For most of the past year, the trade was simple: long chips, short software. It worked almost perfectly.

The IGV software ETF is down roughly 14% year-to-date. Salesforce has fallen about 38%. ServiceNow shed more than a third of its value. The market had a name for it: the SaaSpocalypse. And it was not random panic. The logic was real. If AI agents can execute the same workflows these platforms were built to automate, why pay $150 per seat per month?

Here is what is interesting, though. The thesis was directionally right but analytically lazy. And now a very specific signal just arrived that says the pendulum may have swung too far.

On July 1st, Guggenheim analyst John DiFucci upgraded both Salesforce and ServiceNow to Buy, titling his note “Armageddon Called Off.” He was explicit that this was not an AI endorsement. He still believes AI poses real long-term risks to these businesses. His argument was simpler and harder to dismiss: the worst-case scenario is now priced in.

ServiceNow is currently trading roughly 50% below its 52-week high of $211.48. Guggenheim set a $125 price target on NOW, valuing it at 7.5x EV to next-twelve-months recurring revenue. For Salesforce, DiFucci cited a valuation of 3.7x recurring revenue and 11x EV/NTM consensus free cash flow as an attractive entry point for a business that is comfortably profitable and growing at double digits. The upgrade carries weight precisely because it comes from a skeptic. He moved ServiceNow from Sell to Neutral in December 2025, then watched it fall another 35%. His shift to Buy is a valuation call, not a change of heart on AI.

Slight tangent, but it matters: the gap between software and semiconductor price action is now the widest on record. Every stock in the S&P 500 software universe is currently trading below its 200-day moving average, while nearly nine out of ten semiconductor stocks sit above theirs. That is not a normal divergence. A rotation from chips into software does not need much of a catalyst. It just needs the fear to stop accelerating.

The underlying business data is harder to dismiss than the stock prices suggest. ServiceNow guided for full-year 2026 subscription revenue of roughly $15.75 billion, implying approximately 20% constant-currency growth. In Q1, subscription revenue rose 22% year over year to $3.67 billion. Evercore ISI maintained its Outperform rating with a $150 price target ahead of Q2 earnings on July 22nd. On the Salesforce side, Agentforce ARR crossed $1.2 billion in Q1 FY2027, up 205% year over year, with combined Agentforce and Data 360 ARR reaching $3.4 billion. That is faster AI product scaling than any other enterprise SaaS company has reported this year. Salesforce also completed its $8 billion acquisition of Informatica in November 2025, deepening its data management infrastructure and laying the foundation for expanded agentic AI deployments across enterprise customers.

The business risk is real. AI agents are compressing the per-seat model. Net revenue retention is stalling across the sector. New AI-native competitors are funded and not dependent on operating cash flow. None of that has changed.

But here is what the market appears to be missing: gross retention across the software sector is still sitting around 90%. Enterprises have a decade of proprietary data living inside these platforms. Migrating off Salesforce is not a software decision. It is a digital-foundation rebuild. The moat is not impenetrable, but it is not thin either.

The highest-probability outcome is not a V-shaped recovery. It is a grind higher for the strongest platforms, the ones with genuine workflow lock-in, healthy cash flow, and the balance sheet to embed AI before competitors dislodge them. The sector has been pricing in permanent decline. The underlying businesses have not confirmed it yet.

That gap does not stay open forever.