Oracle locks in $16 billion for massive AI data center campus in Michigan amid supply shake‑up and investor scrutiny

Home » Python » Oracle locks in $16 billion for massive AI data center campus in Michigan amid supply shake‑up and investor scrutiny

Oracle is pushing hard into the next phase of the cloud and artificial intelligence race with a huge bet on physical infrastructure in Michigan. The company has arranged roughly $16 billion in financing to build what is set to become its largest AI data center campus in the United States, even as markets question the cost, the growing debt load and the pace at which these investments can be turned into profit.

Behind the headlines about record funding, there is a more complex picture: surging demand for AI compute, aggressive capital spending, a reshuffling of hardware suppliers and mounting investor nerves. Oracle is trying to close the gap with hyperscale rivals like Amazon, Microsoft and Google, but it has to do so while managing regulatory risk, legal disputes and a balance sheet that is getting heavier by the quarter.

The Barn: Oracle’s gigawatt‑scale AI hub in Michigan

At the center of this expansion is a project nicknamed “The Barn” in Saline Township, near Ann Arbor, Michigan. Oracle is building a campus of three single‑story buildings with a combined surface of about 1.65 million square feet. Designed to exceed one gigawatt of power capacity, the site is being positioned as one of the largest AI‑ready data center complexes in the country.

The Michigan campus is closely intertwined with Oracle’s partnership with OpenAI. The facility is expected to support large‑scale training and inference workloads for generative AI, tying into broader plans for a massive high‑performance computing build‑out sometimes linked to OpenAI’s long‑term “Stargate” ambitions. In practice, this means racks of GPUs, high‑bandwidth networking and dense power distribution on a scale that only a handful of operators can handle.

Construction on the site quietly began in February, well before the financing structure was fully disclosed. By locking in a dedicated funding package, Oracle is essentially ring‑fencing the capital needed for The Barn, rather than relying exclusively on its corporate balance sheet to pay for every phase of the project.

In internal and public commentary, Oracle has framed Michigan as a strategic node in its global Oracle Cloud Infrastructure (OCI) network. The location gives access to industrial land, regional power grids and a labor pool that is less expensive than in coastal tech hubs, while still being close to major transport and fiber routes across the Midwest.

How the $16 billion package is structured

The funding arrangement for The Barn stands out not only because of its size, but also because of its structure. According to people familiar with the deal, around $16 billion has been assembled through a mix of equity and long‑term debt, making it one of the largest single data center financing packages seen so far in the U.S. market.

On the debt side, the heavy lifting comes from the bond market. Pacific Investment Management Co. (PIMCO) has reportedly taken roughly $10 billion in privately placed notes, issued under Rule 144A and maturing in 2045 with a coupon of about 7.5%. These securities are designed for institutional investors, effectively locking in fixed financing costs over a long horizon while giving Oracle the upfront cash to build out capacity.

Equity contributions fill in part of the remaining gap. Related Digital and Blackstone‑linked funds are said to be injecting around $2 billion of equity capital into the project, with the rest of the $16 billion coming from additional debt tranches arranged by a banking syndicate led by Bank of America. In some tellings of the deal, Blackstone’s exposure has been highlighted alongside Bank of America’s role as a lead structuring adviser.

This aggressive build‑out carries consequences for the balance sheet. Oracle’s long‑term debt has swelled to roughly $124.7 billion, and interest expenses have jumped by more than 30% year‑on‑year to around $1.18 billion. At the same time, free cash flow has slipped into negative territory, largely because capital expenditures have ballooned to the vicinity of $48-50 billion as the company funds data center projects like Michigan and other hyperscale facilities worldwide.

For context, project finance of this magnitude for a single data center would have been unthinkable just a few years ago. Traditional deals in the sector rarely exceeded $3-4 billion for one site, but AI has changed the economics: estimates suggest that every new gigawatt of cloud capacity can demand $6-10 billion in capital, depending on land, power and chip costs. Oracle’s Michigan package highlights how far the bar has moved.

Why Michigan became the site of Oracle’s biggest AI campus

The decision to plant this flagship data center in Michigan is not random. As Oracle and its peers race to secure power, land and regulatory approvals, the state offers a specific mix of industrial legacy and energy infrastructure that fits AI workloads.

Decades of restructuring in the auto industry have left large tracts of industrial‑zoned land near good transport links, making it easier to assemble sizable campuses without the community opposition and zoning constraints seen in more densely populated or saturated regions. Additionally, Michigan is tied into the Midwestern power grid, giving operators access to multiple generation sources and interconnection points.

Energy costs loom large in the economics of AI. For hyperscale data centers dominated by GPUs and high‑density racks, electricity can account for 35-45% of total operating expenses. Securing stable, long‑term power contracts at predictable prices can be as important as the cost of the servers themselves. Oracle appears to be betting that Michigan can offer the right combination of availability and tariff stability over a 15‑ to 20‑year horizon.

The region is also becoming part of a broader map of North American compute hubs. Microsoft and Meta have been channeling investment into nearby states, as Northern Virginia—the long‑time epicenter of U.S. data centers—grapples with grid constraints and local resistance to further expansion. Utilities in Virginia have already started to slow approvals for new large‑scale connections, pushing hyperscalers to look elsewhere.

For Oracle, this geographic shift opens a window. By getting an early foothold with a gigawatt‑scale site in the Midwest, the company positions itself to negotiate with large AI customers on more equal footing than in the past, when its capacity trailed the hyperscale leaders by a wide margin.

Record AI demand, a swelling backlog and rising capital needs

The rationale for building The Barn is straightforward: demand for Oracle’s cloud infrastructure has been outpacing available supply. In its recent fiscal third quarter, the company reported that cloud infrastructure revenue jumped about 84% to roughly $4.9 billion, driven largely by AI‑related workloads.

Oracle’s obligations to deliver future services have also exploded. Across its cloud and AI contracts, remaining performance obligations have been cited as high as $553 billion, representing an increase of more than 300% year‑on‑year. In earlier disclosures the company mentioned a backlog above $130 billion, with a big portion tied to deals like the one signed with OpenAI. Whichever figure one focuses on, the direction of travel is the same: contracted demand is piling up much faster than the company can currently build data centers.

To close that gap, Oracle expects to raise and deploy enormous sums over the next few years. Management has talked about bringing in roughly $45-50 billion of fresh capital in 2026 alone to fund capacity expansion, while also guiding toward potential revenues of around $90 billion by fiscal 2027 if the backlog can be translated into active workloads.

That said, there is a notable mismatch between long‑term commitments and short‑term accounting. While the backlog jumped by around $30 billion between recent quarters, short‑term deferred revenue has stayed roughly flat near $9.9 billion. Analysts argue that this gap underscores the time lag between signing AI deals and recognizing them as revenue, particularly when large chunks of capacity are still under construction.

This delay feeds into a broader concern on Wall Street: the cost of building the infrastructure is immediate, but the revenue flows are spread over many years. Until utilization ramps up and AI services are fully monetized, cash flow will likely remain under pressure, even if headline demand stories look impressive.

A $1.1-$1.4 billion server order canceled over legal concerns

Just as the Michigan financing deal was coming together, Oracle’s supply chain and investor narrative were jolted by an unexpected move. On 23 April, the company canceled a massive server order with Super Micro Computer, a prominent maker of high‑end AI racks, triggering a sharp reaction in both firms’ share prices.

According to research from Bluefin, the cancellation involved between 300 and 400 GB300‑NVL72 racks equipped with Nvidia GPUs, with each rack priced at roughly $3.5 million. That puts the value of the dropped contract in the range of $1.1-$1.4 billion—a meaningful number for any supplier, and far from a routine adjustment.

Market sources have linked the decision to regulatory and legal risks rather than a change in Oracle’s AI roadmap. A co‑founder of Super Micro has faced accusations related to the alleged illegal export of chips to China, and Oracle is understood to have stepped back from the relationship to avoid being caught up in compliance issues. In effect, the tech giant appears to be de‑risking its supply chain at a time when AI hardware is already constrained.

The fallout has created opportunities for competitors. Taiwan‑based Wiwynn is reported to have taken over the canceled Super Micro racks, while Dell Technologies and Hewlett Packard Enterprise are being mentioned as alternative suppliers that may benefit from hyperscaler clients looking to diversify away from any perceived regulatory exposure.

Financial markets did not take the news lightly. Oracle shares fell by roughly 4.5-6% in the immediate aftermath of the cancellation, wiping out part of the gains accumulated over the previous month. The drop reinforced a growing sense among investors that the company’s ambitious hardware rollout is exposed not only to execution risks, but also to the legal and geopolitical minefield surrounding advanced chips.

Stock performance, technical signals and analyst views

Despite monumental AI headlines, Oracle’s share price has had a bumpy ride. So far in 2025, the stock is down about 12-13% and trades roughly 48% below its 52‑week high, even after a double‑digit rebound over the last month. Recent quotes in Frankfurt have hovered around €145-146, and from the 1 January level near $161.90, the name has logged a noticeable year‑to‑date decline.

On a technical basis, indicators point to stress as well as opportunity. The relative strength index (RSI) has been sitting in deep oversold territory, at around 20.7 in one reading and roughly 26.5 in another, depending on the specific time window used. At the same time, the stock has climbed more than 15% over the past month and now trades about 21-27% above its early‑February low, while still lagging its 200‑day moving average, which sits closer to €183.

Analysts remain broadly constructive, but not without caveats. Wedbush Securities recently initiated coverage with an “Outperform” rating and a price target of $225, describing Oracle as an undervalued infrastructure provider in the AI sector and arguing that the market may be overestimating the risks tied to its investment cycle.

The broader analyst community is also leaning positive. Out of roughly 40-46 firms following the stock, around three‑quarters rate Oracle as a buy, with consensus price targets clustered near $260-261 and a range that stretches from about $160 to as high as $400. Even so, some houses are getting more cautious: Morgan Stanley has trimmed its target from $213 to $207 and kept an “Equal Weight” stance, citing uncertainties around margins in the GPUs‑as‑a‑service business.

Institutional investors appear to be selectively adding exposure. Recent filings show that firms such as Eagle Wealth Advisors LLC and Anchyra Partners LLC have opened or expanded positions, buying a few thousand shares each. These are modest stakes in the grand scheme of things, but they hint at ongoing portfolio rotations ahead of what many see as a decisive phase in Oracle’s AI story.

Financial results, guidance and the backlog monetization puzzle

Against this backdrop, Oracle’s recent earnings have been solid, but the big question is timing. In the latest reported quarter, the company delivered earnings per share of around $1.79, beating consensus estimates of about $1.71. Revenue came in near $17.19 billion, up roughly 21.7% from the same period a year earlier, helped by strong cloud infrastructure growth.

Looking ahead, management has guided for earnings per share of roughly $1.96-$2.00 in the upcoming quarter, on the back of an expected 22% revenue increase over the prior period. For the fiscal fourth quarter, investors will be watching closely to see whether the combination of backlog conversion, new AI deals and data center ramp‑up can offset higher interest expenses and capital outlays.

The key tension lies in how quickly those massive AI commitments can be transformed into actual cash. Orders and remaining obligations may total hundreds of billions of dollars, but they are spread over multi‑year horizons and depend on new capacity like The Barn coming online and filling up. Until utilization and pricing stabilize at attractive levels, the gap between contracted demand and recognized revenue will continue to shape the narrative.

Capital intensity is front and center here. For 2026, Oracle expects to deploy around $45-50 billion in gross capital inflows and investment, much of it into data centers, GPUs and network infrastructure to support partners such as OpenAI, Nvidia, xAI and Meta. That means the company has to walk a fine line: expand fast enough to satisfy key AI customers, but not so fast that its leverage and free cash flow metrics spook credit markets and equity holders.

Analysts and investors have repeatedly flagged this as the core risk. If revenue growth from AI infrastructure fails to ramp quickly enough, the result could be a prolonged period of elevated debt, pressured margins and volatile sentiment around the stock, regardless of how compelling the long‑term AI demand story looks on paper.

Legal pressures and questions about disclosure

Adding another layer of complexity, Oracle is also dealing with a securities class‑action lawsuit alleging misleading statements about its AI infrastructure strategy. Plaintiffs argue that the company and certain executives overstated how quickly heavy investments in data centers and GPUs would translate into accelerated revenue growth, while downplaying or failing to fully disclose the potential impact on leverage, credit ratings and free cash flow.

Tension around disclosure reached a peak in December 2025. On 11 December, Oracle’s shares fell almost 11% in a single session after the company reported results that missed some expectations and outlined plans for roughly $50 billion in capital spending. For critics, that drop illustrated how sensitive investors are to any suggestion that AI promises may not immediately translate into profitability.

None of this has stopped Oracle from investing, but it has sharpened the debate over risk. Supporters point to strong revenue growth, a deep backlog and major institutional backing from names like PIMCO and Blackstone as evidence that the strategy makes sense over a multi‑year horizon. Skeptics counter that the combination of big checks, long build times and unpredictable AI economics leaves little margin for error if demand or pricing were to cool off.

For now, the fourth‑quarter and upcoming annual results are being treated as a kind of stress test. They will offer the clearest view yet of whether Oracle can start to narrow the gap between headline AI contracts and actual cash generation, and whether projects like The Barn can begin to contribute meaningfully to earnings instead of just to debt totals.

Stepping back from the day‑to‑day market swings, Oracle finds itself at a pivotal moment: a record‑breaking $16 billion financing for a gigawatt‑scale AI campus in Michigan has given the company real muscle in the infrastructure race, but it also locks in years of elevated capital commitments and places the spotlight firmly on execution. With demand for AI compute still running hot, a massive backlog to work through and high‑profile partners depending on fresh capacity, the success or failure of The Barn and similar projects will go a long way toward determining whether Oracle’s current strategy is remembered as a bold, well‑timed move or an overextended gamble on a still‑forming market.

Artículo relacionado:

OpenAI and Oracle unveil Stargate’s 1‑GW Michigan data center campus