So, this just happened, but we don't

really want to talk about it. Actually,

let's talk about it. The history between

Sam Alman and Daario goes all the way

back to 2016 when Daario left Google as

a researcher to join a new startup

called OpenAI. And before OpenAI

released, Daario left to start his own

company called Enthropic. And the

competition between them started to

become a bit silly where after open

started running ads on their platform,

Enthropic followed up with a whole ad

saying how Enthropic will never run ads

on their platform. And Sam Alman wrote a

rather long post on X defending his

position, making it a bigger deal than

it should have. And since this picture

is making all of us more uncomfortable,

let's move on to something more

substantial, which is really breaking

down the competition between OpenAI and

Anthropic. Welcome to Kale Brightes Code

where every second counts. Whether

you're on team openai or team anthropic,

chances are you probably have some

financial exposure to both of these

companies. If you look at the overlap of

investors in OpenAI and Anthropic, Max 7

companies like Google, Amazon,

Microsoft, and Nvidia all have stake in

either or both of these companies. And

even outside of the MAX 7 companies,

these publicly traded banks have also

serviced debts, which means having a

managed retirement account or even an

S&P 500 will likely have some exposure

to what's happening between Enthropic

and OpenAI. But when we look at the

competition between OpenAI and

Anthropic, one of the major differences

is when it comes down to data centers.

While Anthropic relies mostly on

partnerships with AWS and Google Cloud

for training and inference, OpenAI has

famously been trying to be vertically

integrated, including working with

Broadcom to make their own custom AI

chip. So, right off the bat, we're

looking at a significant difference in

the business model between OpenAI and

Anthropic. According to Kushman and

Awakefield's 2025 data center report,

the construction cost to data centers

has been rising since 2020. material

cost to the power supply equipment and

cooling has been rising more than 40%

across the board since 2021. And if you

factor in the lead time that goes into

actually acquiring basic materials that

goes into construction, the two main

takeaway here is first how OpenAI's

partners like Crusoe and Oracle that are

actually constructing these data centers

can't fumble on this since major delays

in construction directly affects

OpenAI's business plan. And second, just

how impressive XAI's construction of

their Colossus facility looking back

when they only took 122 days to turn an

old factory into a 100,000 H100 filled

AI data centers. But delays in data

center construction isn't the only thing

that OpenAI needs to execute on. It's

also financing. Ever since the official

announcement of Stargate back in January

2025, people like Elon Musk have been

questioning how OpenAI will actually

secure $500 billion to finish their 10

gawatt data center plan by 2029. So far,

OpenAI raised about $164 billion to get

to where they are today. And they're

still about 10% of the way as they

continue to expand their Texas facility.

Oracle is finishing up their Wisconsin

data center. Michigan is about to start

construction along with Ohio. We're just

getting started in the Stargate plan.

And all of this requires continual

funding, which means even though OpenAI

raised about $164 billion up to today,

they likely need to secure another $400

billion until 2029 to see this entire

thing through. And getting this kind of

funding requires continual interest in

venture capitals, banks, private

equities, and other companies. Which

means we need a CEO that can not only

create positive sentiment around AI, but

also strategic alignment with cash cow

companies like Amazon and Nvidia that

can help OpenAI invest in their dream.

But recently, OpenAI receiving

investment interest from Amazon led to

Nvidia actually walking back on their

$100 billion commitment because Nvidia

is likely only interested in investing

in OpenAI as long as OpenAI commits to

using Nvidia chips to fill up most of

their data center needs. But OpenAI's

partnership with Amazon likely leads to

using Amazon's training chips, which by

in effect reduces Nvidia's commitment

towards OpenAI. As you can see, when we

look at the whole strategy around data

center alone, unlike Enthropic that

partners with companies to largely meet

their compute needs, OpenAI has their

hands full. Well, I guess not that full.

Now, things look very different when we

start to look at Enthropic. When we look

at Anthropic's revenue source, they

positioned themselves as a market leader

in enterprise AI and coding. Ever since

their first model release of Claude in

March 2023, which nobody really cared

about since Chachi BT was a buzz,

Enthropic had a pretty impressive runup

in their revenue, largely contributed by

their API availability for coding into

more enterprise level adoption in 2025.

And with the release of cloud code and

wider AI adoption, Enthropic grew their

run rate revenue to $14 billion. On the

other hand, OpenAI's 2025 estimate for

their expected ARR is $20 billion, which

is different from the actual revenue for

2025, which is likely closer to $14

billion instead. But regardless, what

OpenAI is saying is that since there

seems to be a linear relationship

between compute measured in gigawatts

and annualized revenue, it's hard not to

think looking at this that their

implication here is that by the time

they finished their 10 gawatt facility,

we could expect them to have a hundred

billion in revenue following their

linear projection, which is hard to

believe that this linear relationship

will continue on without mixing in other

sources of revenue like running ads.

OpenAI's opt-in to run ads on free and

go subscription users was rather

controversial. But what running ads

really show about OpenAI more than

anthropic is not only when it comes to

their revenue growth, but more about

OpenAI quickly needing to stop their

bleeding by essentially subsidizing

their free users from eating away into

their margin. Because the more they

subsidize, the more they need to look

for financing in cost that aren't going

into their data center buildouts. This

kind of precarious situation is what

Daario commented on a recent podcast

where he said that even a slight

miscalculation of capex in data centers

could essentially sink the company which

is why Anthropic is taking more of a

measured stance in how they're unfolding

their business plan. So far, Anthropic's

most recent plan of $50 billion data

center in Texas and New York seems

rather small compared to OpenAI that

wants to spend up to a trillion dollars

in the next 10 years. So the race

between Enthropic and OpenAI is not only

about how they source their revenue from

users, but also how they source their

compute and financing. And in order for

both of these to work, they not only

need to continue proving their relevance

when it comes to their models

performance and benchmarks, they also

need to win the hearts of many end

users, especially the early adopters and

enterprises that tend to spend the most

on their platforms through subscriptions

and APIs. And recently a lot of early

adopters have grown more disgruntled

towards anthropic crackdown on oath of

their max subscription plans especially

as we look at proliferation of openclaw

use cases given that openai quote

unquote acquired the creator of

openclaw. This kind of sentiment has

been stronger as the cost of

intelligence has been dropping because

cheaper alternatives from Chinese models

are increasingly becoming more and more

viable. And whereas before people easily

signed up for multiple subscriptions at

the same time, more options mean people

are seeing the value in being able to

hot swap or load different models from

different providers. But cracking down

on high-paying users to only allow them

to use their model through their

subscription is starting to not play

well among users, especially given that

the alternative option is API pricing,

which is extremely unfavorable for

Enthropic given that their pricing is

much much higher than their competitors.

So given what we covered, do you think

Anthropic's measured approach will win

or do you think OpenAI's economies of

scales approach is actually a better

method? What do you think?