Wednesday, November 18, 2026

Software Repricing: AI Agents and Anthropic IPO Hit Seat-Based SaaS

CRMNOWMSFT
Why today:As of the Nov 17 close, Salesforce (CRM) fell 9.45% to $194.52, ranking No. 3 on the US heat list with discussion at 10.5x normal; ServiceNow fell 6.51% and Intuit 7.17%. X posts broadly say none of the three had company-specific bad news, attributing the selloff to fears that AI agents will replace seat-based subscription pricing and to money rotating out of software ahead of Anthropic's imminent IPO pricing.

The question: Salesforce fell 9.45% on Nov 17. Is that evidence that AI agents are already eroding per-seat pricing, or collateral damage from money leaving software ahead of Anthropic's IPO pricing? And what does a forward P/E of about 12x imply? Data as of Nov 17, 2026, after the US close (~17:06 ET).

Report date: 2026-11-18 | Data cutoff: US close on Nov 17, 2026; after-hours as of ~17:06 ET | Sources: yfinance (prices, valuations, share counts, analyst targets); Salesforce FY27 Q2 press release; coverage of the latest ServiceNow and Microsoft quarters; Bloomberg / Benzinga / mlq.ai / Motley Fool on the Nov 17 session and the Anthropic IPO; our Nov 11 and Nov 18 reports; Aya heat ranking; internal research corpus (Stratechery). Live X search was rate-limited that day, so market views come from media coverage and our previous daily report.

1. About $30 billion gone in a day: what happened

At the close on Tuesday, Nov 17, Salesforce (CRM) was down 9.45% at $194.52. Per mlq.ai, it was the worst performer in the S&P 500. Intuit (INTU) fell 7.17% and ServiceNow (NOW) fell 6.51%. A rough calculation from yfinance share counts puts the combined loss in market value at about $30 billion, about $16.7 billion of it at Salesforce. Per reports carried by Benzinga and TradingView, it was Salesforce's steepest one-day drop since May 30, 2024.

None of the three companies released bad news of its own. Media and X pinned the drop on two things. First, AI agents (AI programs that complete multi-step tasks on a user's behalf) could cut the number of software accounts companies need to buy. Second, Anthropic prices its IPO the next day (Nov 18) and could raise up to about $64.35 billion, so some investors were thought to be selling software to free up cash for it.

After-hours trading did not reverse the move. At press time, CRM was up 0.12% after hours, NOW up 0.45% and MSFT down 0.26%, so we keep the original title.

One piece of context first: software was not the only thing that fell. The S&P 500 closed down 1.03% and the Dow down 1.64%. Per @RonenTzirin, only 75 S&P 500 stocks closed higher, and the VIX rose to 25.9. The software ETF IGV fell 3.29%. Net of the market, CRM underperformed the S&P 500 by about 8.4 percentage points that day. That excess drop is what needs explaining.

Cumulative move since 10/28: NOW broke first, CRM caught down later, MSFT held u

Look at the shape of each line. ServiceNow's collapse came on Oct 29, its own earnings day. Salesforce drifted lower and then caught down about 8 points in a single session on Nov 17. Microsoft is down just 4.66% over the same period. The market is not dumping all software at once; it is sorting stocks by business model.

2. Why per-seat pricing is exposed to agents

Start with the business model. SaaS (software as a service) companies like Salesforce and ServiceNow mostly charge per seat: a customer buys one account for each employee who logs in and pays per user per month. Revenue is stable and predictable. The built-in assumption is that the more people a customer employs, the more the vendor collects.

Agents challenge exactly that assumption. Say a company has 100 support reps working tickets in Salesforce Service Cloud. If AI agents can handle 70% of the work, the customer may renew only 30 seats. The software is doing more work, yet the vendor collects less. Stratechery's interview with Asana founder Dustin Moskovitz (2025-10-20) put the conflict bluntly: if AI really raises productivity, per-seat pricing works against the value the product creates. He also admitted the shift is slow. Asana's new usage-based revenue started at single-digit millions of dollars against a business of about $700 million.

Three metrics come up repeatedly below:

  • ARR (annual recurring revenue): current monthly recurring revenue times 12, a measure of revenue "speed".
  • cRPO (current remaining performance obligations): contracted revenue to be recognized over the next 12 months, the leading indicator investors in subscription companies watch most closely.
  • NRR (net revenue retention): what the same customers pay this year divided by what they paid last year. Above 100% means existing customers are buying more.

Stratechery (2026-02-06) offers a useful frame. Software moats have not vanished: customer relationships, workflow embedding, accumulated data and switching costs all remain. But AI compresses product differentiation, and capital markets often "painfully reprice valuations first", before any change in fundamentals shows up. Nov 17 looks like a demonstration of that frame.

3. Core question 1: Are agents eroding seats, and can in-house agents fill the gap?

Our view: so far, erosion has not shown up in revenue or renewal rates. It has shown up in slowing growth and lower gross margins. In-house agent revenue is growing very fast but is still under a tenth of total revenue, which is not enough to conclude that seat shrinkage can be offset.

Evidence 1: Microsoft's seats are still growing, but the growth engine has changed. In Microsoft's FY27 Q1 (ended September), paid Microsoft 365 commercial seats rose 5% year over year, while Microsoft 365 commercial cloud revenue rose 17%. So growth comes mainly from charging more per seat: premium tiers like Copilot, E5 and E7 are lifting the price. A rough estimate puts revenue per seat up about 11% (1.17÷1.05−1; our estimate, not a company figure). In the same quarter, customers consuming Copilot credits on usage-based billing grew more than 75% from the prior quarter. For the vendor with the largest customer base, then, seats are not shrinking. Price is simply carrying more of the "seats × price" equation, and a usage-based revenue line is starting to grow alongside it.

Evidence 2: Salesforce's contracted revenue is still growing, but organic growth is modest. In FY27 Q2 (ended Jul 31, 2026), revenue was $11.3 billion, up 11%. Subscription and support revenue was $10.8 billion, up 12%, and cRPO was $33.5 billion, up 14%. But the company's guidance says the Informatica acquisition contributes "slightly above 3–4 points", which puts organic revenue growth at roughly 7%–8% (our estimate). On AI, Agentforce ARR passed $1.5 billion, up more than 240%, and combined Agentforce plus Data 360 ARR is about $3.9 billion. We found no seat count or NRR disclosure from Salesforce, so we cannot directly check whether seats are shrinking. That is the biggest blind spot in this analysis.

One mechanism deserves attention. Salesforce's Flex Agreement lets customers convert user licenses into Flex Credits (usage-based points) and back again. Its AELA (Agentic Enterprise License Agreement) offers fixed price, fixed term and unlimited usage. In other words, Salesforce has itself opened a door for customers to swap seat revenue for usage revenue. Whether total contract value grows or shrinks after the swap will show up in cRPO in the next report.

Evidence 3: ServiceNow's renewals are strong, but margins and guidance are weakening. Q3 subscription revenue was $3.988 billion, up 21%. The renewal rate was 98%. AI net new ACV grew 65%, and the full-year AI ACV target has been raised twice, from $1 billion at the start of the year to $1.7 billion. The trouble is in two places:

  • Q4 cRPO growth guidance was only 18.5%, below the roughly 19.8% the market expected.
  • Per indmoney, subscription gross margin fell from about 83% a year ago to about 80%. The most plausible reason is that customers are running more AI on the platform, and AI inference costs real compute money.

CFO Gina Mastantuono called the AI-disruption narrative "false" (per TIKR, 2026-11-06), but the stock did not buy it.

Scale of in-house AI products: still under a tenth of revenue

Look at the proportions: Agentforce alone is about 3.4% of revenue, and only about 8.9% even with data products included.

Putting the three pieces of evidence together:

  • Revenue and renewals show no erosion yet. Microsoft's seats are growing, ServiceNow's renewal rate is 98%, and Salesforce's cRPO is up 14%.
  • The pressure is already visible in the quality of growth. Organic growth is slowing, cRPO growth guidance was cut, and AI inference costs are squeezing gross margins.
  • In-house agents are growing fast but are too small. Current data cannot tell us whether they will fill the hole left by shrinking seats three years from now.

The counter-evidence must be stated too. The strongest case that erosion will accelerate comes from reviews of Anthropic's prospectus (see our Nov 11 report). Net revenue retention excluding consumer is about 900%, and more than 1,900 customers spend over $1 million a year. If enterprise AI budgets grow roughly 9x in a year, that money has to come from somewhere. On May 20, 2026, Intuit announced it would cut about 17% of its workforce (about 3,000 people) in an AI-driven restructuring, after signing multi-year deals with OpenAI and Anthropic. If software vendors are using AI to cut their own headcount, there is no reason their customers won't do the same.

The case that erosion is overstated has weaker evidence. @saastr cited a Stripe index showing non-AI B2B revenue still up more than 30% year over year, but the post drew almost no engagement and the sample definition is unclear.

4. Core question 2: How much money is the Anthropic IPO pulling, and for how long?

Our view: the funding hit is real but one-off, and most of it probably lands within a few sessions around pricing. The more durable effect is that a listed Anthropic becomes a valuation anchor investors can compare against at any time.

Start with size. Anthropic is selling 825 million shares at $70–78. At the top of the range that raises about $64.35 billion, roughly 40% of Salesforce's current market cap. There is also a $26 billion private placement at the IPO price, but the buyers are seven strategic investors such as Google, Nvidia and Microsoft. That money does not come from public-market funds, so it should not count as money pulled out of software. Bloomberg (Nov 12) reported the deal about 6x oversubscribed; another report says about 7x, which we could not verify. Underwriters are reportedly weighing pricing at or above the top of the range.

Then the causal evidence, which needs sorting by strength. Two kinds of evidence support the rotation story:

  • Timing. On Nov 10, the day Anthropic set its range, Microsoft fell 2.2% and ServiceNow 3.07% (our Nov 11 report). Nov 17 was the eve of pricing.
  • Positioning surveys. Per mlq.ai, Bank of America's fund manager survey shows managers cutting tech holdings at the fastest pace since 2006.

But this is correlation, not flow data. We have no data on any actual trade of the "sell software, buy ANT" kind.

Several confounders also apply:

  • A broad sell-off. Nov 17 was a market-wide risk-off day ahead of Nvidia's earnings.
  • Other negatives. Bloomberg cited a string of technology outages hitting government agencies, airlines and banks as another weight on software. The evidence for this explanation is weaker.
  • Uneven declines. Since Oct 28, Intuit is down only 2.25% while ServiceNow is down 38.12%. Pure liquidity draining should not produce such a split.

The roughly $30 billion in lost market value equals about 47% of the maximum IPO raise. Note, though, that market value is set by marginal trades: losing $30 billion of value does not mean $30 billion of money flowed out, and the two cannot be matched directly.

How long will it last? Pricing is Nov 18 and the first trade is Nov 19. If the oversubscription reports are right, investors who don't get full allocations may keep buying ANT in the secondary market, and that money won't flow back to software soon. But the offering itself is a one-time supply event. Once allocation is done, the reason to "raise cash for the IPO" disappears. Our inference: the liquidity pressure lasts days to weeks; the comparison effect lasts much longer.

Revenue scale: Anthropic's run-rate is already ~1.8x Salesforce's annual revenue

Look at the ranking. A five-year-old model company already has a higher revenue run-rate than Salesforce. Once it is listed, every quarter's revenue and retention figures will be read side by side with SaaS seat data.

5. What the market is betting on: the "terminal value" in a 12x multiple

Two terms first. The P/E ratio is share price divided by earnings per share. TTM uses actual profits over the past 12 months; forward uses analysts' forecast for the next 12 months. Terminal value is the part of a valuation model that captures what a company is worth many years out, and for mature software companies it is usually most of the value. "Terminal-value risk" means the market has started to doubt whether the business will still exist in ten years.

The valuation comparison (yfinance, as of the Nov 17 close):

CompanyFwd P/ETTM P/EP/S (TTM)Latest qtr rev growthvs 52-wk high
Salesforce12.1x17.8x3.6x10.8%−27.7%
Intuit11.9x19.5x4.0x13.7%−53.0%
ServiceNow19.3x66.8x6.2x20.5%−46.9%
Microsoft20.9x26.4x10.9x18.8%−9.5%

Forward P/E: CRM and INTU squeezed to ~12x, little more than half of Microsoft's

Look at the two ends: Salesforce, growing about 11%, and Microsoft, growing about 19%, are almost 2x apart on forward P/E.

1. Historically, CRM is in a very low valuation range. CRM's TTM P/E is now 17.8x. Per VCP Scanner, its five-year TTM P/E low was about 20.5x, so it now sits below that floor. A separate market commentary (rallies.ai, publication date unverified) said CRM's forward P/E at around 14x was already its lowest on record, including the financial crisis; today's 12.1x is lower still. We could not get a full forward P/E history, so "historic low" rests on several third-party sources pointing the same way, not on a precise percentile.

There is also a quality caveat. Per Zacks, part of Salesforce's Q2 EPS beat came from investment gains, so the TTM P/E may look lower than it really is because of non-recurring gains.

2. What does 12x imply? A rough back-of-envelope (an assumption, not a forecast): assume investors want a 9% annual return and all profits are paid out as dividends or buybacks. Working backward from the perpetual-growth formula, CRM's forward earnings yield is 8.2% (1÷12.15), which implies perpetual growth of only about 0.8%. The same method gives:

  • ServiceNow: earnings yield about 5.2%, implied growth about 3.8%;
  • Microsoft: earnings yield about 4.8%, implied growth about 4.2%.

In other words, CRM's price roughly assumes its profits never grow again. Yet the company guides to 11%–12% revenue growth (including acquisitions), and Agentforce is growing 240%.

3. Is that expectation reasonable? Both sides have a case.

  • Too much pessimism already priced in: even at 7%–8% organic growth, there is a real cushion over an implied zero-growth case. The average analyst target of $287.08 is 47.6% above the current price (53 analysts, yfinance).
  • Not cheap enough yet: if seats really start shrinking, profits won't be flat, they'll fall. And sell-side targets usually lag the stock, so they are of limited use.

Our read: a 12x forward P/E shows the market is already applying a "terminal-value discount" to Salesforce. Which way the stock goes next depends on whether the next report shows cRPO holding up as seats shift to usage. ServiceNow's forward P/E is still above CRM's because it grows about twice as fast. But the stock's path since Oct 29 shows that once growth guidance is cut, a premium multiple shrinks faster.

6. What to watch next

Date (ET)EventWhat it tells us
Nov 18 (Wed)Anthropic IPO pricing; FOMC minutes (14:00)Pricing at or above the top of the range shows investors will pay up for "AI-native", which reinforces near-term pressure on software; weak pricing would undercut the cash-raising story
Nov 19 (Thu)ANT first trade; Intuit FQ1 results (after close)Day one sets the valuation anchor for model companies; Intuit consensus is EPS of $2.47 and revenue of $4.31B, so watch QuickBooks agent monetization and guidance
Dec 2 (Wed, after close)Salesforce FY27 Q3Three numbers: whether cRPO meets the ~14% guide; whether Agentforce ARR keeps doubling; whether management discloses the net effect of swapping seats for Flex Credits
Late Jan 2027 (yfinance estimate)ServiceNow Q4; Microsoft FY27 Q2For ServiceNow, whether cRPO holds 18.5% and subscription gross margin stops falling; for Microsoft, the gap between paid seat growth and revenue per seat

The single most important signal: after a seat-based vendor actively pushes customers from seats to usage, does total contracted revenue (cRPO) grow or shrink? If cRPO growth holds or improves, "erosion" is a manageable structural shift. If cRPO growth keeps stepping down and companies start avoiding seat disclosures, the terminal-value discount is justified.

7. Conclusion

The Nov 17 software sell-off stacked three causes: market-wide risk-off before Nvidia's earnings, money shifting ahead of the roughly $64.35 billion Anthropic IPO pricing, and a repricing of the long-term outlook for seat-based software. The first two are short-term; the third is what really drives the stocks.

Based on reported results, agent erosion of seats has not yet shown up in revenue or renewals: Salesforce's cRPO is up 14% and ServiceNow's renewal rate is 98%. The pressure shows up in slowing organic growth, cuts to cRPO growth guidance and AI inference costs eating into gross margins. Agentforce is growing fast, but its roughly $1.5 billion of ARR is only about 3.4% of Salesforce's revenue, not enough to show it can offset shrinking seats. CRM's forward P/E of about 12.1x is close to pricing in zero growth. What comes next depends on whether the Dec 2 report shows that the seat-to-usage shift has not shrunk total contract value.


Appendix: Data and Sources

A. Key financials and valuation (USD 100m unless noted)

MetricCRMNOWINTUMSFTSource / date
Nov 17 close ($)194.5292.90320.79506.24yfinance
Nov 17 change−9.45%−6.51%−7.17%−2.27%yfinance
Market cap1,600.9964.2857.337,594yfinance
TTM revenue439.4154.3214.53,464.4yfinance
Fwd / TTM P/E12.1 / 17.819.3 / 66.811.9 / 19.520.9 / 26.4yfinance
Mean analyst target ($, # analysts)287.08 (53)155.82 (46)407.89 (31)636.12 (54)yfinance
Next earningsDec 2, 2026 after closelate Jan 2027 (est.)Nov 19, 2026 after closelate Jan 2027 (est.)yfinance
Latest AI metricAgentforce ARR >15; +Data 360 ≈392026 AI ACV target 17—Copilot usage-billing customers +75% q/qCompany releases / earnings coverage
Contracts / renewalscRPO 335 (+14%); RPO 663 (+11%)cRPO 136.5 (+21%); renewal 98%; Q4 cRPO guide 18.5%—M365 commercial seats +5%; M365 commercial cloud revenue +17%Same

B. Formulas and inputs

MetricFormulaResult
CRM 10/28→11/17194.52 / 242.11 − 1−19.66%
NOW 10/28→11/1792.90 / 150.13 − 1−38.12%
MSFT 10/28→11/17506.24 / 531.01 − 1−4.66%
INTU 10/28→11/17320.79 / 328.19 − 1−2.25%
IGV 10/28→11/17; Nov 1795.74 / 111.91 − 1; 95.74 / 99.00 − 1−14.45%; −3.29%
CRM excess drop−9.45% − (−1.03%)−8.42 pts
One-day value lost$ drop × shares: CRM 20.31×823M; NOW 6.47×1,038M; INTU 24.77×267M$16.72B + $6.72B + $6.62B = $30.06B
Value lost / max IPO raise300.6 / 643.546.7%
Max IPO raise / CRM market cap643.5 / 1,600.940.2%
Agentforce ARR share15 / 439.383.4%
Agentforce+Data 360 share39 / 439.388.9%
NOW AI ACV target share17 / 154.3111.0%
Anthropic run-rate / CRM TTM revenue810 / 439.381.84x
MSFT implied revenue-per-seat growth1.17 / 1.05 − 1~11.4%
CRM organic growth (estimate)11%–12% − "slightly above 3–4 pts"~7%–8%
Earnings yield1 / fwd P/E: CRM 1/12.15; NOW 1/19.34; MSFT 1/20.928.2%; 5.2%; 4.8%
Implied perpetual growth (assume r=9%, full payout)g = 9% − earnings yieldCRM 0.8%; NOW 3.8%; MSFT 4.2%
CRM upside to target287.08 / 194.52 − 1+47.6%
CRM vs 52-wk high194.52 / 269.11 − 1−27.7%

C. Aya heat data (Nov 17, 2026 close)

TickerRankHeatDiscussionBull/Bear (self-tagged)Drivers
CRM36810.5x normal53% / 47% (15 posts)Discussion spike, sharp drop
NOW10592.9x normal88% / 12% (26 posts)Discussion spike, sharp drop
MSFT13571x normal65% / 35% (17 posts)—

Note: the heat ranking shows NOW at −6.7%; the text uses the −6.51% from the quote snapshot throughout. CRM's near 50/50 split shows deep retail disagreement; NOW's 88% bullish reading looks like retail buying the dip.

D. Timeline

  • May 2025: Salesforce introduces Flex Credits pricing (per Redress Compliance)
  • May 20, 2026: Intuit announces a ~17% workforce cut (~3,000 people) in an AI restructuring
  • Aug 26, 2026: Salesforce FY27 Q2; Agentforce ARR passes $1.5B
  • Oct 28, 2026: Microsoft and ServiceNow report; NOW −26.95% on Oct 29
  • Nov 10, 2026: Anthropic sets its $70–78 range; MSFT −2.2%, NOW −3.07%
  • Nov 12, 2026: Bloomberg reports the IPO about 6x oversubscribed
  • Nov 17, 2026: CRM −9.45%, INTU −7.17%, NOW −6.51%; Nvidia reports after the close
  • Nov 18 / 19: Anthropic pricing / ANT first trade; Intuit reports on Nov 19
  • Dec 2, 2026: Salesforce FY27 Q3

E. Market views

  • RBC Capital Markets analyst Rishi Jaluria: "the AI companies are going to try to destroy Salesforce entirely" (via mlq.ai, Nov 17, 2026)
  • ServiceNow CFO Gina Mastantuono called the AI-disruption narrative "false" (via TIKR, Nov 6, 2026)
  • @techprosv: "No company news explains the drop."; @aslampalia29: "AI agents eating the subscription model, with Anthropic pricing its IPO tomorrow" (quoted in our Nov 18 daily)
  • @WealthSignal739: "what happens when AI agents do the work the seats used to do?" (same)
  • @saastr: Stripe index shows non-AI B2B revenue up 30%+ YoY (same; low engagement)
  • Stratechery (Oct 20, 2025; Feb 6, 2026): per-seat pricing conflicts with AI productivity gains; valuations often reprice ahead of fundamentals (internal corpus)

F. Sources

For information and research discussion only. Not investment advice.

<!-- AYA_PODCAST -->

🎙️ Podcast script (Chinese, ~20 min, 泓观阳略): read it here

AI-generated by Aya Research from public data and disclosures, for information and research purposes only. Not investment advice.