GPT-6 Astra Sent Salesforce, ServiceNow, and Intuit Down 4% to 5% in a Day. Qualcomm and Intel Rallied That Same Afternoon.

Salesforce closed down 4% on Tuesday. ServiceNow fell nearly 5%. Intuit, already down more than half from its 52-week high, slid another 4%. None of the three companies reported earnings or issued new guidance that day. What moved their stock was a product launch at a different company entirely: OpenAI's release of GPT-6 Astra, an AI model built to operate software rather than just talk about it.

OpenAI describes Astra as capable of filling out online forms, updating customer records inside a CRM, organizing a calendar, drafting a document, installing and testing software, and fixing whatever it sees go wrong on a screen. On OSWorld 2.0, a benchmark built around exactly these kinds of computer-operating tasks, Astra scored 72.6%, up from 65.7% for OpenAI's previous flagship model. The jump on that particular test matters because it measures the clicking, typing, and navigating that has historically needed either a human being or a piece of purpose-built software standing between the two.

[[IMG1]]

The selloff wasn't evenly distributed. Investors sold the companies whose businesses look most like a set of forms, workflows, and record-keeping tasks, the exact category Astra is built to automate. Intuit has taken the hardest and longest hit of the three. Its stock traded above $700 within the past year and closed Tuesday near $319, a decline that began with the company's own disappointing fiscal 2027 guidance in August and has widened with each AI product announcement since.

This Is the Third Time This Year

Software stocks have now had this kind of afternoon three separate times in 2026. In January, Anthropic announced an AI agent capable of automating legal research, data analytics, and sales tasks, and Salesforce, Intuit, and Workday all fell between 7% and 10% that same day. The selloff extended into February, when the S&P 500 Software and Services Index lost roughly $1 trillion in value over the following week, a stretch traders nicknamed software-mageddon.

In July, OpenAI released a governance tool called Presence, built to give companies guardrails for running AI agents inside their own operations, and the selloff spread over two trading sessions: Workday fell nearly 10% and Salesforce dropped more than 7%. September's reaction to Astra is smaller in magnitude than either of those episodes, but the pattern is now familiar enough that Wall Street has a name for the trade before the ink on the press release is dry.

The Other Side of the Trade

Money leaving software stocks has to land somewhere, and this week it landed in chips. Qualcomm rose 5% to $177.60 the same day, after announcing a custom AI chip partnership with Amazon Web Services covering inference silicon and high-speed optical connectivity. The company hasn't disclosed committed order volume or a delivery date, but it's now targeting $15 billion in annual data-center AI chip revenue by fiscal 2029, triple its earlier fiscal 2027 goal of $5 billion. Broadcom, which lost roughly 5% of its own value on a guidance miss earlier this quarter despite reporting 221% growth in AI chip sales, picked up 3% on the same afternoon. Intel added nearly 10%, though its move traced back to a separate story entirely: a DigiTimes report that the company plans a third CPU price increase of 2026, backed by an analyst upgrade citing a persistent chip shortage.

Commentary circulating after Tuesday's close described the pattern in blunt terms: Astra had reignited the software disruption trade, and capital was rotating back into the chips and data centers that make AI agents possible, at the expense of the software companies now competing with them. That framing lines up with the day's numbers even if it oversimplifies which stock moved for which reason. Qualcomm's gain traced to its own Amazon deal, and Intel's traced to a pricing report with nothing to do with AI agents at all. Both still landed on the winning side of a trade that's become easy to spot once you know to look for it.

What to Watch Next

Watch how Salesforce, ServiceNow, and Intuit talk about AI agents on their next earnings calls, and more specifically, whether they can show actual revenue or customer retention holding up against the threat rather than simply acknowledging it exists. Analysts have started drawing a line between software companies that hold deep customer data, permissions, and workflow history, which may be harder for an outside AI agent to casually replace, and narrower point solutions built around a single repetitive task, which look more exposed. Also worth tracking is whether Astra's rollout, still limited to select organizations for its more advanced capabilities, broadens enough in the coming months to turn this from a sentiment-driven selloff into something that shows up in actual renewal rates.

The Pulse24 Take

Three selloffs tied to three separate AI launches in eight months adds up to a trend that's getting harder to wave away as overreaction. That doesn't mean Salesforce, ServiceNow, and Intuit are doomed. Enterprise software has absorbed plenty of predicted disruptions before by building the new technology into its own product, and all three companies are already selling their own AI agent features rather than only defending against everyone else's. What's changed is the burden of proof. Every quarter without visible AI-driven customer losses now counts as evidence the market takes seriously, and every quarter with them gets read as confirmation of the thesis rather than noise.

The chip side of this trade is simpler to underwrite than the software side. Whoever ends up winning the agent war still needs the silicon to run it, which is part of why Qualcomm, Broadcom, and Intel keep finding buyers on days when software sells off, for reasons that in Intel's case had nothing to do with agents at all. Watch the software companies' own AI product revenue disclosures over the next two quarters. That number, not the next model launch, will settle whether this trade reflects a real structural shift or just a very persistent case of headline risk.