Pre-Market Briefing on Sept 30: Oracle Jumps 3.99% on 25 Agentic Apps While Micron Reports Today

All three major indexes edged lower on Tuesday, September 29: the S&P 500 closed at 7,670.04 for −0.18%, the Nasdaq at 26,797.54 for −0.09% and the Dow at 51,349.92 for −0.26%. The indexes barely moved, but the sectors inside them changed places: among Nasdaq sectors utilities averaged 1.50% at the top and technology −0.61% at the bottom. The stock in focus is Oracle (ORCL), which closed at $137.90 for a 3.99% gain after shipping Fusion Claw, a layer that connects AI reasoning to the deterministic computation an enterprise runs on, with 25 agentic applications released at once. What was announced is the count and the scope of those applications, and how fast customers adopt the 25 will not be disclosed until a later quarterly report. Today, September 30, August core PCE and the final second-quarter GDP reading land at 12:30 UTC, and Micron (MU) reports at 20:00 UTC after the close. The data in this article is as of the September 29 US close, and all times are UTC.

 

Unit: %, single-day change on September 29 against the prior close. All three sit inside 0.3%: the Dow fell 0.26%, the most, and the Nasdaq 0.09%, the least.

 

Three bars hugging zero is usually not a view, it is a wait. The card's explanation is plain: the data and the results are stacked from Wednesday onward, so money did not add ahead of them. When the tape looks like this, the useful question is not why the index fell but what sits on the calendar next. Flat is often a timing problem rather than a valuation one.

 

Unit: %, Nasdaq sector averages for the session, with a zero line because the set spans positive and negative. Utilities averaged 1.50% at the top and technology −0.61% at the bottom, 2.11 points apart.

 

The index moved 0.09% while the ends of the sector table sat two points apart, so the money that left did not leave the market; it moved into the defensive box. This is why reading the index alone loses half the information: an index is a weighted average, and only the sector table shows direction. On a day when the headline barely moves, open the sector table first, because that is usually where the positioning shows up.

 

Unit: %, single-day change on September 29. These six are grouped by hand as enterprise software and cloud platforms, which is not an exchange industry classification. Oracle led at +3.99%, Adobe came second at +0.94%, ServiceNow fell 1.15% at the other end, and the ends sit 5.14 points apart.

 

Set the basis first: NYSE software infrastructure averaged 1.70% the same session, and that is the industry-field average, not an average of the six shown. The average was positive, so Oracle led a rising group rather than moving against it; three of the six still fell, which shows that "the group rose" and "its peers rose" are different statements. The extra 2.29 points came from Fusion Claw, released that day, which connects AI reasoning to the deterministic computation an enterprise runs on and arrived with 25 agentic applications plus two compliance tools. Reasoning can guess and a ledger cannot, and that layer joins the two.

 

To judge whether a gain belongs to the company, find the industry average first, then see whether the peers followed. A positive average with a split group usually points back to something the company announced that day.

 

Unit: 0–100, against peers and its own one-year history, 0 at the centre and 100 at the rim, all five axes on one scale, over a one-year window rather than the single session. Peer rank scores the full 100, relative strength reads 61 and volatility control 51, while trend position reads 11 and valuation heat reads 1.

 

The shape is unusual: one axis reaches the rim and two collapse almost to the centre. It is the same stock measured with two different rulers. Peer rank answers where it placed in this group today; trend position and valuation heat answer where it has travelled over a year. Strongest on the day and low in its own range can both be true at once.

 

When you read a radar, separate the axes that measure today from the axes that measure a year. Read them together and "rose the most today" turns into "has already risen a lot", which is a different claim.

 

Unit: US$ billion, quarterly revenue on the company's reported basis, four bars drawn from zero. ⚠️ The last bar, $50B, is the midpoint of company guidance, plus or minus $1B, and not a reported result. The four quarters run $13.64B, $23.86B, $41.46B and then the $50B guide, with the slope rising the whole way.

 

Why read the quarter rather than the year? Memory is cyclical, and an annual figure mixes in last year's low base, so only the quarterly series shows whether the slope is still lifting. The card adds that this bar is not the one to read first: revenue is the number the company already guided, and gross margin is the one the market cannot pre-empt.

 

Unit: %, year-on-year revenue growth for the four business lines last quarter, with the size of each line in brackets, four bars drawn from zero. The denominator here is each business line rather than total revenue above, so the two charts cannot be read across. All four lines more than tripled, and only core data centre jumped to 653%.

 

The difference is who the customer is. Cloud memory, $13.8B, sells to the cloud providers and carries the HBM; that line was already full a year ago. Core data centre, $11.5B, sells ordinary server memory to enterprises and server makers, and it only filled up this quarter: $11.524B against mobile and client at $11.521B, level for the first time, where a year earlier it was under half the size.

 

When a company grows fast, ask which line is doing it. A shortage confined to the top of the chain and a shortage that has spilled downstream describe two very different widths of demand.

 

Unit: %, drawn from 80% rather than from zero, because both readings fall between 84 and 86 and the 1.4-point gap would otherwise be invisible. Last quarter reported 84.6% and the company guides the current quarter to about 86%. ⚠️ Guidance is not a reported result, so the two bars are different kinds of number.

 

A commodity carries no brand premium, so its margin is set by supply and demand. Printing 84.6% has one explanation: buyers are paying up to jump the queue for the same output. Guiding another 1.4 points on top says the queue is still there.

 

For a commodity producer, the gross margin is a direct reading of how tight capacity is. The quarter it fails to hold is the first signal that the pricing cycle has peaked, and it arrives earlier than revenue and is harder to dress up.

 

Unit: %, calculated as (latest close − the 52-week low) divided by (the 52-week high − the 52-week low), times 100, six bars drawn from zero on the same basis. Meta reads 84.2 at the top, ServiceNow 43.6, and Oracle, today's stock in focus, only 11, so one ruler spans 73 points.

 

The order barely matches how well the agents are selling. Salesforce grew Agentforce recurring revenue past $1.5B, up more than 240%, and reads 64.3. ServiceNow passed $1B in AI contract value with subscription revenue up 24.5%, and reads 43.6. Meta's Muse reached the top of the US App Store free chart ten days from launch, and Meta reads 84.2.

 

The difference is not the product; it is whose book the revenue lands on. Meta opened one more entry point on an existing social and advertising book, so every extra use is a net addition. Software priced per seat works the other way: the agent finishes the workflow, so the customer needs fewer seats, and the better it works the more it eats its own pricing model. For ServiceNow, the good news and the bad news are the same fact.

 

When a company ships an agent, do not start by asking how good it is. Ask whether that revenue is added or swapped. Added revenue is growth; swapped revenue has to pay off the old line first. That question works on any company now selling AI features.

 

 

A: No. The 1.70% is the NYSE software infrastructure average and covers far more companies than the six shown, which are a hand-picked comparison set.

 

A: It led. The industry average was positive and it was the strongest of the group. Moving against the group requires a negative average, which was not the case.

 

A: No. It measures where the price sits between the high and the low of the past year, which is altitude rather than valuation, and it carries no view on what comes next.

 

A: $50B is the midpoint of company guidance, plus or minus $1B, not a reported result. The card suggests reading the margin first: 84.6% last quarter against guidance of about 86%, and a miss there matters more than a small miss on revenue.

 

A: Because it shows the width of the shortage. Cloud memory was already full a year ago, while core data centre only filled up this quarter, which means the crowding from AI servers has spilled into general-purpose servers.

Disclaimer: This article is compiled and written by the MEXC RealStocks team. The data in this article is based on the closing of the US stock market on September 29, 2026. The content is a compilation of public market information, and individual stocks are publicly discussed targets, which do not represent the recommendation or opinion of MEXC and do not constitute any investment advice. More US stock content: @MEXC | @Alpha_MEXC | @MEXC_Research