If you invested $10,000 into Armsttock

when it went public less than 3 years

ago, you'd have $35,000 today. And if

you invested that money in Palanteer

when it went public back in 2020, you'd

have close to $150 grand. Well, this

company makes AI chips, that should be

physically impossible, and they just

went public. My name is Alex, and I

spent 8 years as an electrical engineer

and AI researcher at MIT, and I've never

seen chips like this. So, let me show

you what Cerebra Systems does and how

I'm investing in it. Your time is

valuable, so let's get right into it.

IPOs can make investors a lot of money

or they can destroy portfolios if you're

not careful. So, let's start with how

IPOs actually work. IPO stands for

initial public offering. That's when a

company starts selling shares on a

public stock exchange for the first

time. Before that, the company is

private. Private equity is usually

reserved for big institutions and

accredited investors that can afford to

lock up lots of money for a long period

of time. That's because most private

companies are still building their core

products or finding their first big

customers. So, they're burning cash and

raising money to survive while they do

it. The catch is that private companies

don't have to report earnings. They

don't have to go through audits and they

don't have to hit external deadlines. A

startup can show you a pitch deck

projecting billions of dollars in

revenue with zero obligation to show you

what they actually made last quarter.

Everything changes when a company goes

public. Public companies have to report

earnings every quarter. Independent

accounting firms audit their books and

every major risk, every major business

change and every dollar of compensation

has to be disclosed in writing on a

fixed schedule. Otherwise, the SEC comes

knocking. But here's the catch. All of

this starts after the company IPOs. The

day a company goes public, investors

only have the S1 form, which is the

initial filing that a company submits to

go public. The S1 covers the company's

business model, its biggest markets,

competitors, and risks, and some basic

financials. But what it doesn't show you

is how the company actually competes in

those markets, how they deal with their

margins going down, or if they'll ever

even hit their revenue guidance in the

first place. Investors don't find out

those things until the first real

quarterly report 90 days later. So when

any company goes public, the market is

buying a story. And that story comes

with real risks. That's why the pattern

for every IPO is almost always the same.

The stock skyrockets and then the real

clock starts. Industry analysts start

comparing them to companies with

stronger numbers. Market share starts to

matter more. The stock price moves with

every headline and then the lockup

period ends and insiders start selling

their shares. When a company goes

public, employees and early investors

can't sell their shares right away.

They're locked out for a fixed period of

time, usually around 180 days. When the

lockup period expires, billions of

dollars worth of new shares can hit the

market all at once as the insiders

finally start to sell. That selling

pressure drives the stock price down and

it can drive it down a lot. That's not a

warning. It's actually an opportunity if

you know the schedule and you can plan

around it. Here's how big this

opportunity can be. Palanteer went

public on September 30th, 2020 at a

price of $10 per share. By January of

2021, it was at $35, a 250% gain in just

4 months. The lockup period expired on

February 18th and 80% of all their

outstanding shares hit the market all at

once. 1.8 billion shares and the stock

price dropped by 30% over the coming

weeks which gave investors a much better

entry point. Palanteer went on to be one

of the best performing stocks of the

last 5 years and this channel's second

biggest winner only after Nvidia. Meta

Platforms went public in May of 2012 as

Facebook, one of the most anticipated

IPOs of all time. It opened at $38 per

share. Then it stalled out and dropped

by over 50% that summer. Investors who

waited and bought it at $18 per share

made more than 30 times their money over

the next decade. The IPO was not the

best buying opportunity, but the crash

after the lockup period was. ARM went

public on September 14th of 2023. This

was actually ARM's second time going

public. SoftBank acquired them for $32

billion in 2016 and then relisted them 7

years later. So unlike most IPOs,

including Cerebras, ARM already had

decades of revenue history and a proven

track record when it went public again

at $51 per share. It popped over 25%

that day, but within a week it was back

below its IPO price and by early October

of that year, it was down by 27%. ARM is

worth over $200 per share today.

According to Market US, the global

artificial intelligence market is

expected to almost 19x in size over the

next 9 years, which is a compound annual

growth rate of 38.5%

through 2034. But many of the companies

building next generation AI applications

are not publicly traded. Think about the

90s and early 2000s. Companies like

Amazon and Google went public very early

in their growth cycle. But today,

they're waiting an average of 10 years

or longer to go public. That means

investors like us can miss out on most

of the returns from the next Amazon, the

next Google, the next Nvidia. That's

where VCX comes in, the sponsor of this

video. VCX is the public ticker for

private tech. Venture capital is usually

only for the ultra wealthy, but VCX by

Fundrise gives everyday investors access

to some of the top private preIPO

companies on Earth. They have an

impressive track record already

investing over $500 million in some of

the largest, most in demand AI,

infrastructure, and space launch

companies. So, if you want access to

some of the best late stage companies

before they IPO, check out VCX by

Fundrise with my link below today. All

right, so the pattern is pretty clear.

When insiders sell, the stock price

drops. Meta Platforms and Palunteer

Insiders sold because they were sitting

on massive gains from when those

companies were still private. Cerebras

Insiders are in that same position. The

company's valuation went from $8 billion

last September to $95 billion today, an

11x gain in just 8 months. Once the

lockup lifts around November of this

year, I expect a lot of insider selling.

But is Cerebras actually worth investing

in? To understand that, we need to

understand the science behind this

stock. For 75 years, the semiconductor

industry made the same assumption. Chips

should be small. The logic is pretty

simple. When chips are made, defects can

happen. A single speck of dust or a

microscopic flaw in the silicon crystal.

They're random and sometimes they're

unavoidable. The bigger the chip, the

higher the odds that a defect lands

inside it and kills the entire thing.

That's why chip makers keep them small.

For example, the actual compute die

inside an Nvidia Blackwell B200 is

roughly 740 mm, which is about the size

of a postage stamp. Cerebrus is betting

their entire company that this approach

is wrong. Every chip on Earth gets

stamped out of a large silicon disc

called a wafer. And that wafer gets cut

into hundreds of individual chips.

Cerebra skips that step entirely and

turns the whole wafer into a massive

chip that they've called the wafer scale

engine or WSE for short. The current

generation is the WSE3 and the die size

is over 46,000

mm or over 60 times bigger than

Nvidia's. If Nvidia's chips are the size

of postage stamps, cerebruses are the

size of dinner plates. But as you know,

size doesn't matter, it's how you use

it. Transistors are the fundamental on

andoff switches that make computation

possible. The more transistors, the more

operations a chip can do at once.

Cerebras chips have 4 trillion

transistors, 19 times more than Nvidia's

B200's. But the chip itself is 62 times

bigger, which means Nvidia actually

packs around three times more

transistors into the same area because

they're made on a more advanced process

node by TSMC. If transistors are like

switches, then AI cores are like

workers. Each one is a processing unit

that handles a piece of the math. More

cores means more work can happen in

parallel. Cerebras' wafer scale engine

has a whopping 900,000 cores, 44 times

more than Nvidia. Onchip memory capacity

is how much data can be held close to

the processors so that it's ready to use

right away. Nvidia has 192 GB of high

bandwidth memory stacked right outside

the chip. While Cerebras has 44 GB of

SRAMM, which is a faster type of memory

built into the die itself, which limits

how much they can fit without

sacrificing compute. Chips constantly

need to transfer data between these

cores and memory. So, memory bandwidth

is the speed at which that happens.

Think of it as the width of a highway. A

wider highway can move more cars even if

those cars are all going the same speed.

The wafer scale engine moves data at 21

pabytes per second, which means it can

move around 2600 times more data than

Nvidia's B200's at a time. So Nvidia's

chips can hold four times more data in

memory, but Cerebrris can move it 2600

times faster. That's a huge deal for AI

inference performance. As a result,

Cerebrris can run MetaLama 4 Maverick

model at 2500 tokens per second, which

is roughly 2.4 four times faster than

the Nvidia B200. The big difference in

inference performance is because NVIDIA

moves data between chips, across cables,

and through switches, all of which adds

extra time to every transfer, while

Cerebrris moves data across a single

chip. No hops, no cables, just compute.

While the overall speed advantage goes

to Cerebrris, the actual difference

depends on the exact workload. Cerebras

wins when it comes to real-time

inference applications like voice and

translation, coding agents, and

reasoning models that spend time and

tokens thinking before they answer. Any

workload or workflow where speed really

matters is one where Cerebras has an

edge. But what does that mean for

Nvidia? Well, they win basically

everywhere else, like batch inference

processing, which is where thousands of

requests get handled at once and total

throughput matters much more than the

speed of any one response, or like

inference for massive frontier models

that don't fit on a single wafer scale

chip. Another example would be workloads

that mix training and inference.

Companies that train and serve models

from the same AI infrastructure don't

usually run two separate chip

ecosystems. Hyperscalers are the

exception there, not the norm. But the

biggest thing is CUDA. Two decades of

software, developer tools, and

infrastructure that every major AI team

is already running on. Switching

architectures means rewriting

fundamental software. And most AI teams

won't do that unless the speed gains are

game-changing. So, let's talk about

who's actually buying these wafer scale

chips and the associated risks. Right

now, Cerebras has three sets of major

customers, and the order really matters

here. First and foremost are two

entities in Abu Dhabi that make up 86%

of Cerebris's revenue in 2025. A

university of artificial intelligence

and an AI cloud company called G42. The

university alone accounted for 62% of

Cerebras' revenue. That's not exactly a

diversified customer base. That's one

relationship in one country, accounting

for almost all of their income last

year. The second major customer is

OpenAI, which signed a compute agreement

valued at over $20 billion between now

and 2029. According to the IPO filing,

OpenAI committed to purchasing 750

megawatt of cloud compute capacity from

Cerebrris with options to expand that to

2 GW. But the deal also comes with

warrants for OpenAI to buy roughly 10%

of the company for basically nothing.

That means existing shareholders will

get diluted. But it also means that

OpenAI has a huge financial incentive to

make sure Cerebras succeeds. And the

third big customer is AWS. A couple

months ago, Amazon agreed to integrate

Cerebras into their AI development

platform, Amazon Bedrock. Every

developer building an AI application on

AWS can now run inference workloads

directly on these wafer scale chips.

That's serious distribution. Cerebras

just got access to the biggest customer

base on Earth without having to build a

sales team to reach them. Three big

customers, one big relationship driving

all revenues for 2025 and two new deals

that haven't hit their income statement

yet. Speaking of which, let's look at

their financials next. Cerebras reported

$24.6 million of revenue in 2022, $79

million in 2023, $290 million in 2024,

and $510 million in 2025. That's 20x

revenue growth in 3 years, and 76%

growth year-over-year. In quarter 4 of

last year, they made $171 million,

putting them closer to a $700 million

annual run rate when they IPOed. Gross

margins were reported to be 39% in 2025,

down slightly from 42% in 2024. Their

hardware business runs at 43% gross

margins, while their cloud business runs

at 30%. That's because running data

centers costs a lot more than just

selling chips. That margin gap matters

because more of their growth is coming

from the lower margin cloud business,

not from hardware. That means a big part

of the bull case for Cerebras is that

cloud margins will keep improving as

more customers fill the capacity that

they're building right now. And the bare

case is the flip side of that. Competing

with AWS, Microsoft, and Google on cloud

infrastructure might force their margins

to stay low forever. So margins are one

of the biggest numbers that we need to

watch over their next few earnings

calls. Cerebras also reported a gap net

income of $238 million, which makes them

sound profitable on paper, but that

includes a $363 million one-time

non-cash gain from unwinding a financial

contract tied to preferred stocks. Said

another way, this one-time gain has

nothing to do with how the technology or

the business are actually performing

today. And if we remove it, Cerebrris

actually posted an operating loss of

$146 million and an adjusted net loss of

$76 million. That means the underlying

business is still burning cash.

Operating cash flows came in at minus

$10 million back in 2025, but they had

over $700 million in cash on hand, plus

another billion loan from Open AI.

That's roughly a $1.7 billion war chest,

but they spent almost 400 million of

that on capex last year alone. So cash

is burning fast. On the flip side,

Cerebrris does have a $24.6 billion

backlog that stretches into the 2030s.

That's almost 50 times last year's

revenue. About 80% of that comes from

the OpenAI deal I just mentioned. But

there are two other disclosures in the

S1 form that investors need to know

about. First, when they were a private

company, the same person was writing and

reviewing their accounting, which is a

basic internal controls failure. They

fully disclosed this and they're fixing

it now. But we should wait and see what

their first financial audit turns up now

that they're public. And second, their

CEO, Andrew Feldman, settled securities

charges back in the dot era. I'm not

really worried about this since it was

for a completely different company 18

years ago and he went on to sell his

last company to AMD for $334 million

before running Cerebras for a decade.

Still, I figured it was worth mentioning

since it's also disclosed in the S1. All

right, let's put everything together and

see if Cerebra stock deserves a spot in

our portfolios. And if you feel I've

earned it, consider hitting the like

button and subscribing to the channel.

It really helps and it lets me know to

make more content like this. Thanks.

Now, here's how I'm investing in this

stock. Cerebras built a chip that should

be unbuildable. They partnered with TSMC

to develop a process that didn't exist.

They spent years planning and building

for a workload that didn't have a market

yet. Realtime AI inference. That means

they saw Agentic AI coming. They waited

for the rest of the world to realize and

now they're worth close to a hundred

billion dollars. Besides being run by

literal visionaries, the bullcase for

Cerebras comes down to three big

factors. First, the market they're

targeting is massive and it's growing

fast. Like I said earlier, the global

artificial intelligence market is

expected to grow at a compound annual

growth rate of 38.5%.

Three times faster than the growth of

the S&P 500. So Cerebras doesn't need a

huge market share to see huge growth

over the next few years. Second, their

open AI deal is real commercial

validation for their wafers scale

architecture, $20 billion in contracts,

a separate $1 billion loan and a warrant

for roughly 10% of the company at a

strike price of basically $0. OpenAI is

betting big on Cerebras' success. So if

you think OpenAI is smart money, then

that's a useful signal. And third, AWS

is already solving the hardest problem

for any new chip company, distribution.

Getting into Amazon Bedrock means every

developer building AI applications on

AWS can now use Cerebrris's wafer scale

engines without sales calls, without

contract negotiations, or without going

through a procurement process. They're

already on the biggest cloud platform in

the world. Amazon also has a warrant for

up to 2.7 million shares of Cerebrris at

$100 per share. So they're also betting

big on their success today. Cerebras can

run reasoning models 2.4 times faster

than Nvidia. The more AI reasoning

models get adopted, the more valuable

this weight forcale architecture will

become. That's a good position to be in

this early in the AI revolution. But my

plan right now is to wait for their next

earnings call to see their audited

financials to see how they're doing on

their existing contracts and how they're

acquiring new customers to lower their

overall concentration risk. The lockup

period for employees and early

shareholders expires around November,

180 days after the IPO. Based on

everything we saw with Palanteer, Meta

Platforms, and ARM, that could be a

great opportunity to get rich without

getting lucky. Let me know in the

comments if you're buying Cerebra stock

today, waiting for their next earnings,

or waiting for the lockup period to

expire. And if you want to see what else

I'm buying to get rich without getting

lucky, check out this video next. Either

way, thanks for watching, and until next

time, this is Tickerol U. My name is

Alex, reminding you that the best

investment you can make is in you.