Oracle, the Comeback Kid: Nothing says “we’re back” like a $30B cloud contract
How GenAI reset the gameboard and gave Oracle its second chance.
For most of the 2010s, Oracle was seen as a relic of a different era.
It had the revenue. It had the customer base. But it had missed the moment.
Oracle sat out the first wave of public cloud. It clung to its on-prem cash cows. It felt like Microsoft’s older, stiffer cousin- technically sound, but culturally out of step.
And that comparison is worth pausing on: Microsoft is actually older than Oracle (1975 vs. 1977), yet somehow has reinvented itself every decade like a particularly ambitious liberal arts major. Windows, Office, Azure, GitHub, OpenAI, Copilot… you get the idea.
Oracle, by contrast, stayed Oracle. The guy in the corner of the party who’s still explaining why PeopleSoft was actually a smart acquisition.
As cloud-native tools became the new stack and developers became kingmakers, Oracle stuck with a license-driven, CIO-first model long after the market moved on. It’s only now - nearly a decade behind - that Oracle is stepping into the AI spotlight and actively shedding its “legacy tech” label. But like most overnight successes, this one was years in the making.
🕰 A brief history of missing the point
Oracle’s cloud story started late and shaky. While AWS launched in 2006, Azure in 2010, and Google Cloud in 2011…Oracle OCI didn’t land meaningfully until 2016–2018. And even then, its early cloud products were barely competitive.
It wasn’t until OCI Gen2 (around 2018–2019) that Oracle had a modern offering and by then, a full decade had passed. The major players had already defined the category, developer preferences had ossified, and the world had mostly stopped checking to see what Oracle was up to.
Oracle’s early missteps stemmed from its DNA: on-prem enterprise software. Rather than rethink its products for a cloud-native world, it spent years trying to “cloud-wash” its legacy stack and defend its margins. Its early cloud offerings were often difficult to deploy, not developer-friendly, and optimized for existing Oracle customers, not net-new builders. Perception lagged because usage lagged. Developers weren’t choosing Oracle by default, startups definitely weren’t.
🤖 GenAI reset the gameboard
Then something interesting happened. The rise of generative AI created an insatiable demand for compute - fast, cheap, and scalable. This disrupted the cloud status quo and gave Oracle a rare second chance. Unlike traditional workloads where AWS and Azure dominate, GenAI:
Requires specialized infrastructure (GPUs, low-latency networks)
Doesn’t care much about legacy developer ecosystems
Wants fast scaling and cost-efficiency
Oracle built for this.
“The largest and most important workload for the next decade will be training AI models. We’re building out our infrastructure to handle the largest AI workloads in the world.”
— Larry Ellison, Oracle earnings call, Dec 2023
Instead of trying to match AWS feature-for-feature, Oracle found edge cases that would become mainstream: high-performance AI workloads, sovereign deployments, multi-cloud flexibility. Being late doesn’t matter - if you bet where the market is going, not where it was.
Oracle’s pivot wasn’t cosmetic. It made bold, clear-headed decisions and executed on them with conviction:
Scrapped Gen1 and Rebuilt the Stack for AI. With OCI Gen2, Oracle rebuilt its cloud from the ground up:
Bare-metal compute for direct GPU access
Remote Direct Memory Access (RDMA) networking for ultra-low latency
Non-blocking, flat network fabric to support distributed training
GPU superclusters designed for frontier-scale models
Gen2 was built for GPT-scale AI. Thanks to these architectural decisions, Oracle claims 2x training speed at half the cost of competitors.
Doubled Down on Enterprise Strengths. Oracle leaned into its core strengths - security, compliance, data control - and embedded GenAI directly into its enterprise apps, not as a chatbot, but as agentic workflows inside ERP, HCM, and CX systems. They created a compelling platform for private, secure, enterprise-grade AI. Exactly what regulated industries and sovereign entities were asking for.
Bet Big, Bet Early. Oracle didn’t wait for validation. Once the strategic direction was clear, it moved fast - and big:
$7B in equity to co-found Stargate, the AI infra project with OpenAI, SoftBank, and MGX
$40B in Nvidia GPU supply, leapfrogging capacity-constrained rivals
Broke ground on 1GW+ data centers in Texas, built specifically for frontier model training
Expanded internationally with major projects like the ByteDance-backed hub in Johor, Malaysia - quietly turning it into one of the largest AI infrastructure zones outside the U.S.
This was Oracle acting before market validation - and it’s why it was ready when OpenAI started looking for scale beyond Microsoft.
💥 The $30 Billion Deal That Proved the Point
Yesterday, Oracle disclosed a $30B cloud contract - the largest in its history. While the customer wasn’t named, all signs point to OpenAI:
Oracle and OpenAI are core partners in Stargate
Larry Ellison has publicly teased an order for “all available cloud capacity”
And frankly, there are very few customers on Earth that need compute at that scale
This mega-deal serves as both a payoff and a signal:
▪️ Validation: The deal - nearly 3x Oracle’s entire 2025 data center revenue-proves that its AI infra strategy is working.
▪️ Disruption: It positions Oracle as a credible, scaled alternative to AWS, Azure, and GCP, especially as model labs move toward multi-cloud strategies.
▪️ Timing: Oracle showed up with availability, neutrality, and performance exactly when the hyperscalers were tapped out.
While others fought for exclusive AI partnerships, Oracle was the scalable, neutral Plan B. Which, it turns out, is a fantastic place to be when Plan A runs out of GPUs.
From database dinosaur to AI infra juggernaut, Oracle’s story is a reminder that being early is good, but being right on time, with the right product, is better. In a world obsessed with disruption, Oracle’s comeback shows the power of disciplined reinvention.
PS: FMAI