Broadcom vs Nvidia: The AI Chip War Explained
From GPUs to XPUs: The Shift That Could Redefine Tech Giants and Your Bank Account.
This chip just got real.
This isn’t me trying to get clicks either.
You won’t find an ad or affiliate link anywhere.
This is about making sure you know something huge happened last week.
And if you fully understand what’s going on - you could make a lot of money.
There’s layers to this, so let’s peel back the onion so you have a clear understanding of what’s going to happen next.
Let’s begin with Nvidia NVDA 0.00%↑ .
It’s the Michael Jordan of the GPU business.
You don’t need to know what a GPU is, but if you have a general idea that’s great.
The main point is Nvidia has been working on GPUs for longer than any company and has competitive advantages that other companies don’t have.
In the past GPUs were used almost exclusively for video games or to deliver superior graphics to a computer screen.
But when AI startups found GPUs were superior to anything else in the market - Nvidia’s addressable market exploded.
And shares of Nvidia stock skyrocketed.
The competitive advantage was so great for Nvidia, the company could basically charge whatever it felt like.
Not only that, but many of Nvidia’s customers were the richest companies in the world.
Microsoft, Amazon Meta, Google have hundreds of billions in cash along with wildly profitable business models.
This essentially allowed Nvidia to print money over the past few years.
Sounds great right?
Not so fast.
That’s because Nvidia made a critical mistake.
I saw it coming over a year ago.
I wrote about it here.
Instead of catering to the desires of the largest cloud computing providers, Nvidia made investments in smaller cloud companies like CoreWeave.
It allocated the red-hot GPUs to companies many people never heard of.
Industry sources indicated that startup cloud providers like Lambda and CoreWeave were getting more allocation from Nvidia than Google and Amazon.
I said a year ago Nvidia should have catered to the megacap companies instead of playing silly games.
Lock these filthy rich companies into the Nvidia ecosystem forever.
But that didn’t happen.
For years both Amazon and Google have been developing their own in-house custom silicon designs.
In particular, Google’s Tensor Processing Unit or TPU was on par with Nvidia’s offering.
The difference was Google wasn’t selling its TPU to other companies, instead using it for its own Search, Maps and YouTube services.
Google’s secret weapon?
Broadcom AVGO 0.00%↑
The two companies collaborated to create Google’s custom data center chips leaving Google in a unique position.
It didn’t need Nvidia.
It also didn’t need to pay 75 cents of every dollar to Nvidia’s bottom line.
With AI being a game of shoving as many computer chips into a data center as humanly possible, shaving costs was a wise move.
This week it was revealed on the Broadcom conference call that virtually every megacap company is now going to copy what Google did.
We’ll unpack what $60 - $90B in revenue opportunities in 2027 mean for Broadcom in a moment.
But it gets better.
Broadcom CEO revealed the $60 - $90B in revenue might actually be conservative. That’s because Broadcom was selected by two additional hyperscalers and are in advanced development for their own next generation AI XPUs.
Naturally Broadcom stock skyrocketed on this news.
This is extraordinary guidance.
The salesman in me might want to mention I recommended buying Broadcom to Brownstone subscribers in August 2023 when shares were (split adjusted) $89.
And again about a month ago on the Equity Empire YouTube channel.
But my main mission is to teach you how picking stocks can be this easy.
Just give up watching all TV except Cincinnati Bengals football games and study the semiconductor business for a few years.
Luckily you don’t need to give up TV since I’ll break it down for you now.
Broadcom’s 2027 outlook is absolutely a game changer.
I hate the term “game changer” … only because it’s an overused term and it sounds kind of cheesy.
But it’s warranted in this moment.
That’s because Broadcom’s revenue for all of 2024 was $51.6B.
58% of the company’s revenue comes from semiconductors and the other 42% comes from enterprise software.
More interesting than that was the company only had $12.2B in semiconductor revenue that was related to AI in 2024.
Do you realize what this means?
Broadcom just told investors that AI revenue is going to go from $12.2B to potentially $60B or higher by 2027.
!!!!!!!!!!!!!!!!!!!!!!!!!
Keep in mind that Broadcom is calling the $60 - $90B “serviceable addressable market” or SAM.
It doesn’t mean they’ll get to $60 - $90B … but rather that’s the opportunity.
Analysts have chimed in and estimated that Broadcom could capture 70% of this market.
Even if it’s 50% it’s nothing short of extraordinary.
If Broadcom can get anywhere close to $60B+ in AI revenue by 2027 it’s a 3x opportunity from here. I explain this in detail on the YouTube channel:
So now you know how the stage is set.
What’s Next?
That’s a good question.
2025 will still be Nvidia’s best year ever.
The Blackwell design is far superior to anything on the market and every company is knocking down Nvidia’s door for allocation.
That won’t change in 2025.
Nvidia will announce an upgrade to Blackwell in 2025 that will probably ship in the first quarter of 2026, but it will be an incremental improvement rather than a giant leap.
Nvidia could get every hyperscaler to pump the breaks on developing AI chips with Broadcom by slashing its gross margins, but I would be shocked if that happens.
Instead, Nvidia will rely on the “total addressable market” to continue to expand.
That’s not the worst strategy.
If I’m correct and robotics becomes a huge area of investment, there should be plenty of market to address.
Broadcom should have a decent 2025 without the uplift from custom silicon coming until a year or two later.
That’s because Broadcom has an AI data center networking business that can win alongside Nvidia winning.
If Broadcom remains on track to hit ~ $80B in revenue by 2027 the stock will track into the $2T - $3T valuation as long as the broader markets don’t take a dump.
All in, Broadcom is one of the few megacap companies with a clear path to double its valuation over the next few years.
Whereas for Nvidia to go from $3T to $6T … while not impossible … is obviously a taller task.
Who is the big winner?
What if I told you the real winners are not even mentioned yet? (!!)
During my last job they always wanted me to come up with these stupid marketing tag lines and messages.
I ignored the request most of the time … but I did come up with a good one.
Hardware … Software … Everywhere
It was cute because it not only rhymed but it was true.
The hardware phase of a technological rollout is exciting.
I remember getting an 18k modem, then waiting a few months and getting a 36k modem back when the internet was just getting started.
Personal computers, Cell phones and smartphones all had its hardware phase.
What’s interesting about this hardware phase is it’s all in the data center.
That’s why you have some investors claiming “AI is a bubble” or it’s “fake” … that’s because they don’t get to see & touch the hardware like we did with a personal computer or even the internet.
But this shit is real.
The software that exists today is quite good, but the rate in which it’s improving is remarkable.
Just as an example I sell baseball cards as a hobby.
Right out of college I opened a baseball card shop with my brother and it was a good idea until the Great Financial Crisis hit.
But for fun I still buy and sell baseball cards.
But it’s an annoying task to list 1,000+ baseball cards for sale on eBay. One that could take you several days.
Now there’s AI software that lists the cards automatically by recognizing the image and pairing it with a database of known examples.
You can now list 1,000 baseball cards in about 1 hour.
AI software basically allows me to run a full-time reselling operation in my spare time.
This is just one example.
Computer coding, coming up with headlines for YouTube, thumbnails, research, fact checking, etc is all being done with AI.
Where am I going with all this?
Good question.
This is the next phase of AI.
With billions being spent on chips from Nvidia and Broadcom - the future software applications will be extraordinary.
Pricing will continue to compress lower.
Access will become wider.
Industry will change.
Change = Opportunity
Trust me … there’s going to be some of you that are going to worry about people losing jobs or legacy industry going away.
Let me ask you this ….
When the internet was just taking off in the late 1990’s should you have worried about the print newspaper business getting demolished or invested in Google?
When e-commerce was taking off and it was clear retailers like Kmart and Sears were going to get demolished - should you have worried about that or invested in Amazon?
The choice is really obvious.
And if you get focused AI could make you a ton of money.
So let’s recap.
Nvidia is still going to have a great 2025.
At the very least the company will buy back an insane amount of shares to keep it performing quite well.
You continue to hold Nvidia.
As long as Nvidia doesn’t slash pricing - Broadcom stock could easily double in valuation by 2026/27.
There’s a ton of ancillary businesses that benefit as well, but that can be saved for another time.
All this funnels into software with Robotics being the holy grail opportunity.
Good news is I’ll be at the 2025 Consumer Electronics Show all week in January.
I’ll bring you all the interesting insights and stories from companies that stand to benefit from the AI chip war.
Make sure you follow the YouTube, Instagram and X accounts below for all the latest.
Thanks!
Colin
Thank you Colin! Given this, what happens to Nebius?
Thank you Collin. Greetings from Switzerland