Oracle defused ‘the key risk going into 2026,’ BofA argues, but the market isn’t buying it

“Every morning the opening screen on my Bloomberg is what’s going on with CDS spreads on Oracle debt,” Morgan Stanley Wealth Management CIO Lisa Shalett told Fortune in October, seeming to speak for a market that was increasingly worried about the bursting of a bubble in artificial intelligence. CDS, as students of the 2008 financial crisis know, stands for “credit default swaps,” a financial instrument to hedge against giant debt loads elsewhere in the market. And the reason Shalett highlighted Oracle’s CDS was that the Larry Ellison–founded software giant has stood out as a relative anomaly among the “hyperscaler” companies fueling billions in data center investment for having just too much debt.

“If people start getting worried about Oracle’s ability to pay,” Shalett told Fortune, “that’s gonna be an early indication to us that people are getting nervous.”

That’s why Bank of America Research wrote on Tuesday that “the lack of clarity on hyperscaler borrowing was the key risk going into 2026,” and why a single press release from Oracle on Sunday carried so much weight, not just with Oracle investors but for the entire AI trade.

Announcing its financing plan for 2026, Oracle said it expects to raise $45 billion to $50 billion of gross cash proceeds, and plans to achieve this funding objective by “using a balanced combination of debt and equity financing to maintain a solid investment-grade balance sheet.” The most significant bit, according to BofA Situation Room analysts Yuri Seliger and Sohyun Marie Lee, is that Oracle plans for a single bond deal to cover its debt borrowing needs for the full year, after which it priced $25 billion of bonds on Monday.

“This transparency on the timing and the amount of Oracle supply is supportive for the broader market,” the analysts wrote, given how nervous credit markets and analysts like Shalett had been through the back half of 2025. This announcement “chips away at hyperscaler supply risks” by providing absolute certainty on both the timing and magnitude of Oracle’s market participation, the analysts wrote. The equity market didn’t exactly agree.

A catalyst for stability

By defining the upper limit of its borrowing, BofA argued that Oracle turned a potential supply glut into a supportive signal for the broader high-grade market. The positive ripple effects were evident almost instantly, with BofA noting that bonds for fellow hyperscaler Meta were trading about three basis points tighter on Monday afterward.

BofA suggested that this set a constructive precedent for the sector. Future bond deals from other tech giants are now likely to act as positive market catalysts rather than disrupters. For a new deal to act as a negative catalyst now, the supply would need to be significantly larger than these aggressive expectations, a scenario analysts view as challenging, given that the market has already priced in up to $300 billion of hyperscaler supply.

There’s just one problem with this thesis: what happened to Oracle stock later on Monday—and so far on Tuesday. The reason why says a lot about the importance of corporate communications at this juncture in the AI hyperscaler trade.

The OpenAI conundrum

The positive vibes from Oracle’s Sunday press release were erased—and more—by a lone tweet from the company.

“The Nvidia-OpenAI deal has zero impact on our financial relationship with OpenAI,” the company posted on X at noon. “We remain highly confident in OpenAI’s ability to raise funds and meet its commitments.” The stock immediately reversed, erasing a roughly 2% gain and trading down 2% instead, before extending its fall on Tuesday, down more than 3%.

Oracle’s been having a rough go of it. The stock is down almost 12% in just five days, and more than half its value has been wiped out since its September high. Investors are punishing the company over its increasingly unruly balance sheet: Oracle is already carrying roughly $100 billion in debt, with plans to take on another $50 billion to fund what it has cast as the crown jewel of its AI strategy: massive data centers built largely to serve OpenAI.

So far, that strategy has proved difficult to turn into pure growth.

For one, demand is outpacing supply. Oracle has said its data-center expansion is running into labor and equipment shortages, delaying some build-outs and pushing revenue further into the future. “The world of bits moves fast. The world of atoms doesn’t,” data center expert Jonathan Koomey previously told Fortune. “And data centers are where those two worlds collide.”

Second—and more troubling for investors—Oracle is increasingly exposed to a single and highly opaque customer. A significant share of those data centers are being built for OpenAI, a private company with over $1 trillion in obligations and only about $20 billion in revenue. Investors have begun questioning how OpenAI can scale its revenue without continual, massive funding rounds, and because the company is private, markets have none of the transparency they would normally demand from an entity this systemically important.

That anxiety is spilling over into the public markets. Microsoft shares dropped 12% after the company disclosed that 45% of its future cloud growth is tied to OpenAI, while Nvidia has slid in recent days amid reports that its expected $100 billion OpenAI investment may be smaller than anticipated.

However, the risks matter more for companies that have already taken on leverage to meet OpenAI-driven demand. Oracle has nearly $250 billion in long-term leasing commitments tied to data centers with life spans of 15 to 20 years, many of which it expects to sublease on shorter time horizons. If demand falters, or capital tightens, Oracle could be left holding the debt long before the cash arrives.

This story was originally featured on Fortune.com