Oracle: The Enterprise Database Giant Betting Big on AI Infrastructure

The Executive Intelligence Series · Company Deep Dive | September 20, 2026

Oracle spent four decades as the most profitable enterprise software company. The cloud era challenged that position. The AI era has given Oracle something it rarely had in the cloud wars: a genuine shot at becoming indispensable infrastructure. This deep dive examines whether Oracle can turn its database dominance into a durable position in the AI infrastructure stack.

The Thesis

Oracle Cloud Infrastructure revenue grew 121% year over year to $7.4 billion in fiscal Q1 2027. Remaining performance obligations reached $664 billion. GPU utilization ran at 97.9%. Oracle signed more than $30 billion in new AI cloud contracts in a single quarter.

But the transformation comes at a cost unlike anything Oracle has undertaken before. Capital expenditures reached $55.7 billion in fiscal 2026. Free cash flow turned negative, at $23.7 billion. Long-term debt climbed to $122.3 billion. The company is on the cusp of a junk-grade credit rating as investors question whether the AI infrastructure bet will pay off.

From Database Giant to AI Infrastructure Contender

Oracle's transformation began long before the AI boom. For decades, the company's business model rested on selling database licenses to enterprises, then charging for maintenance and support. The model was enormously profitable. It was also increasingly vulnerable as cloud computing shifted enterprise workloads to AWS, Microsoft Azure, and Google Cloud.

Oracle's response was Oracle Cloud Infrastructure, launched in 2016 as a second-generation cloud platform built from the ground up. OCI was designed with a flat, non-blocking network architecture and RDMA (Remote Direct Memory Access) networking that allowed it to connect thousands of GPUs more efficiently than competitors. That design decision — made years before the AI boom — turned out to be Oracle's most important strategic asset.

By fiscal 2026, the transformation had reached an inflection point. Cloud revenues contributed roughly 50% of total revenues, and OCI was growing at rates that outpaced every major hyperscaler. In Citi's Q2 2026 analysis, Oracle OCI surged 92% year over year, compared to Google Cloud at 82%, Microsoft Azure at 43%, and AWS at 37%.

The CODEW Lens: Oracle did not pivot to AI infrastructure because of the AI boom. It built the infrastructure that made the AI boom possible for its customers — and the market arrived faster than anyone expected.

The AI Infrastructure Opportunity

Oracle's AI infrastructure business is driven by a simple observation: demand for GPU compute is growing faster than supply. As Oracle's own earnings release stated, "customer demand for AI Cloud Training and Inferencing Services continues to grow faster than supply."

Metric	FY2026	FY2027 Q1	Growth
OCI Revenue	$18B (est.)	$7.4B	+121% YoY
RPO	$638B	$664B	+363% YoY
GPU Utilization	—	97.9%	—
Data Center Capacity	1.2 GW	850 MW	—
New AI Contracts	—	$30B+	—

Oracle delivered 850 megawatts of AI capacity in fiscal Q1 2027 alone, containing more than 300,000 GPUs. That single quarter's delivery was 73% of everything Oracle delivered in all of fiscal 2026. The company's largest planned builds target up to 800,000 GPUs in a single cluster.

The revenue trajectory is equally aggressive. Oracle expects OCI revenue to grow from $18 billion to $144 billion over the next four years — an eightfold increase that would make OCI the largest single contributor to Oracle's revenue.

The CODEW Lens: Oracle is not chasing AI infrastructure because it wants to be a cloud company. It is chasing it because its database customers need somewhere to run AI. The infrastructure is the means. The database is the business.

The Stargate Partnership and OpenAI Concentration

No single deal defines Oracle's AI strategy more than its $300 billion, five-year contract with OpenAI, signed in September 2025 as part of the Stargate initiative. The agreement made Oracle the backbone of OpenAI's data center buildout, with Oracle managing OpenAI's flagship facility in Abilene, Texas.

The Stargate project — a joint initiative between Oracle, OpenAI, SoftBank, and MGX — aims to deploy up to $500 billion and 10 gigawatts of capacity across multiple U.S. data centers. Oracle's portion of that commitment is approximately 4.5 gigawatts of compute capacity over five years.

The concentration risk is significant. Approximately $300 billion of Oracle's $664 billion RPO comes from OpenAI contracts alone. OpenAI has not yet turned a profit, and its ability to pay depends on its ability to continue raising capital. Oracle's own SEC filing acknowledged: "Our business is, and may continue to be, exposed to risks of customer non-payment and non-performance."

But the partnership also has structural protections. Oracle structured much of the demand to reduce capital strain, with customers either funding equipment upfront or supplying their own hardware. Pre-paid and customer-supplied hardware contracts now total $75 billion. The company has also secured over 10 gigawatts of power and data capacity for AI infrastructure over the next three years, with more than 90% partner-funded.

The CODEW Lens: Oracle is not building AI data centers for OpenAI. It is building them with OpenAI — and the distinction matters. When customers fund the equipment, Oracle's capital risk is lower, and its margin profile is different. The question is whether that structure holds as the relationship scales.

Database as Strategic Distribution Advantage

Oracle's most underappreciated asset in the AI era is not its data centers. It is its database franchise.

Oracle's database business has an installed base that spans virtually every large enterprise on the planet. That installed base gives Oracle something no pure-play cloud provider has: a pre-existing relationship with the customers who need AI infrastructure. When those customers want to run AI workloads against their existing Oracle databases, OCI is the natural destination.

This dynamic is visible in Oracle's multicloud strategy. Oracle AI Database workloads can now be deployed across AWS, Microsoft Azure, Google Cloud, and Oracle Cloud Infrastructure (OCI) through Multicloud Universal Credits — a single commitment that works across all four clouds with unified pricing. In fiscal Q2 2026, multicloud database consumption surged 817% year over year.

Oracle is positioning itself as a cloud-neutral AI and data layer rather than a traditional hyperscaler. The strategy is to embed Oracle's database inside every major cloud provider, making Oracle the connective tissue between enterprise data and AI workloads regardless of which cloud the customer chooses.

The CODEW Lens: Oracle's cloud strategy is not to beat AWS at cloud. It is to make Oracle databases so deeply embedded in enterprise operations that every cloud needs Oracle — and Oracle's AI infrastructure becomes the path of least resistance for running AI on that data.

AI Infrastructure Economics and the Margin Question

Oracle's AI infrastructure business is growing faster than its margins. That is the central tension in the investment thesis.

Internal financial documents suggest Oracle's GPU deals with OpenAI and other customers have seen an average gross profit margin of just 16%, with margins varying between 10% and 20%. Morgan Stanley has modeled GPUaaS gross margins in the 12–19% range for fiscal 2028 — well below Oracle's long-term target of 30–40%.

The challenge is structural. Renting NVIDIA GPUs is a commoditized business. Oracle buys the same chips that AWS, Microsoft, and Google buy, and competes primarily on price and availability. The margins are thin because the underlying hardware is expensive and depreciates quickly.

Oracle's bull case is that margins will improve as the business scales, as Oracle optimizes its data center designs, and as higher-margin services — database, analytics, AI inference — layer on top of the GPU compute. The company has said it can generate $10 billion in annual revenue at $6.4 billion in cost, implying a 35% gross margin on those contracts.

But the near-term reality is that Oracle is trading margin for growth. The company is building infrastructure at a pace that requires it to accept lower returns to capture market share.

Metric	FY2026	FY2027 Q1
Capital Expenditure	$55.7B	$28.5B (quarterly)
Free Cash Flow	−$23.7B	−$5.4B
Long-Term Debt	$122.3B	—
GPUaaS Gross Margin	~16%	~14–20%
Target Gross Margin	—	30–40%

The CODEW Lens: Oracle is not building a high-margin business yet. It is building a high-growth business with the expectation that margins follow. The question is whether the margin expansion arrives before the debt becomes a problem.

Competitive Positioning: OCI vs. the Hyperscalers

Oracle occupies an unusual position in the cloud market. It is the clear fourth-place provider behind AWS, Microsoft Azure, and Google Cloud, yet it is growing faster than all three.

Provider	Q2 2026 Growth	Market Share	AI Strategy
AWS	37%	41%	Broadest service ecosystem; Trainium/Inferentia chips
Microsoft Azure	43%	24%	OpenAI partnership; enterprise hybrid integration
Google Cloud	82%	24%	TPU infrastructure; Gemini integration
Oracle OCI	92%	~6%	GPUaaS; database embedded across clouds

Oracle's differentiation is not breadth of services. It is performance-per-dollar for GPU workloads and database integration. Its RDMA networking architecture allows it to link thousands of NVIDIA Blackwell GPUs more efficiently than some larger rivals. Its flat network design reduces latency and improves training throughput.

The company is also the first hyperscaler to offer a publicly available AI supercluster powered by 50,000 AMD Instinct MI450 GPUs, starting in Q3 2026. That partnership gives Oracle access to GPU supply beyond NVIDIA, reducing dependency on a single vendor.

But Oracle remains a distant fourth in the cloud market. Its $7.4 billion quarterly OCI revenue is a fraction of AWS's $30+ billion. Its data center footprint is smaller. Its service ecosystem is narrower. And it is competing against companies with far more capital and far more established enterprise cloud relationships.

The CODEW Lens: Oracle does not need to beat AWS to win. It needs to be the best place to run AI workloads that are connected to Oracle databases. That is a narrower market — but it is a market Oracle is uniquely positioned to own.

Risks: Debt, Concentration, and Execution

Oracle's AI infrastructure bet carries risks that are different from those facing AWS, Microsoft Azure, or Google. The company is smaller, more leveraged, and more dependent on a handful of customers.

Debt load. Oracle's long-term debt climbed from $85.3 billion to $122.3 billion, and the company plans to raise roughly $40 billion more in fiscal 2027. Credit rating agencies have grown increasingly cautious, with Oracle approaching junk-grade status. The company's debt is approximately 4.3 times EBITDA, a leverage ratio that limits financial flexibility.

Customer concentration. OpenAI accounts for roughly $300 billion of Oracle's $664 billion RPO. If OpenAI's business model fails or its funding dries up, Oracle's backlog could evaporate. The company's own SEC filing acknowledged the risk of "customer non-payment and non-performance."

Execution risk. Oracle is building data center capacity at a pace it has never attempted before. The company delivered 1.2 gigawatts in fiscal 2026 and plans to deliver 850 megawatts per quarter going forward. Any delay in construction, equipment delivery, or customer onboarding could disrupt the revenue trajectory.

Margin compression. GPU rental is a low-margin business. If Oracle cannot layer higher-margin services on top of GPU compute, the AI infrastructure bet will generate revenue without generating profit.

Competitive pressure. AWS, Microsoft, and Google are all investing heavily in AI infrastructure. Google Cloud is growing at 82%, Microsoft at 43%, and AWS at 37%. Oracle's 121% growth is impressive, but it is growing from a smaller base — and the larger players have more capital, more customers, and more established enterprise relationships.

The CODEW Lens: The biggest risk is not that Oracle's AI infrastructure fails. It is that the infrastructure works, the revenue arrives, and the margins never do. Oracle could become a high-revenue, low-margin utility — a very different business than the database company it was built on.

What's Next

Oracle's trajectory over the next 24 months will be defined by three variables:

Backlog conversion. Oracle expects roughly 12% of its $638 billion backlog to become revenue within 12 months. The pace at which RPO converts to revenue will determine whether Oracle's growth is real or simply contractual.

Margin expansion. GPUaaS margins need to move from the mid-teens toward the 30–40% target. If they do not, Oracle will be a high-revenue, low-margin infrastructure provider — a very different business than the database company it was built on.

Customer diversification. Oracle's AI business is heavily dependent on OpenAI. The company needs to broaden its customer base to reduce concentration risk and demonstrate that its infrastructure is valuable beyond a single, capital-hungry model developer.

The company is also expanding its multicloud footprint. Oracle AI Database workloads are now available natively inside AWS, Azure, and Google Cloud, with unified pricing and simplified management. That strategy positions Oracle as a data layer that works across every cloud — and gives it a distribution channel that does not depend on customers choosing OCI.

The CODEW Lens: The next 24 months will tell us whether Oracle's bet is a strategic repositioning or a financial overreach. The backlog is real. The question is whether the economics are.

The CODEW Analysis: Can Oracle Turn Database Dominance into Durable AI Infrastructure?

The short answer is yes — but the company that emerges will be different from the database giant of the last four decades.

Oracle has something that no other AI infrastructure provider has: a database franchise that touches virtually every large enterprise on the planet. That franchise gives Oracle a distribution advantage in AI infrastructure that AWS, Microsoft, and Google cannot easily replicate. When enterprises want to run AI workloads against their existing Oracle databases, OCI is the path of least resistance.

The company has also built infrastructure that is genuinely competitive for AI workloads. Its RDMA networking, flat network architecture, and GPU cluster designs give it performance-per-dollar advantages for specific large language model training and inference workloads. The 97.9% GPU utilization rate is evidence that customers are using the capacity Oracle is building.

But the risks are real and material. Oracle is financing its AI buildout with debt, and its free cash flow is deeply negative. It is dependent on a single customer — OpenAI — for roughly half of its contracted backlog. Its GPU rental business generates margins in the mid-teens, not the 30–40% the company ultimately targets. And it is competing against companies with vastly more capital, more customers, and more established cloud businesses.

The CODEW verdict: Oracle will not become the largest AI infrastructure provider. AWS, Microsoft, and Google will remain dominant. But Oracle can become the most strategically embedded infrastructure provider — the one that connects enterprise data to AI workloads across every cloud, the one that enterprises turn to when their AI initiatives need to run against the databases that already hold their most valuable information.

That is a narrower position than being the biggest cloud. It is also a position that is harder to displace. Oracle's database franchise is its moat. The AI infrastructure buildout is its bridge to the next era of enterprise computing. Whether that bridge holds depends on whether Oracle can convert its backlog into revenue, expand its margins, and reduce its dependence on a single, capital-hungry customer.

The CODEW Lens: Oracle is not betting on AI infrastructure because it wants to be a cloud company. It is betting on AI infrastructure because its database customers need somewhere to run AI. The infrastructure is the means. The database is the business.

The AI Infrastructure Glossary

OCI (Oracle Cloud Infrastructure) — Oracle's second-generation cloud platform, built with a flat network architecture and RDMA networking for high-performance computing.

RPO (Remaining Performance Obligations) — Contracted revenue not yet recognized. Oracle's RPO reached $664 billion in Q1 FY2027.

GPUaaS — GPU-as-a-Service. Renting GPU compute capacity to customers on a pay-as-you-go or contracted basis.

Stargate — A joint AI infrastructure initiative between Oracle, OpenAI, SoftBank, and MGX, targeting up to $500 billion in investment and 10 gigawatts of capacity.

RDMA (Remote Direct Memory Access) — A networking technology that allows direct memory access between servers, reducing latency and improving throughput for distributed AI training.

Multicloud Universal Credits — Oracle's commercial model allowing customers to commit once and deploy Oracle database services across AWS, Azure, Google Cloud, and OCI.

Capital Expenditure (Capex) — Long-term investments in physical assets like data centers, chips, and infrastructure. Oracle's fiscal 2026 capex reached $55.7 billion.

Free Cash Flow — Operating cash flow minus capital expenditures. Oracle reported negative free cash flow of $23.7 billion in fiscal 2026.

Utilization Rate — The percentage of available GPU capacity that is actively serving customers. Oracle reported 97.9% utilization in Q1 FY2027.

Junk-Grade Credit Rating — A credit rating below investment grade (BB+ or lower), indicating higher risk of default and higher borrowing costs.

FAQ

Q: Why is Oracle growing faster than AWS, Microsoft, and Google in cloud infrastructure?

Oracle is growing from a much smaller base. Its OCI revenue was $7.4 billion in the most recent quarter, compared to AWS at $30+ billion. Oracle's growth is driven by AI infrastructure demand — specifically, GPU compute for training and inference — where demand exceeds supply. The company also benefits from its database installed base, which gives it a natural customer base for AI workloads that need to run against Oracle databases.

Q: What is Oracle's biggest risk?

Customer concentration. Approximately $300 billion of Oracle's $664 billion RPO comes from OpenAI contracts. If OpenAI's business model fails or its funding dries up, Oracle's backlog could shrink significantly. The company's own SEC filing acknowledged the risk of "customer non-payment and non-performance."

Q: Can Oracle's GPU rental business become profitable?

Oracle's GPUaaS margins are currently in the mid-teens, well below the company's 30–40% target. Margins could improve as the business scales, as Oracle optimizes its data center designs, and as higher-margin services layer on top of GPU compute. But GPU rental is a commoditized business, and Oracle competes against larger players with more purchasing power.

Q: How does Oracle's multicloud strategy work?

Oracle AI Database workloads can now be deployed across AWS, Microsoft Azure, Google Cloud, and OCI through Multicloud Universal Credits — a single commitment with unified pricing. This positions Oracle as a cloud-neutral data layer rather than a traditional hyperscaler. Multicloud database consumption surged 817% year over year in fiscal Q2 2026.

Q: Is Oracle's AI infrastructure bet a good investment?

That depends on whether Oracle can convert its $664 billion backlog into revenue at acceptable margins while managing its debt load. The company is growing faster than any major cloud provider, but it is also more leveraged and more dependent on a single customer. The bull case is that Oracle becomes the most strategically embedded AI infrastructure provider for enterprise data. The bear case is that it is financing a capital-intensive race it cannot win against larger competitors.

The CODEW Stat

$664B backlog · 121% OCI growth · −$23.7B free cash flow Oracle's remaining performance obligations reached $664 billion in fiscal Q1 2027, up $209 billion year over year. OCI revenue grew 121% to $7.4 billion. GPU utilization ran at 97.9%. But free cash flow was negative $23.7 billion in fiscal 2026, long-term debt climbed to $122.3 billion, and roughly $300 billion of the backlog comes from OpenAI alone. Oracle is building the infrastructure for the AI era. The question is whether the economics of that infrastructure can support the debt required to build it.

The CODEW