Uber to shut down its own data centers as it strikes cloud agreement with Oracle and Google

[PriyaWalia]

Recap: Ride-hailing giant Uber plans to move away from its own data centers and shift to cloud services offered by Google and Oracle. The move is unusual because Uber has historically relied on its own server hardware, however the company is opting for a new strategy after suffering post-pandemic supply chain disruptions and longer IT hardware delivery timelines.

Uber has struck agreements with Oracle and Google to migrate 95% of its data center contents onto the cloud, with plans to leverage additional services such as Google Ads, the Google Maps platform, and database service Cloud Spanner. Both agreements would last seven years. Uber's transition to the Google Cloud Platform (GCP) and Oracle Cloud Infrastructure (OCI) will enable the company to improve its efficiency and scalability while reducing costs significantly. The transfer should be concluded "within a few years," Uber CEO Dara Khosrowshahi said.

Interestingly, Google and Oracle's cloud offerings are not the leaders in the sector falling behind Microsoft's Azure and Amazon's AWS in market share. Under the agreement, Uber will transfer its data and apps to Google Cloud to modernize its infrastructure and utilize Google Cloud services, which include artificial intelligence, machine learning, security, and microservices.

The partnership between Oracle and Uber is reciprocal. Oracle is signing on as a global Uber for Business client, selecting Uber as its preferred ridesharing service for its employees' international travel and dining. The two companies have also pledged to work together to develop new retail and delivery solutions. "To deliver on [our] promise for customers while building value for shareholders, we needed a cloud provider that will help us maximize innovation while reducing our overall infrastructure costs," Uber CEO Dara Khosrowshahi said in a statement.

Although Uber likely operates at a different scale, just recently Ruby on Rails creator and Basecamp maker, 37signals was announcing they were leaving the cloud due to the sheer annual cost of running on rented hardware. According to Barron's, analyst Derrick Wood estimated that the Oracle deal is likely worth more than $1 billion.

Kamran Zargahi, Uber's senior director of technology strategy, told The Wall Street Journal that moving to a cloud provider will lessen the firm's reliance on hardware supply chains and allow it to take advantage of certain cybersecurity defenses and compliance standards.

Permalink to story.

https://www.techspot.com/news/97603-uber-shut-down-own-data-centers-strikes-cloud.html

 

[human7]

Execs love touting going to the cloud, and the plan to "migrate 95%" of their "data center contents" onto the cloud makes me wonder how they came up with that number.

My company made the push to cloud. Took a couple of years to get so much out there. Now? We're going back on-prem because of the opex cost. Lot's of other companies are doing the same.

It's no surprise that GCP and Oracle are offering them tons of money (they tried to poach us from AWS and succeeded in a few areas), and I suspect that Uber may regret their choice.

Capex is expensive. But once it's paid for, it's paid for. There's always some opex, but nothing makes the opex go through the roof like, well, the cloud. I wish Uber the best of luck. But I suspect that they, like many enterprises before (and inevitably after), will be rethinking this move in a few years.

 

[takaozo]

My brilliant CEO is pushing the Cloud migration also. I made some forecasts for On-Prem, Co-located and Cloud.
For a single machine with 64 cores, 128Gb Ram and 6TB storage the on-prem cost for 5 years is 25k.
Same machine Co-located in a big datacenter is about the same cost and only depends on how many cabinets are rented.
Now the cloud.........15K per month.

Of course the cloud wins because the man want's this because "it's the future and everybody is doing it"

 

[toooooot]

Execs love touting going to the cloud, and the plan to "migrate 95%" of their "data center contents" onto the cloud makes me wonder how they came up with that number.

My company made the push to cloud. Took a couple of years to get so much out there. Now? We're going back on-prem because of the opex cost. Lot's of other companies are doing the same.

It's no surprise that GCP and Oracle are offering them tons of money (they tried to poach us from AWS and succeeded in a few areas), and I suspect that Uber may regret their choice.

Capex is expensive. But once it's paid for, it's paid for. There's always some opex, but nothing makes the opex go through the roof like, well, the cloud. I wish Uber the best of luck. But I suspect that they, like many enterprises before (and inevitably after), will be rethinking this move in a few years.

It makes sense. These companies gave great deals to attract people, now they can make real money and do some real earning. That was the idea from the beginning.

 

[human7]

My brilliant CEO is pushing the Cloud migration also. I made some forecasts for On-Prem, Co-located and Cloud.
For a single machine with 64 cores, 128Gb Ram and 6TB storage the on-prem cost for 5 years is 25k.
Same machine Co-located in a big datacenter is about the same cost and only depends on how many cabinets are rented.
Now the cloud.........15K per month.

Of course the cloud wins because the man want's this because "it's the future and everybody is doing it"

That's an expensive cloud offering. In AWS it would only be $3,953.52 per month (current on-demand pricing for a i4i.16xlarge with 64 cores, 512 GiB Ram, 4 x 3750 GiB Nvme; and you could get cheaper by using a c6a.16xlarge and EBS for storage and even less if you use a 3-year capacity reservation).

But whatever, the final numbers aren't really important. If you need fully-provisioned performance, in the long term the cost of the machine is always lower on-prem. The cloud wins in price in general, though, since you:
1. Don't have to pay for hardware failures or electricity
2. Don't have to pay someone's salary to maintain it
3. Only pay for what you use
4. Flexible scaling

You technically do pay for 1 and 2, but only indirectly, and with economies of scale, the business using the cloud can have a much lower total cost of ownership than the business doing it on-prem. For 3, the cloud is a lot cheaper if you have a bursty workload: you don't have to pay for a 64 core machine and have it idle 90% of the time in your environment. You pay for the 64 cores only when you need them. For 4, if 64 cores turns out to not be enough, you can easily scale up. If it's too much, you can scale down. So it's a lot easier to right size your purchase.

The real power of the cloud is not in the infrastructure offerings, though, it's in the platform and software offerings that can be connected together to build powerful solutions.

All that said, yes, the cloud is still expensive, and there are still very valid reasons to do things on-prem from a cost perspective (my company is doing that, for example). We're still staying in the cloud for some things and reducing our on-prem footprint as it relates to older data centers, but there's definitly more acceptance these days for expanding our on-prem infra instead of going to the cloud for some things. Versus 2 years ago when it looked like closing down most of our on-prem infrastructure was the eventual goal.