Four months ago the market decided B2B software was dead.
In a 48-hour stretch in February, roughly $285 billion in software market cap evaporated. The trigger was the fear that AI agents would make per-seat licensing obsolete, that the entire model of charging companies for seats their employees would no longer need was about to collapse. The market gave it a name: the SaaSpocalypse. The IGV software index fell more than a third from its September highs, hit a 52-week low on April 10, and at the bottom around 75% of software stocks were screening as technically oversold.
But since then, the broad software index has ripped back more than 40% off the April low. But not everyone has ripped back. Three of our most iconic B2B leaders — Salesforce, HubSpot, Adobe — kept sliding to fresh 52-week lows. The recovery left them behind. One of them is arguably the cheapest large-cap software stock in a decade. One is the fastest grower of the three and quietly cheap on revenue. And one is carrying the most direct exposure to the exact fear that started the panic.
Are these Big 3 … oversold? Is 3x ARR a bargain for B2B leaders with real AI revenue?
The fear that drove the selloff
Salesforce, HubSpot, and Adobe built franchises by charging per seat. More employees, more licenses, more revenue. If AI agents do the work of those employees, the logic goes, headcount stops growing, seats stop growing, and the revenue model caps out or shrinks.
It is a clean thesis. It is also testable. The way you test it is by watching whether these companies can convert AI from a threat into a line item that grows faster than the core. So far, the results are mixed. “AI revenue” is up and real, but it hasn’t fueled material net growth for these Big 3. Yet.
Salesforce: the cheapest it has looked in years
Salesforce entered earnings as the worst-performing stock in the Dow this year, down about 32% before the print. The market had effectively concluded that Agentforce was a science project and that the core CRM seat business was the most exposed in enterprise software.
The Q1 FY2027 numbers told a different story. Revenue of $11.13 billion grew 13% (12% in constant currency). Non-GAAP EPS jumped 50% to $3.88. And the non-GAAP operating margin hit 34.8%, a record and a 250-basis-point expansion in a single year. For a company that spent a decade absorbing criticism over dilution and bloated spend, that margin number reframes the entire profitability debate.
The headline that mattered most was Agentforce. ARR crossed $1.2 billion, up roughly 205% year over year. Combined with Data 360, the agentic and data layer is now around $3.4 billion in ARR, growing north of 200%. That makes Agentforce the fastest-scaling AI product line any enterprise software company has disclosed this year.
Two details inside that number do real work. First, more than half of Agentforce and Data 360 bookings came from existing customers expanding their deployments. That is the textbook signal of product-market fit, not pilot churn. Second, Agentforce is not priced per seat. It is priced per Agentic Work Unit, a consumption charge tied to workflow output. If that holds at scale, Salesforce revenue per customer is no longer capped by headcount. It is capped by workflow volume, which can grow even as headcount shrinks. That is the same architectural shift that turned consumption-based data and infrastructure businesses into the most valuable franchises of the last decade.
The bear case did not disappear. Full-year guidance of roughly $45.9 to $46.2 billion implies about 11% growth, and once you strip out the Informatica acquisition, organic growth is closer to 6 to 7%. Commerce Cloud and Tableau were both flagged as soft. The core is maturing, and every dollar of net-new Agentforce ARR carries some risk of substituting for a Sales Cloud or Service Cloud seat rather than adding to it.
But look at the price. Salesforce gave the entire post-earnings pop back and then kept falling. The stock now sits around $152, down roughly 43% on the year and parked right on its 52-week low, which works out to about 11x forward non-GAAP earnings against a software industry near 27x. What dragged it back to the lows was not the quarter. It was a fresh round of layoffs in early June that touched Agentforce, MuleSoft, and Marketing Cloud, the acquisition of usage-based billing platform m3ter, and a broad AI-driven software selloff. A Rule-of-40 company (about 47.8 here) trading at roughly 11x forward, retiring 10% of its shares through a $25 billion accelerated buyback, carrying a PEG near 0.47 and a 12% free cash flow yield, with the largest AI revenue line in the category, is not the setup the “structurally impaired” thesis describes. The market is pricing the cannibalization risk as if it has already happened. The income statement says the opposite. At least one sell-side shop agrees: Monness upgraded the stock to Buy this week, calling the valuation compelling with shares 58% off their all-time high.
Why It’s Oversold
- Agentforce ARR at $1.2 billion growing 205% is the largest and fastest AI revenue line in enterprise software, and it is priced per workflow rather than per seat, so it is not capped by customer headcount.
- A record 34.8% operating margin plus a $25 billion buyback retiring 10% of shares pushes earnings per share higher even at 11% revenue growth.
- At roughly 11x forward earnings, a 0.47 PEG, and a 12% free cash flow yield while sitting at 52-week lows, it is priced for a decline the income statement is not showing.
Why It Isn’t
- Strip out the Informatica acquisition and full-year organic growth is closer to 6 to 7%, so the core is maturing.
- Commerce Cloud and Tableau are soft, and per-seat Sales and Service Cloud licenses are the most exposed in software to agent substitution.
- The June layoffs across Agentforce and MuleSoft teams cut against the “AI is pure upside” story and signal cost management, not just demand.
HubSpot: fastest grower, hit the hardest
HubSpot is the most violent chart of the three. The stock is down about 56% this year, trading around $176 just off a fresh 52-week low, against a 52-week high above $600. On price alone it looks like the market gave up on it entirely.
The operating numbers do not match that abandonment. Q1 revenue grew 23% as reported (18% in constant currency) to $881 million. ARR reached $3.45 billion. The customer base crossed nearly 300,000, up 16%, with 10,800 net new adds in a single quarter. And the part that should get more attention: non-GAAP operating margin expanded 4 points to 17.8%, with the company raising full-year margin guidance to 21%, hitting its 2027 target a full year early.
The AI monetization signal is early but pointed in the right direction. Active core seat users grew 90% year over year. Credits consumed grew 67% quarter over quarter. The Spring Spotlight launch shipped Customer Agent, Prospecting Agent, and Data Agent, and HubSpot is layering in outcome-based pricing for AI credits on top of core seats. Management’s framing is that AI adds two monetization levers, seats and credits, rather than replacing the seat model outright. Larger deals ($60K+ ARR) grew up to 64%, which means the upmarket motion is real, not aspirational.
HubSpot is the most exposed to the original fear because it sells primarily to SMB and mid-market go-to-market teams. The per-seat-to-agent transition hits hardest where the seat is a salesperson or a service rep an agent can plausibly stand in for. That is the legitimate risk. The counter is that clean, unified customer data is the prerequisite for any agent to work, and HubSpot’s whole pitch is being that data and AI foundation for companies that do not want to stitch together a Salesforce-plus-Informatica-plus-everything-else stack.
On valuation, HubSpot is the interesting case. After the drop it trades around 2.5x ARR and roughly mid-teens forward non-GAAP earnings, for a business still compounding revenue above 20% with accelerating margins. That is not the multiple of a company the market believes is broken. It is the multiple of a company the market stopped believing in temporarily. The faster grower of the three got the steepest price cut, which is exactly the kind of dislocation that tends to correct when the next two quarters confirm the AI levers are converting.
Why It’s Oversold
- The fastest grower of the three at 23%, now trading around 2.5x ARR and mid-teens forward earnings after a 56% drop, a multiple that does not fit a 20%-plus compounder.
- It hit its 2027 operating margin target of 21% a full year early, proving it can expand profitability and growth at the same time.
- AI monetization is already in the numbers: core seat users up 90%, credits consumed up 67% quarter over quarter, and $60K-plus deals up as much as 64%.
Why It Isn’t
- It sells to SMB and mid-market go-to-market teams, the exact seats an agent can most plausibly replace, which makes it the most directly exposed to the original fear.
- Management flagged short-term disruption from sales enablement and packaging changes, so the next quarter or two could stay choppy.
- AI credit revenue is still early and small next to the core seat business, so the second lever is more promise than proven engine.
Adobe: the cheapest, with the messiest narrative
Adobe is the value-trap-or-bargain debate in its purest form. The stock is down about 49% over the past year and trades near $195, at roughly 11x trailing earnings and high single digits to low double digits on a forward basis. Adobe has not traded this cheap on earnings in well over a decade. Its ten-year average P/E is north of 40.
The Q2 print was a beat-and-raise. Record revenue of $6.62 billion, up 13%. Non-GAAP EPS of $5.96, ahead of estimates. Total ending ARR of $27.1 billion, up 12.5%. The company raised full-year revenue and EPS targets.
The AI line is now material. AI-first ARR crossed $500 million and tripled year over year. Firefly ARR is approaching $300 million, growing about 50% quarter over quarter. GenStudio ARR grew over 25%. Adobe also has a moat the AI-native generators cannot easily copy: Firefly was trained on licensed and public domain content, which lets Adobe offer paying customers contractual IP indemnification for commercial use. For any regulated enterprise, that legal backstop is a direct reason to pay Adobe rather than a free image generator. It helps explain why the business-professional and consumer subscription line is growing faster than the core creative base.
The stock fell 5.5% on the beat-and-raise for two reasons, and neither is about the quarter. First, the CFO announced he was leaving in four days, to Marvell, with no advance notice. That came on top of a CEO transition already underway, leaving Adobe heading into the back half of its fiscal year with two of its three most senior roles filled on an interim basis. Second, Adobe is deliberately suppressing near-term ARR by going aggressive on a Creative Cloud freemium model. Free monthly actives jumped from over 50 million to over 90 million, and traffic to adobe.com rose 40 to 50%, but management has not disclosed the free-to-paid conversion rate. Until it does, investors are being asked to accept lower ARR today on faith that the conversion math works later.
That is the real Adobe risk, and it is more nuanced than “AI eats Photoshop.” The competitive pressure from Canva, Figma, and AI-native design tools is real. But the more immediate question is execution and trust during a leadership vacuum, plus a freemium bet with an unverified payoff. The stock is priced as if AI permanently lowers the value of the creative bundle. The earnings say AI is currently adding half a billion in recurring revenue and growing it 3x. Both cannot stay true for long.
Why It’s Oversold
- The cheapest of the three on earnings at roughly 8 to 11x forward, against a ten-year average above 40, after falling about 49% over the past year.
- AI-first ARR tripled past $500 million and Firefly is nearing $300 million, so AI is now a material, fast-growing line rather than a cost center.
- Firefly’s licensed training data gives Adobe contractual IP indemnification that Midjourney, OpenAI, and Stability cannot match, a real moat with enterprise buyers.
Why It Isn’t
- No permanent CEO and no permanent CFO heading into the back half of the year, a governance vacuum at the worst possible moment.
- The freemium push is deliberately suppressing near-term ARR, and management will not disclose the free-to-paid conversion rate that would justify the trade.
- Canva, Figma, and AI-native generators are real competition for the creative bundle, and the stock is pricing that pressure as permanent.
And Who The Market Is Rewarding Instead
The selloff was never a vote against all software. The same year the application names cratered, a set of infrastructure companies ran the opposite way, and four of the biggest winners tell you what the market is buying.
Datadog is up about 64% on the year, near a 52-week high, on roughly $4 billion in ARR. The thesis is almost too clean. AI makes systems more complex, and more complexity means more to monitor. Every model, agent, and pipeline a company puts into production is one more thing Datadog gets paid to watch.
CrowdStrike is up about 46%, on ARR around $5.2 billion with record net new ARR last quarter. It spent the year repositioning Falcon as the security control plane for the agentic enterprise and shipped Continuous Identity for AI Agents in June. Every agent an enterprise switches on is a new thing that has to be secured, and CrowdStrike is selling the layer that does it.
Okta is up about 36%, working the same shift from the identity side. Its newer products, the bucket its AI work sits in, already drive around 30% of new bookings, and deals that include them carry roughly 40% higher contract value. The logic is plain. Every AI agent needs an identity and a set of permissions, and the count of non-human identities inside enterprises is climbing fast. IDC expects the identity security market to roughly double, from about $29 billion to $56 billion by 2029, largely on the back of it.
Twilio is up about 31%, and it may be the most literal version of the trade. AI agents need a way to reach customers, by voice and by text, and Twilio is the pipe they run on. Voice revenue grew 20% last quarter, its fastest in nineteen quarters, and software add-ons like branded calling and conversational intelligence more than doubled year over year. AI-native players like Sierra and Bland.ai are building their agent platforms directly on Twilio, which raised full-year guidance and now bills itself as the communications layer for the age of AI.
Look at what those four share, and then at what they do not. None of them sells a seat to a human an agent might replace. They sell the layer that watches, secures, and authenticates the agents themselves.
The more agents a company deploys, the more these businesses make. The market is not anti-software. It is selling the applications most exposed to agent substitution and buying the infrastructure that grows every time an agent is switched on. Salesforce saw the direction early, which is why Agentforce is metered on work units and why it bought m3ter for usage-based billing. On the multiple the market is using to price these winners, Salesforce trades at about 3.1x its $40 billion in ARR, against Datadog near 20x. The market is paying for the companies furthest along, and discounting the ones it has decided are behind.
AI Revenue is Real in the Big 3. All-In Growth Has to Come From It Next.
For anyone building in B2B, the more useful signal is not the stock prices. It is what these three quarters reveal about how AI revenue shows up.
It does not show up as the core collapsing. It shows up as a second monetization layer, consumption units at Salesforce, credits and seats at HubSpot, Firefly and GenStudio ARR at Adobe, growing 50% to 200% on top of a core that is still expanding. The market modeled substitution. The companies are reporting addition, at least so far.
That is the real test the next two quarters will settle. If Agentforce ARR keeps compounding at triple digits, if HubSpot’s credit consumption keeps accelerating, if Adobe’s AI-first ARR holds its 3x pace, then the SaaSpocalypse was a repricing of fear, not fundamentals, and these three were oversold. If the AI lines stall while the core decelerates, the discounts were earned.
Right now the evidence points one direction. The companies the market left for dead are posting the fastest-growing product lines they have shipped in years. Being down 38% to 56% while growing revenue and printing record margins is not what structural impairment looks like. It is what fear looks like when it overshoots.
