Signals Inbox·July 28, 2026·AI Infrastructure

Who is the next hyperscaler?

Oracle has the strongest claim to become the fourth global hyperscaler, while CoreWeave is emerging as the leading AI-specific cloud and Alibaba remains the biggest regional alternative.

Q1What does “the next hyperscaler” mean here?

For this article, the next hyperscaler means the next company that can become a fourth global cloud platform alongside AWS, Microsoft Azure and Google Cloud.

Large data centers alone do not qualify. Meta owns vast computing infrastructure, OpenAI buys enormous amounts of capacity, and CoreWeave runs some of the world’s densest GPU clusters. Most outside companies still cannot use those platforms to host everything from a basic website to a banking system.

A full hyperscaler needs several things at once. It must sell computing, storage, networking, databases, security and developer tools. It must operate across many countries, support regulated customers and make capacity available without designing every deployment as a special project.

Scale also has to be repeatable. One enormous AI contract can fill several data centers, but it does not prove that thousands of unrelated companies will keep increasing their spending.

That leaves three serious types of candidate. Oracle is trying to build a full global cloud. Alibaba already runs a broad platform but remains strongest in China and Asia. CoreWeave is building a narrower cloud designed mainly for AI.

Q2Is there really room for another global hyperscaler?

A fourth hyperscaler can still emerge because the cloud market is expanding fast enough to create a large new business without immediately stealing customers from AWS, Azure or Google Cloud.

According to Synergy Research Group, companies spent $129 billion on cloud infrastructure services in the first quarter of 2026. Spending had risen 35% in one year and reached an annualized rate above $500 billion.

At that size, one percentage point of global market share represents roughly $1.3 billion of quarterly spending, or more than $5 billion a year. A provider can add billions in revenue while barely moving on a market-share chart.

AI has made that opening larger. Companies now need huge clusters for training, inference, data processing and model development alongside their existing cloud workloads. Many also want a second or third supplier so they are not completely dependent on one platform.

The opening exists. The hard part is that every major provider can see it. The challenger must expand while Amazon, Microsoft, Google, Alibaba and dozens of specialist clouds are all spending heavily at the same time.

Q3How far behind are Oracle and the other hyperscaler challengers?

Oracle currently leads the challenger group, but the distance between Oracle and Google Cloud remains enormous.

Synergy Research Group estimates that AWS controls 28% of global cloud infrastructure spending, followed by Microsoft at 21% and Google Cloud at 14%. Oracle and Alibaba each hold around 4%.

The gap between Google and Oracle is therefore ten percentage points. At the current market size, that represents almost $13 billion of quarterly customer spending.

Even reaching an 8% share would require Oracle to double its current position while the overall market continues growing. By the time Oracle gets there, 8% could represent a much larger revenue figure than it does today.

CoreWeave, Huawei and Tencent remain another step behind. Each is estimated at roughly 1% to 2% of the global market, depending on the provider and measurement period.

Estimated global cloud infrastructure share, Q1 2026

Q4Why is Oracle now the leading hyperscaler candidate?

Oracle is currently the leading candidate because its cloud infrastructure business is growing fast enough to change the company’s position, rather than merely keeping pace with the market.

Oracle Cloud Infrastructure generated $18.1 billion in revenue during fiscal 2026, up 77% from the previous year. Revenue reached $5.8 billion in the final quarter alone, representing 93% year-over-year growth.

Oracle also moved from roughly 3% to 4% of the global infrastructure market. One percentage point may look small, but it was enough for Oracle to overtake Alibaba in Synergy Research Group’s ranking.

The speed of the change is unusual for a company of Oracle’s size. Infrastructure revenue rose from $3 billion to $5.8 billion in four reported quarters. The amount added each quarter increased too, rather than fading after an early burst.

Oracle has another advantage over newer AI clouds: it already sells essential software to many of the world’s largest companies. Banks, governments, hospitals and multinational businesses have spent decades building systems around Oracle databases and applications.

That existing relationship makes OCI easier to introduce. Oracle can start with a database or an AI cluster and gradually sell computing, storage, analytics and security around it. CoreWeave and other specialists usually have to win the customer relationship from scratch.

Q5Is Oracle’s cloud growth broad, or is one AI boom distorting it?

Oracle’s cloud growth is real and unusually fast, although a small number of giant AI contracts are driving much of the latest acceleration.

OCI infrastructure revenue grew from $3 billion to $5.8 billion across four quarters. Growth also accelerated from 52% to 93%. That progression is too large and too consistent to dismiss as a single announcement.

Oracle’s backlog tells a more complicated story. Remaining performance obligations reached $638 billion, up from $138 billion one year earlier. Most of the increase during the last two quarters came from large-scale AI agreements.

Some customers prepaid Oracle to buy GPUs, while others purchased the GPUs themselves and supplied them to Oracle. Those prepaid and customer-supplied hardware commitments reached $75 billion.

This lowers Oracle’s financial risk because customers are helping fund the equipment. It also shows how concentrated the new demand is. A few multiyear AI contracts can create hundreds of billions in backlog without producing the same customer diversity as AWS or Azure.

Two things are happening at once: Oracle has built a genuinely fast-growing cloud business, and an exceptional wave of contracted AI capacity is pulling that business forward.

OCI infrastructure revenue by reported quarter

Q6Can Oracle become more than a place to rent GPUs?

Oracle can build a broader hyperscaler business because OCI already combines AI infrastructure with databases, enterprise software and more than 200 cloud services.

Oracle currently operates more than 50 public cloud regions across 28 countries. It also offers dedicated and sovereign versions of OCI for governments and regulated companies that cannot place sensitive systems in a normal public cloud.

The database business gives Oracle its best route into those customers. Many large organizations would rather keep an Oracle database than spend several years rewriting the applications connected to it.

Oracle has made that relationship easier to preserve. Oracle database services can now run directly inside AWS, Microsoft Azure and Google Cloud data centers. Customers can keep their main cloud provider while using Oracle technology nearby, with less delay and fewer complicated data transfers.

That multicloud strategy looks odd because Oracle is helping customers stay on rival platforms. It still gives Oracle a way to earn infrastructure revenue wherever the application happens to run.

The next step is harder. Oracle needs customers to add wider OCI services around those databases, including general computing, storage, networking, analytics and security. Otherwise, Oracle may remain an important supplier inside other clouds rather than becoming the customer’s main cloud.

Q7Can Oracle afford to spend like AWS, Microsoft and Google?

Oracle can finance its current expansion, but the buildout is putting much more pressure on Oracle’s balance sheet than the same spending puts on the Big Three.

Oracle generated $32 billion in operating cash flow during fiscal 2026. Free cash flow was negative $23.7 billion, implying roughly $55.7 billion of capital expenditure and related investment.

The company raised $43 billion through debt and another $5 billion through equity during the year. Oracle expects to raise around $40 billion more through debt and equity in the following fiscal year.

Those figures place Oracle’s spending firmly in hyperscaler territory. They also expose a basic disadvantage. Amazon, Microsoft and Alphabet can fund infrastructure using cash generated by several highly profitable businesses.

Oracle has a strong software business, but its cash engine is much smaller. It therefore relies more heavily on customer prepayments, external financing and long contracts tied to individual data-center projects.

That model can work while contracts are secure and customers keep taking the promised capacity. The financial pressure would rise quickly if deployments were delayed, chip prices changed or customers reduced future orders.

Oracle does not need Amazon’s balance sheet. It does need to reach a point where operating cash flow covers a much larger share of new construction.

Q8Is CoreWeave already an AI hyperscaler?

CoreWeave already deserves to be called an AI hyperscale cloud, but it still falls outside the broader hyperscaler category occupied by AWS, Azure and Google Cloud.

CoreWeave generated $2.08 billion in revenue during its latest reported quarter, more than double the previous year. Its revenue backlog reached $99.4 billion.

The company has also passed one gigawatt of active power and secured more than 3.5 gigawatts of contracted power. CoreWeave expects that figure to exceed eight gigawatts by 2030.

Those numbers put CoreWeave among the most important AI infrastructure companies in the world. Meta, Anthropic, OpenAI, Mistral, Cohere and several financial firms have chosen CoreWeave for demanding workloads.

CoreWeave offers more than bare GPU rental. Its platform includes managed Kubernetes, storage, networking, model-development tools and dedicated inference services. The company has built an impressive technical stack around AI training and production inference.

Its limits are still easy to see. A normal enterprise cannot move most of its technology operations onto CoreWeave. The platform does not offer anything close to the range of business applications, databases, integration tools and industry services available from the established hyperscalers.

CoreWeave may become the largest specialist AI cloud. That would be a major outcome, even if it never becomes the fourth general-purpose hyperscaler.

Q9Can CoreWeave escape its dependence on a few huge customers?

CoreWeave’s growth is spectacular, but customer concentration and financing costs make the current business much more fragile than its backlog suggests.

CoreWeave’s latest SEC filing showed that one customer produced 45% of quarterly revenue and a second produced 20%. Two customers therefore generated almost two-thirds of the company’s sales.

The concentration has improved. One customer accounted for 72% of revenue one year earlier. Even so, the loss, delay or renegotiation of one major agreement could change CoreWeave’s outlook overnight.

Financing creates another layer of risk. CoreWeave reported $536 million in quarterly net interest expense on $2.08 billion of revenue. It lost $740 million after interest and other costs.

The company had $11.8 billion outstanding under delayed-draw loan facilities. These loans are generally linked to contracted projects and are repaid as customer revenue arrives, which is more controlled than borrowing without secured demand. The debt burden remains heavy.

CoreWeave also disclosed that all GPUs in its infrastructure currently come from NVIDIA because its customer contracts require them. That close relationship helps CoreWeave win leading hardware early, while leaving little room to switch suppliers or develop cheaper custom chips.

CoreWeave concentration and financing measures

Q10Could Alibaba Cloud become the next global hyperscaler instead of Oracle?

Alibaba Cloud remains Oracle’s closest large-scale rival, but its global position has barely moved despite strong growth at home.

Alibaba’s Cloud Intelligence Group generated RMB41.6 billion, around $6 billion, in its latest quarter. Revenue grew 38%, while revenue from outside Alibaba’s own businesses increased 40%.

AI has become a serious business inside the division. AI-related product revenue reached almost RMB9 billion and recorded triple-digit year-over-year growth for the eleventh consecutive quarter. Alibaba said AI products represented 30% of external cloud revenue.

Alibaba is also investing heavily. Its annual capital expenditure reached RMB126.1 billion, up from RMB86 billion, with cloud infrastructure and computer equipment among the main uses of that money.

These are hyperscaler-level numbers. Alibaba already offers computing, storage, databases, security, networking, AI models and developer services across a wide international footprint.

The problem is the lack of global share gains. Synergy Research Group has placed Alibaba at roughly 4% for several years. Revenue is growing, but the overall market is growing at a similar pace.

Alibaba’s strongest position remains China and parts of Asia, where local relationships, regulations and its Qwen model ecosystem give it real advantages. Breaking into large Western enterprise accounts has proved much harder.

Oracle currently looks more likely to create a fourth worldwide platform. Alibaba looks more likely to remain a powerful regional hyperscaler with selective international strength.

Q11Could Huawei Cloud or Tencent Cloud jump ahead of Oracle?

Huawei Cloud and Tencent Cloud are too strong to dismiss in China, but neither currently shows a transparent global path past Oracle or Alibaba.

Both providers hold roughly 2% of the global infrastructure market according to Synergy Research Group. Their domestic ecosystems give them a substantial customer base, and both can fund years of infrastructure development.

Huawei is particularly difficult to judge. The company generated RMB880.9 billion in total revenue during 2025 and spent RMB192.3 billion on research and development. Its technology spans telecom networks, data centers, cloud services, AI chips, operating systems and enterprise hardware.

Huawei does not publish enough separate financial information about Huawei Cloud to show how quickly the platform is gaining external customers or whether it is closing the gap internationally.

Tencent has a similar reporting problem. Its cloud operation benefits from WeChat, gaming, payments and a large Chinese developer ecosystem, but Tencent does not disclose a clean cloud-infrastructure revenue figure comparable with OCI or AWS.

Both companies may become larger regional powers, particularly as China builds a more independent computing stack. The public evidence supports Oracle more strongly because Oracle is gaining measured global share and reporting rapid standalone infrastructure growth.

Q12Are OpenAI, Meta, xAI and the neoclouds changing the hyperscaler race?

OpenAI, Meta and xAI are creating hyperscale demand, while neoclouds are selling specialized capacity. Neither group currently offers a fourth broad public cloud.

OpenAI has signed enormous infrastructure commitments with Microsoft, Oracle, CoreWeave and AWS. Meta is buying capacity from CoreWeave while building its own data centers. xAI has assembled large clusters to train and run its models.

These companies operate or consume infrastructure at a scale comparable with major cloud providers. Their main purpose is still to power their own models and products.

Turning that infrastructure into a public cloud would require billing systems, customer support, security certifications, developer tools, service guarantees and a much wider product catalogue. Owning servers covers only part of that work.

Neoclouds such as CoreWeave, Nebius, Lambda and Crusoe are closer to being cloud providers. They sell infrastructure to outside customers and often deploy new AI hardware faster than the largest platforms.

Their likely role is becoming clearer. They can handle specialist workloads, provide emergency capacity and build large clusters for customers that cannot wait for a traditional cloud region.

That creates a permanent new layer in the market. A specialist can become a multibillion-dollar supplier while customers continue using Azure or Google Cloud for everything surrounding the AI workload.

Q13Are AWS, Azure and Google Cloud now too far ahead for Oracle to catch?

AWS, Azure and Google Cloud are widening the absolute gap even while Oracle grows faster in percentage terms.

AWS generated $37.6 billion in its latest reported quarter, up 28%. AWS also produced $14.2 billion of operating income, giving Amazon a large pool of profit to reinvest in chips, regions and data centers.

Microsoft reported 40% growth for Azure and other cloud services. Microsoft Cloud revenue reached $54.5 billion, supported by Microsoft 365, security, databases, developer tools and a large enterprise sales network.

Google Cloud has accelerated sharply. Its latest quarterly revenue reached $24.8 billion, up 82%, while operating income rose to $8.8 billion. Google’s cloud backlog reached $514 billion.

Oracle’s $5.8 billion infrastructure quarter was still less than one-sixth of AWS revenue. Oracle’s reported figure excludes cloud applications, while AWS includes a somewhat broader mix, so the comparison is not exact. The order of magnitude is still useful.

Percentage growth can mislead. Adding 90% to a $3 billion business creates less new revenue than adding 30% to a $30 billion business.

Oracle can become the fourth hyperscaler without catching AWS. Catching Google Cloud would require several years in which Oracle adds more revenue in absolute dollars, not merely a higher percentage.

Q14What could stop Oracle from becoming the fourth hyperscaler?

Oracle’s biggest risk is ending up with enormous contracted AI capacity while the wider OCI business stays too concentrated.

The first danger is customer concentration. Oracle has said that most of the recent jump in backlog came from large AI agreements. A few customers can make growth look extraordinary while leaving the company exposed to changes in their model strategy, funding or hardware choices.

The second danger is financing. Oracle is currently spending far more cash on infrastructure than its normal operations generate after other costs. Customer prepayments reduce that burden, but they do not remove execution risk.

The third danger is weaker economics once GPU supply catches up with demand. Renting similar NVIDIA clusters can become price-sensitive when several providers have comparable capacity. Oracle needs databases, software and managed services to protect margins.

The fourth danger is construction. Power connections, chips, cooling systems and data-center buildings must all arrive in the right order. A delay in one component can postpone billions in revenue.

Oracle also needs ordinary enterprise customers to follow the AI buyers. The company will struggle to earn full hyperscaler status if most new usage comes from a handful of model developers.

None of these risks makes failure likely on its own. Put together, they explain why Oracle’s recent revenue growth is convincing while the final verdict remains open.

Q15What would prove that Oracle has truly become a hyperscaler?

Oracle will have earned hyperscaler status when its market share, customer mix and cash flow all improve at the same time.

A sustained global infrastructure share around 8% would be a meaningful starting point. At the market’s current size, that would represent more than $40 billion in annual customer spending, and the required figure will keep rising.

Oracle also needs broader growth. Databases, general computing, storage, networking, analytics and security should expand alongside the giant AI clusters.

Customer diversity will be equally important. We would want to see many companies increasing normal OCI consumption, rather than a few contracts creating most of the backlog and construction demand.

The financial test is straightforward. Oracle should return to positive free cash flow while continuing to expand capacity. Repeated debt and equity issuance cannot remain the normal funding model forever.

These targets are demanding because the label should mean something. Building a few enormous campuses demonstrates technical scale. Running a global platform requires customers, products and cash generation that remain strong across several cycles.

What would establish Oracle as a true fourth hyperscaler

Q16Who is the next hyperscaler?

Oracle is the next hyperscaler, although calling it a fourth equal to AWS, Azure and Google Cloud would still be premature.

Oracle currently combines the strongest set of ingredients: rising global market share, fast infrastructure growth, more than 50 cloud regions, a broad service catalogue and deep relationships with large enterprises.

CoreWeave has the best chance of becoming the dominant AI-specific cloud. Its growth, backlog and infrastructure scale are already remarkable. Customer concentration, debt and a narrower product range keep it in a different category for now.

Alibaba Cloud remains a real hyperscaler across China and much of Asia. Its revenue and AI growth are strong, but several years of flat global share make an international breakthrough less convincing.

Huawei, Tencent and the smaller neoclouds may become major regional or specialist providers. None currently matches Oracle’s combination of global reach, enterprise software and measured share gains.

Oracle has moved beyond being a distant cloud challenger. It now has a credible route to becoming the fourth worldwide platform.

The transition is unfinished. Oracle still needs to double the size of OCI, bring in a wider range of customers and prove that the infrastructure can eventually fund itself.

For now, the global cloud market remains a Big Three followed by Oracle. Among the challengers, Oracle is the one most likely to turn that ranking into a genuine four-company hyperscaler market.