Slipping Into the Close: Oracle's $300B Wager, Adobe's Near Miss, and RH's Melting Profit

Selling pressure built into the close: the Dow fell more than 300 points, the S&P 500 closed down a bit over half a percent, the Nasdaq-100 lost about 1 percent, and the market posted its fourth straight down day. The first 20 minutes were calm with no fresh opening lows, but the Q indexes slipped visibly in the final stretch.

The macro backdrop worked against stocks: the 10-year yield pushed toward 4.95 percent, up more than 11 basis points, Brent crude moved higher, and gold struggled to make a new high despite geopolitical tension. Attention turned to the inflation data due the next day; the oil-plus-yields combination weighed on equities.

The evening menu was set: Oracle and Adobe reported about five minutes after the bell, Restoration Hardware about five minutes after that. Adobe carried extra tension with a new CEO appointed roughly a week earlier, while all eyes were on Oracle because of the data-center narrative.

Adobe posted third-quarter revenue of 6.76 billion dollars against expectations near 6.7 billion, only a nominal beat. Earnings per share came in about five cents, roughly one percent, above estimates, with a tight full-year range. Computers will print this as a beat, but it is hardly one to celebrate.

Guidance was the real test: the fourth-quarter outlook midpoint of 6.825 billion landed about 40 basis points below the 6.85 billion consensus. The stock first dropped around 2 percent, then recovered part of the loss, a pattern suggesting sellers may be exhausted. If the break out of the long downtrend holds, it could breathe life into software broadly.

The CEO change adds spice: Anil Chakravarthy succeeds Shantanu Narayen, and the first market reaction was a drop of about 2 percent. The market read it as a strategy confession; the stream cited Meta, recalling the pivot from metaverse excess to AI. In my view a leadership change raises near-term uncertainty, but change at the bottom is sometimes mandatory.

Adobe's AI-first annual recurring revenue grew 150 percent, though off a small base, while the core business stays under pressure. The host described canceling his own subscription, saying chatbots now handle work he once did in Photoshop and Lightroom. That is anecdote, not data, yet it captures the competitive squeeze on design software.

Oracle's picture was more conflicted: license revenue of 655 million missed the 718 million consensus clearly, down 14.5 percent year over year, and the software line missed slightly too. Against that, adjusted total revenue of 19.35 billion beat the 19.13 billion estimate, and the operating margin of 42 percent topped the 40.8 percent forecast.

There were strengths as well: operating income of 8.15 billion beat the 7.81 billion estimate, and adjusted earnings of 1.92 dollars per share beat the 1.75 consensus decisively. Revenue grew about 29.6 percent year over year while expenses rose 18.4 percent, so the top line outran costs. But cost of revenues rose 77 percent, outpacing the related revenue gain, showing growth did not come cheap.

Cost discipline came with job cuts: services expenses fell 4 percent, sales and marketing 12 percent, research and development 4 percent, with software support down 1.3 percent. As the host noted, a data center is no car factory; it needs dozens of specialists, not thousands of workers. The company is cutting where it loses and investing where it wins.

Cash flow shows the price of aggressive growth: operating cash near 23 billion against capital spending of 28 billion, leaving free cash flow around negative 5.4 billion. The company filled the till with roughly 19 to 20 billion of stock issuance and repaid 4.2 billion of debt; cash of 36.3 billion barely covers 33 billion of near-term bills. Long-term debt sits near 35 percent of market value and will grow further to fund the OpenAI contract.

Guidance is bold: second-quarter revenue growth of 30 to 34 percent, full-year revenue of 90 billion, full-year adjusted earnings of 8.10 dollars per share. On that number the stock trades near 20.1 times earnings, with a pre-growth PEG near 0.65. That math makes Oracle look cheap; without the license decline the picture would be far cleaner.

The night's real bomb came not from the statements but from the contract: roughly 300 billion dollars with OpenAI over five years, 4.5 gigawatts of capacity starting in 2027. Annual payments land near 50 to 60 billion, a giant burden next to OpenAI's disclosed revenue near 10 billion. Oracle carries the debt, OpenAI carries the payment promise; the wager runs both ways.

Unit math clarifies the picture: 300 billion over 4.5 gigawatts over five years means 11 to 13 billion per gigawatt per year. That sits inside the 10 to 17 billion band of market comparables. The 50 billion per gigawatt per year scarcity price discussed around xAI Colossus is the exception; long-term contracts come nowhere near it. That compression calls optimistic scarcity-based forecasts into question.

The lesson for the neocloud complex is sharp: short-term spot deals fetch premium prices, long-term contracts do not. The picture is great for hardware and chips, acceptable for Oracle, but it means betting on a single customer. That is why the risky-gamble framing and the 108 billion debt load matter; a bet on Oracle increasingly becomes a bet on OpenAI.

Restoration Hardware looked like a headline beat, but the detail was weak: operating income collapsed from 141 to 107 million, and pre-adjustment income fell 19.5 percent from 72 to 58 million. Net revenue rose 2.5 percent while selling and administrative costs jumped 20.2 percent; the gap widens every quarter.

The tariff detail is critical: the company pulled roughly a year of inventory ahead of tariffs, with cash down to about 41 million; cost of goods sold fell year over year because old stock is being worked through. Six-month revenue growth of 0.5 percent faces expense growth of 9 percent, with bills near 630 million against cash of 125 million. Debt is being paid down, yet the business still bleeds; true margins appear only once market-price purchasing resumes.

Elsewhere the board was mixed: Apple rose 3.4 percent, recovering about 16 points off the bottom; Google held near 331, with 263 in view if it breaks. SanDisk bounced off its technical line, Micron, SK Hynix and Marvell were lower, Red Robin fell 8 percent. Headlines such as Blackstone's interest in AI financing also crossed the stream.

My net read: a good night for the hardware trade, a selective one for software, cheap but OpenAI-linked for Oracle, and too early for Restoration Hardware. Oracle ranks clearly ahead of RH in my book; Adobe's trend turn still needs confirmation.