in 2008
He just underestimated how much a combined marketing spend from three of the largest companies can convince people that an API is enough to justify extremely slow hardware at a 5-10x markup.
(And AWS was pretty overhyped in 2008.)
EDIT: I missed the news from yesterday that OpenAI signed a $300 billion (!) contract with Oracle Cloud:
https://techcrunch.com/2025/09/10/openai-and-oracle-reported...
So this big cloud revenue jump may be from a single giant customer.
if you’re using Azure, your CFO probably made the decision.
if you’re using AWS, your CEO likely made the decision.
if you’re using Google cloud, likely your CTO made the decision.
If your company relies on government contracts; they dictate the cloud provider; and Oracle knows this.
I’m not at that level in my career, but I have had very negative issues with Google, their product graveyard, and their lacking of human support for serious issues for paid offerings.
I would not put my business in their hands, ever.
Turns out this problem disappears if you spend enough money.
GCP does a _lot_ of things as well or better than AWS. I particularly mostly like their IAM model (the entire stack including their nice managed identity implementation and use of public key cryptography instead of API keys) and how "projects" (cloud accounts) can inherit properties in a hierarchy.
Their solutions are more technically sound, their API and UX significantly more consistent; they make the billing system available through the same interfaces that you use for other data processing on the platform.
I wrote a more detailed diatribe on the failings of other providers[0]; but largely the “correctness”, “consistency” are the main reasons. For example their IAM actually works for everything, including GKE pods; and they don’t oversubscribe and try to align VMs to NUMA zones (AWS didn’t do that when I last checked, Azure still doesn’t even try).
Also, as a previous customer; I got much better support from Google than I ever got from AWS; but this is a very variable thing depending on your size, where you are and who your TAM is. In the games industry in Sweden its seriously 10/10- I even talked to the hypervisor, storage and SDN/Network teams directly.
They’re not choosing Hetzner because they think its the unsafe option.
Wild - I’d love to be the sales guy who got that commission
E.g. a few years back I worked with a guy (at an IT sector company) who had a quota of 3 million in the region, but won a single deal worth 30 million + had a great quarter in general anyways. He ended up not getting too much more than if he just had the great quarter.
Sorry to share - there's no sales guy here. It all happened at the top, hint - Stargate.
Oracle Cloud's free tier includes a 4 CPU / 24 GB RAM Ampere A1 ARM VM. On AWS / GCP, an equivalent VPS would be $100+/month, but on Oracle cloud it's free.
And technically can go poof at any time, I guess?
edit: and 200GB of free block volume storage WITH backups? That seems like a good replacement for standard cloud storage.
The company said Tuesday after the bell that it has $455 billion in remaining performance obligations, up 359% from a year earlier.
Oracle now sees $18 billion in cloud infrastructure revenue in fiscal 2026, with the company calling for the annual sum to reach $32 billion, $73 billion, $114 billion and $144 billion over the subsequent four years.
$114b? $144b for Oracle? per year?
Who? Can someone make a serious attempt at speculating who and how revenue will get to 144b/year?
I know Oracle as a Stargate partner. Maybe this is OpenAI being extremely bullish on its growth figures?
Edit: Ok. I have a theory. Oracle is the only big cloud company NOT developing their own AI silicon chips. AWS, Google, Meta, Apple, Microsoft are all making internal AI chips that compete against Nvidia. Nvidia prioritizes Oracle as a partner so the first and the bulk of Nvidia's chips go to Oracle. Therefore, AI labs are going to Oracle for Nvidia access. Video on Ellison having dinner with Jensen Huang "begging" him for GPUs [0].
Partnering with the non-Amazon cloud providers meant spending several millions in surplus infrastructure costs to hire more engineers to un-fuck the alternate cloud shitshows, but unlocked billions in partner deals (think Microsoft promoting our products running on Azure to Azure customers). And then every 3-5 years when we negotiated new cloud resource SKU discount rates we had leverage on common resources like VMs, storage buckets and SQL databases. "If you can't beat this rate we'll move part of our capacity to the other clouds" kinda thing.
- The $300B from OpenAI doesn't even start until 2027 and runs through 2031
- We don't know anything about OpenAI's options to cancel, delay or modify the deal
- There's huge execution risk for Oracle, they have to scale up their operation by more 20x or 30x, obtain gear and staff and gigawatts of power
- Even if everything goes perfectly, and both Oracle and OpenAI perform, that $300B is revenue, not profit -- 70% or 80% of that goes suppliers for capex and operating costs
- That's an uncertain promise of a future stream of profits, and deserves a sizable discount to get back to present value
- Ellison owns about 40% of Oracle, so how does the share attributed to him personally get to $100B in today's dollars?
"Oracle is clearly leveraging a number of advantages in its cloud software/hardware businesses to attract the largest of the AI enterprises, including OpenAI, xAI, Meta, NVIDIA and AMD,"
What exactly are these advantages?
Related:
OpenAI, Oracle Sign $300B Computing Deal, Among Biggest in History
> Why are we doing this? What if we just made better ways to consume content?
Yes! Yes!
> It has window snapping, keyboard shortcuts, and a bookmark app. It works as well as you’d expect an operating system to work in a browser.
Oh no...
bdcravens•4mo ago
> Oracle’s founder, Larry Ellison, added $100 billion to his net worth on Wednesday. Bloomberg reported that he had topped Tesla CEO Elon Musk as the world’s richest person.