF "Y" is >< 2000/ Then LET Y = 2000. Such a hoax! I was broke then.. I have "Dry Powder" today.

I will buy the dip. Nvda run so much this year for sure will come down at some point when the market goes bearish.

Still interested to know how these chips are earning profits for companies, but I guess we'll figure that out later. What are the applications?

Hi You I wrote the same thing last week that Nvidia will be the largest company by Mcap by 2025 Christmas. And with AGI already here and Microsoft and OpenAI spending $100 Billion on StarGate Super Computer with a million plus chips in it and there are several of these. So they don't spend that much money without AGI, and now we know that ChatGPT 7 is nearly ready and they have said that has AGI, and i'm not talking about the GPT 5 that's on media, but the GPT 7 version. EDIT: I thoroughly enjoyed GTC.

A lot of investors just look at graphs and try to predict what the graph will do based on weird graph shape mumbo jumbo. It's like reading star signs and totally disconnected from reality.

Thank you so much.

Pls - what is black wells ticker symbol or vidia ticker symbol

The so called analysts don't understand a fundamental law of generative AI - all other things made equal, increasing Scale of compute makes the AI more intelligent and they are needed yesterday for any company that wants to build a leadership LLM. NVidia is giving them the solution and also has an upgrade pathway as Scale continues the go up. Any company or sovereign country that wants to lead in LLM are compelled to buy NVidia solutions ASAP or be left in the dust by their competitors. A new purchase, in turn, creates stickiness on upgrade paths into the forseeable future. Instead, these talking heads are busy talking about comparing NVidia to CIsco in 1999 bubble, talk of the Osborne Effect, the potential rise of other GPU contenders etc etc.

Alex you should look into Astera labs and Soun thanks

amazon bought a bunch of pico vr headsets and we used them for traning and it was alot faster than their old way of traing its will surely take over

Why not run 8000 Blackwell gpus, refill displacement, and see the total power draw vs Hopper In that manner....??? They cut displacement, then sell the same rate of speed. Vs power draw and space savings. If the space was already created so was initial displacement Can improve architecture but what are they doing making super computer laptops or refilling the space with modern architecture to improve performance in the same 8000 gpu package, sell em on space savings architecture, and matching speed with an old architecture while reducing space. Obviously reducing power consumption, but ok. Same compute power now Mobile, what about displacement like sli, and the 4090. Tdp of 450w, 525w load, two 1050w, looks like quad sli displacement, 4 Titan v same at 250w tdp and while 4 drawing the same..at least you get your property bsck. Mini atx laptops or full atx etc with 2x 90 series.... Sli works best on maximum vram and memory architecture, expanding gfx anti-aliasing, resolution. Gfx software game modding, ultra detail settings future proofed with thise settings, bragging rights, fun, and longevity X2 of everything and not offered 4090 is basically two 4080 crammed together at 450w tdp instead of 250-280w single 4080, and agian no SLI SO NO REPLACEMENT FOR DISPLACEMENT. 2 4090 basically would be quad sli.. at 1050w full loaf. 5 series ect. But there playing a game selling reduction when in fact the same amount of power can be placed to create more gpu rendering power at the same wattage as before, double the total vram, even though vram.per gpu is the same. Fill rate, cuda cores, raytracing cores. Pixel rate, 2x rendering at same wattage as quad sli being what's missing along either 3 way 4080 series. But ai so great why doesn't it write excellent driver sets for sli, along with a hardware style nvlink on the mothrrboard 5090 in sli would be wonders. No replacement for displacement and given power factor no matter the architecture change upgrade Why the 9800 gtx uses 160w , and 5870 188w,'250 w overclocked. Why a 3dfx voodoo 5 5500 had two chips at 15w a piece for 30w, but definitely tdp displacement While a voodoo5.9000 has 32 way sli 32 vsa 100 chips 2 gb-4gb ram depending on model, and with the correct mothrrboard windows xp dx 9.0c and driver sets, crysis runs better than a mondern sli aetup from 2007 lol. 32 chips 60 watt 4 chip lanes 8 lanes wide What's the tdp Agp power plus 4 molex About 420-450w tdp There was also a v5 8000, 8 chips, setup in sli, for 4 cards. To get 32 way sli... Sli really only needed to stretch the limits of rendering computing and provide extreme visuals as a gfx workstation studio Gaming systems. 8k with 8x or 16x anti-aliasing with filters maxed and ultra gfx with gfx mods are what's good now. Hitting beyond a 32k filtered resolution on 8k, but where is the sli to enhance 8k max detail settings and gfx modding with raytracying and some are rich, what about 16k resolution doing the same. 4000 dollars Two 5090s you think it would be achievable. Mabe on 6 or 7090 when they decide to actually add in video memory rather than miniscule increases creating g p u bottlenecking at higher res or detail anti-aliasing settings, wheb in fact the gpu had enough steam.when v ram becomes literally upgraded off the board and exchanged. Tdp Sli Vram Memory architecture Imagine two 5090 you could afford even if the price tag is 2000 a card when new, at say 48 gigabyte vram 512 bit bus, upgraded cudas raytracing cores ect, two in sli would be basically 96gb ram Even though 48gb per gpu, and 1024bit even though 512, with all the ai and frame generation, i don't see what the issue is other than displacement getting cut to basically hidden sli between the 90series ad 3090-4090- and hopefully changes with 5090 and up. I want my third and 4th performing displacement, Sli on the top end card will stretch the limits of rendering power not only at available resolutions bit with antialiasing, filtering, uhd file sets, gfx modding the game, under ultra quailty presets. All the rgb, like the guys out there with fake fast cars acting as if there fast. It's almost embarrassing to know the truth, and the fact they focus less in the advantage of pc and more on other systems not in the sane form factor with much more diminishing returns. 9800gtx 3 way sli 1.5gb ram While 8800gtx was 764mb ram, 3 were 2.3 gb Crysis Maxed ultra settings and being a gfx molded game. Eveb more so, ran into issues with frame buffer or v ram right at 1920x1080 4x aa, 16x af, but worked at 2x aa and even better without. The 8800gtx, would run these settings 3 way and more fps on 4 way due to extra memory and its memory controller. Solid 40fps or more without hitching and 7fps drops like the newer cards they sold. 9800gtx Same trends seem to continue and have become worse now with no sli on newer architecture, messing with MAXIMUM DISPLACEMENT. Increasing displacement and architecture but then removing MAXIMUM DISPLACEMENT 3 WAY SLI QUAD SLI I HAD THREE 5870 NEW FIRST MONTH, TECHNICALLY I HAD A FASTER SYSTEM. GIVEN I COULD RUN 3 4080S NOW OR THREE NEW 5080S OR 2 5090S OR 2 4090S Start looking at power draw and displacement among form factor, the curve is a trick I'm the illusion Can't replace power factoring, why a laptop is a laptop , desktop is a desktop, supercomputers are supercomputers size of rooms or buildings If I could have a 25 watt video card that ran photon mapping ray traycing at 500 fps and 16k resolution, it must be one hell of an architecture, what if we make it larger jack it up to 450w tdp, and then add a second card to link between those two, We could get 276k resolution with 32x aa at 240hz Newer quantum architecture, just replaced displacement at 1060w with it though....

The earnings need to double in the next couple of years in order to justify this valuation. I think it is possible but is it probable? - I am not sure.

But why did they sell denmark the old one to use to build a new supercomputer that is just not Nice.

Great video. I think Nvidia has a very bright future and I also see it being number 1 company some time soon too. I've got a diversified portfolio currently. Different growth stocks, ETFs, dividend stocks etc. But my 2 highest conviction stocks currently are Nvidia and SuperMicro. I invest 50/50 into both of these daily. But have been thinking about about aiming more towards Nvidia for higher weighting. Any thoughts would be great.

How much of their stock do you own? Seems like a lot of your investment is riding on this

Why do you prefere fortinet and palo alto over crowdstryke?

Thanks!

I agree that Al stocks will dominate 2024. Why I prefer NVIDIA is that they are better placed to maintain long term growth potential, and provide a platform for other Al companies. I know someone who has made more than 200% from NVIDIA last year.

But what do you think about Groq and LPUs? Is Groq going to impact Nvidia's AI chip demand?

Should call them cabinets, instead of racks, the racks slide inside the cabinets.

Is there a particular reason you PANW and FTNT instead of CRWD? Just curious.

Agree!!!

You can see the limits of scaling this technology right there in front of you: the silicon estate in use shows that this is at the limit.

These chips suck for inference and if open-source keeps kicking closed-sourced cost effectiveness out the door = the demand for training will be tiny and crowdfunded both BDRX and Dolphin-Mixtral beat gpt3.5 and in many scenarios gpt4 and they can be run on a M3 Max way faster than you can read...

My top 3 investments are: Nvda, tsla, crowdstrike…

No mention to NIM? It's the real deal that AWS fast went to post about once he mentioned it on stage😂

Why dont they integrate the two dies bypassing the middle communication chip, making the dies into one single oblong die, the same as the did with the first die?

The fact you didn’t hold me until the end of the video was enough for me to hit the Subscribe button and like the video 👍 Awesome channel.

Nice overview. Own a boatload of NVDIA MSFT and AMAZON already - and it has been amazing - so not adding even more, still I hope you are right. The counter argument is that there are many out there who are trying to copy/best NVIDEA’s hardware advantage. May take time, but not sure the company can lock in their customers like say Apple were able to.

damn this looks cyberpunk, rather dystopic looks, good to see it being more energy efficient, but looking at the whole picture inncluding how people in the room react to thigns sounds insanely dystopic already

Holy shit, I knew it. At this rate we will hit well into hundreds of zettaflops or even hit yottaflops of performance by 2030. As generous as my predictions have been, I think even mine might be too conservative judging by how fast this is scaling up.

Great video as always.

Great vid 😊

Nvidia's chip is deflationary. There is a market. When the price of the product decreases the size of the market increases. For example the cost to provide cloud services has decreased. The size of the building and the amount of electricity has decreased. It does not take much of a sales pitch to convince an accountant that these chips are a good investment. But wait there is more. More customers can now afford this product and can grow their business faster. This is the beginning of an investment cycle or upgrade cycle. At one time there was a hand held cel phone, then there was Blackberry and now there is Apple. This is one of those times you don't want to be in 2nd. place.

The speeds of these new products sound amazing. But how does it matter when everything else connected to it will bottleneck that speed. Storage devices, wifi, internet speeds. None of those will come close to utilizing those speeds for a long time

14:00. 120kW per rack. Going to need a nuclear power plant to power just that one room.

GenAI dev here. You're right. Nvidia has years on AMD. They aren't slowing down. It's a new age

This is all great information and true about the long term potential of the company. But right now are we in an AI bubble, is the big question ?!

So what exactly is our Super computer powered with?

potentially but right now i am just day trading it, cant hold to it, too unsure

I saw the video for Nvidia Blackwell a couple of days ago and I was thinking "Man, people don't understand what Nvidia has unleashed right now". I showed it to my wife and she also didn't understand the implications. To make it easy to understand. In the 90s we already had computer. If you used a 486 back in those days the processor had over 1 million transistors. With that you could write letters, make calculations in spreadsheets and other stuff but watching videos is impossible for example. Only one high end workstation could display a video full screen back then, and it wasn't in full HD. Today you would complain about bad video quality. Today most of us are using computer power which is 1,000 times greater as back then. It gave us the possibility of full HD video and using software with AI algorithms. Now think that Nvidia's new chip is again 200 times faster and I'm talking of a single GPU only. The board is using two of them. This gives a whole lot more possibilities what we can do which we didn't ever imagined. Also AI revolution has just began. The next 10 years we will see drastical changes and Nvidia will be on top of everything. I'm pretty sure of that!

The great content.

This will not doubt come across as an amateur hour question, but I purchased Nvidia stock at $41 / share. How can I now justify buying more of it at 21.5 x that price? I can't do it! Great video as usual.

Munger and Buffett have both achieved an incredible feat with Berkshire. They've turned thousands to billions, and have made a lot of people wealthy in the process. I really saw the potential of the stock market by reading Berkshire's annual letters. I recently sold my $674k apartment in the Bel Air area and I'm hoping to throw it into the stock market. I just don't want to lose everything.

Yes I admit that Nvidia is doing pretty well. But there is another point, Wall Street is also missing: this company is fabless and totally dependant on TSMC. Intel is not. Keep that in mind.

I couldn't agree more with your analysis. AI is going to transform most businesses. The larger wealthier companies will be the first to adopt and implement the tech. Other companies will be forced to adapt to using these tools or be left behind, lose out on efficiency, growth, advances. It will be similar to the internet but on steroids. The beneficiaries will be NVDIA, a few other chipmakers, large existing companies, FB, MSFT, GOOGLE. We'll see what other businesses choose to adapt and innovate or who wants to be a dinosaur.

It would be fascinating if the US government had this similar tech decades ago but now the consumer market is just catching up.

On your right = His left hand 😊😍 but... Otherwise... great info - hope you're right about the value / price - thanks

Alex -no one’s answered the question - which companies benefit most from their association with the Blackwell chip and Nvidia ??? There has to be some smaller cap businesses that do ??? 🎉🎉

It all comes down to performance and longevity.....

Agreed!