[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

![FransBakker9812 Avatar](https://lunarcrush.com/gi/w:24/cr:twitter::1732263448640413696.png) Frans Bakker [@FransBakker9812](/creator/twitter/FransBakker9812) on x 7123 followers
Created: 2025-07-10 12:54:33 UTC

$IREN $CIFR $ORCL Oracle Is Hot 🔥

Ever since Oracle had their earnings call mid-June, the company has seen a lot of media coverage, and chatter on fintwit and other platforms.

Ever since @IREN_Ltd closed the deal on the Sweetwater site, we have been contemplating who will be the hyperscaler to sign with IREN.

Now Horizon is being built, we are doing the same for the Childress site. 

A few weeks ago, everyone had Oracle as the number X choice. A big reason for this was that Larry Ellison said they have a customer who will take "all the capacity they can get, no matter where it is".

Just over a week ago, SemiAnalysis wrote a piece about Oracle, and last week Oracle announced a contract with an unnamed client, to deliver $30B of annual revenue starting in 2028.

In this post, I want to dive a bit deeper into this story, and find out if Oracle could potentially be a future client of $IREN. 

Let's start with this post I found on LinkedIn which had some very interesting comments from industry expert @Tech_Journalism and Matt Caldwell from Hyper solutions. 
So the main questions are:
- Who will build the data centers for Oracle?
- Who can provide the power?

During @Oracle's last earnings call,  Larry Ellison said:
"Oracle will be the #1 cloud database company. Oracle will be the #1 cloud applications company, and Oracle will be the #1 builder and operator of cloud infrastructure data centers. 
We will build and operate more cloud infrastructure data centers than all of our cloud infrastructure competitors combined."

Now, as far as I know, Oracle doesn't build anything. They just throw money around and lease things.
In this situation, SemiAnalysis agrees with me, for a change:
"As of today, Oracle doesn’t have datacenter self-build capabilities, a disadvantage relative to other hyperscalers who can combine self-development and leasing, and often use the former for large compute clusters."

So, if Oracle says they will build more infrastructure than all of our competitors combined. How much would they actually need others to build for them?

Who are Oracle's competitors, and what is their FY25 Capex for AI hardware & Infrastructure?
- AWS $85B
- Azure $70B
- Google $65B
- Meta $65B
So this year alone, Oracle's competitors combined, will spend $285B on AI.

Now these Hyperscalers are builders, and Oracle isn't forecasting to spend this amount on building.

If we look at what I call the "Altman Benchmark", we can roughly estimate what AI Capex can bring you:
So $XXX Billion investment, gives you 5GW of data centers, and X million GPUs.
Based on $3M per NVL72 GB200 system, we can conclude that the X million chips in this "benchmark", toal to roughly $83.33B in GPUs.
That leaves $16.67B worth of data centers and other ancillary equipment.  

Oracle has a FY25 Capex forecast of "just $21.2B". which could imply—if that's just for GPUs—they are planning to invest in roughly 1.27GW this year.

For FY26, Oracle has guided for $25B in Capex, and SemiAnalysis has estimated it will be over $30B.

No matter the number, it won’t come close to beating the competition - let alone combined. 

So maybe Larry was being hyperbolic, because literally, it can't be true, at least not for the coming 1-2 years.

But what about beyond 2027?
Just a week ago, Oracle announced a $30B deal with a "mystery customer"
"Oracle ( $ORCL ) signed a cloud services deal with an unnamed client expected to generate over $XX billion in annual revenue starting in fiscal year 2028."

Now, on X and LinkedIn, people are wondering if it's @OpenAI again. Since Stargate is not even a real JV yet, and the deal in Abilene consists of Oracle directly with OpenAI, could this new client be just OpenAi again?

Does Oracle want to avoid the "customer concentration" worries, and keep the customer's name in the dark for now? I guess we will find out sooner or later.

The real question is: who will build the datacenters that is going to bring $30B of annual revenue by FY2028?

OpenAI is now paying $10B per year for "Stargate 1" in Abilene:
"OpenAI committed to buying all the compute capacity from that campus, these five-year contracts sum up to close to $10B annual expense for OpenAI"

According to SemiAnalysis, Stargate X is ~880MW of Critical IT load, out of 1200MW of gross Megawatts. 

That would mean, the $30B client would need roughly X times that much, which translates to 3.6GW of power, fully delivered and constructed by 2028.

Now let's go back to the earnings call, and look at what CEO Safra Catz said about their Capex:
“We don’t order, we don’t build unless we've got orders for our capacity to be built out.”

So Oracle has the orders - and at least one that will need 3.6GW XXX years from now - but who has the power for this capacity?

Catz:
“We are still in a position where our supply is not meeting our demand.
We actually currently are still waving off customers or scheduling them out into the future so that we have enough supply to meet demand.”

Ellison:
“The demand is astronomical… People are calling up and asking us, 'Please, can you find us more capacity? We'll take it wherever. If it’s in Malaysia, we’ll take it.”

So to sum up:
X. Oracle has a $30B "order" for 2028 (Since July 1st).
X. Oracle only "builds" when they have "orders for their capacity".
X. $30B could indicate 3.6GW of power needed.

From SemiAnalysis:
This led to the emergence of a new market segment: the AI-first developers. New entrants, often from the cryptomining world, understood that power was becoming the #1 concern and leveraged their existing interconnection agreements to plan large datacenters, away from population centers but with ample power.
 
X. Oracle already works with AI-first developers (Crusoe), away from population (Abilene) with ample power (West-Texas).
X. OpenAi has no problem with the compute coming from West-Texas, they took all the compute in Abilene for at least X years.

Here is Michael Pariser from Oracle's OCI team, who recently joined as the Senior Principal Renewable Energy & Utilities, and is based in Houston, Texas:
X. Oracle is looking for ways to unlock MW capacity as quickly as possible, for active and planned infrastructure projects globally.
X. Oracle is still looking for XXX% renewable energy to cut their emissions in half by 2030.
X. Oracle is looking to connect with innovative energy developers to supply power for the massive cloud and AI infrastructure buildout underway at OCI.
X. Oracle is looking to hire people with deep US power expertise.

So, let's see:
- Who is this MegaCampus builder? (it's not a REIT, they don't have the power, it's not Crusoe/Lancium, they don't have enough power).
- Who is this Innovative Energy Developer with XXX% renewable energy?
- Who is this AI-First Developer with 2.75GW power available in US markets by 2028?
- Who can help Oracle unlock MWs to reach their required capacity by 2028 to satisfy their signed client?

Does this not look like the perfect fit for $IREN's Sweetwater 1+2, supplemented by Childress and maybe @CipherInc's capacity?

Is there any way that Oracle can meet their signed capacity, without $IREN and $CIFR? 

Did Oracle just become my #1 candidate on the list of potential $IREN clients? 👀

![](https://pbs.twimg.com/media/GvfYNN0akAIOqqb.png)

XXXXXX engagements

![Engagements Line Chart](https://lunarcrush.com/gi/w:600/p:tweet::1943292759512879551/c:line.svg)

**Related Topics**
[iren](/topic/iren)
[quarterly earnings](/topic/quarterly-earnings)
[oracle](/topic/oracle)
[$cifr](/topic/$cifr)
[$iren](/topic/$iren)
[cipher mining](/topic/cipher-mining)
[stocks bitcoin treasuries](/topic/stocks-bitcoin-treasuries)
[$orcl](/topic/$orcl)

[Post Link](https://x.com/FransBakker9812/status/1943292759512879551)