Newsletter · · Ashutosh Agarwal
The SaaSpocalypse Debate Splits Software Into AI Winners and Losers - Weekly SaaS / Software Podcast Recap - Week of August 2, 2026
A synthesis of what investor, VC and operator podcasts said about SaaS and software for the week of July 26 to August 2, 2026, including the SaaSpocalypse framing, the coming reckoning over token-based pricing, and why ServiceNow, Intuit and Palantir dominated the bull and bear arguments.
Weekly SaaS / Software Podcast Recap
Week of July 26 to August 2, 2026: The SaaSpocalypse Debate Splits Software Into AI Winners and Losers
Top of mind this week
One word ran through almost every software conversation this week: the "SaaSpocalypse." That is the podcast world's nickname for the fear that artificial intelligence, especially chatbots like ChatGPT and Claude, and the new habit of "vibe coding" (building working software just by describing it in plain English), is about to eat the traditional software-as-a-service business. Steve Eisman built a whole episode around it. Founders quoted a viral post claiming "300 VC-backed startups just died" the day Anthropic added a new Claude feature. A VP at Braze described AI inside his own company in one word: "chaos."
But the more careful voices this week kept landing on the same nuance: it is not "all software dies." It is that the bad software, the stuff that raised prices for years without adding value, the simple horizontal tools you can now rebuild yourself, is in real trouble, while software with proprietary data, deep integrations, a system of record, or a human on the hook is not only surviving but is being called an AI winner. That split, loser versus winner inside the same sector, was the throughline of the week, and it showed up most concretely in the beaten-down names people are now arguing are cheap: ServiceNow, Intuit, Adobe, and Palantir.
Below: the dominant themes, the live debates (each framed as one side versus the other), and every company named with the bull and bear case attached.
1. Dominant themes
The "SaaSpocalypse" is the frame everyone is arguing inside. Steve Eisman devoted an episode to it on The Real Eisman Playbook, "The AI Revolution: Who Wins, Who Loses & the SaaSpocalypse" (July 27). His guest Gil framed the core idea plainly: companies now have to spend so much on AI that they are hunting their budgets for things to cut, and "there's a crowding out of unimportant software… products that didn't make their customers happy." The question for any software company, he said, has narrowed to just one thing: "Who's a good company and who's not a very good company?"
"Vibe coding" moved from buzzword to management headache. On [Un]Churned, "Vibe Coding Is a Bandaid on a Bandaid" (July 29), Saumyo Mukherjee, VP of Business Systems at customer-engagement platform Braze, described what happens when every employee can build their own software: "Somewhere at Braze, a sales rep is vibe coding their own automation. Three desks over, someone else is building the same thing, slightly different. Neither knows the other exists." He calls it "good chaos," but warns the result is often "a bandaid over another bandaid," a sprawl of overlapping, ungoverned tools. His fix is to consolidate and pick the highest-value projects: one AI build (using GPT-4o) automated the handoff from sales to services, a task that used to take "at least one and a half business days per rep" wrestling through 20 documents, spreadsheets and contracts. On Futureproof Founder, "Can AI Replace Your Accountant?" (July 28), founder Sai Dhanak captured the anxiety directly: "Every single day… there was a tweet from someone who said 300 VC-backed startups just died," referring to a new Claude release. He also admitted the disruption "turned out to be far more threatening, actually, than I thought it would be."
A pricing reckoning: seats versus usage versus tokens. This was one of the sharpest, most concrete discussions of the week. On AI Radicals, "Why Enterprise AI Is Entering Its ROI Era" (July 29), Mark Nelson, a Venture Partner at Madrona and the former CTO of Tableau, argued that charging customers "per token" (the unit of text an AI model processes) is heading for a fall, because a token "does not equal value." He drew on two real cautionary tales: Splunk, which charged by data storage until customers realized "the value I got was the last week's worth of data" while paying the same for five-year-old logs, "eventually that pricing model collapsed on itself"; and Tableau, where the team rejected usage-based pricing because charging per query would make users "coin-operated," doing less of the exploration that was the whole point. His conclusion: "I do think there's a reckoning coming… I want to be paying in some unit that eventually accrues to the value I get." Gamma's founder, interviewed on SaaStr 871, "$0 to $100M ARR Fast" (July 29), described living this in real time: Gamma launched with no way to charge at all, hacked together a "credit-based system" when users demanded to pay, and is still "going on a journey of how much should be seat-based versus consumption-based." His rule now: "Pricing and packaging is something you need to revisit constantly."
Capex is out; monetization is in. On The Morning Filter, "5 Stocks to Buy While They're Still Reasonably Priced" (July 27), Morningstar's Dave Sekera framed the shift bluntly: "It's really no longer just about who can spend the most CapEx the fastest, but increasingly it's becoming more and more about who's actually showing how they're monetizing it." He flagged that Alphabet's stock sold off after it raised capex guidance, that investors are now watching when depreciation expense "materially starts ramping up," and that "what we saw with Alphabet last week is they are actually going to be free cash flow negative this year, maybe even next year." That capex debate has a direct read-through to infrastructure software.
Open source and "model routing" as the next layer of the stack. Several shows converged on the idea that companies won't pick one AI model, they'll route between many. Matt Murphy of Menlo, on 20VC, "Leading Anthropic's First Ever Round" (July 27), described "Wave 1 of AI is like, let's just get it going. Wave 2 is like, let's get a lot more sophisticated about what we use and when and how," pointing to OpenRouter (a company he's on the board of), which puts "an intelligent layer that intercepts an API call… what's the best model for me" by price, reasoning, or latency. Eisman's guests made the same point from the investor side, noting Nvidia is now building its own free open-source model ("NemoTron") because "open source models use just as much compute as closed source models."
Security software is the sector being called an AI winner. On the Eisman episode, Dan argued cybersecurity budgets "are going to double the next two or three years" because every AI agent expands the attack surface: "If Steve Eisman has three agents… they have to protect three agents. It's just more surface area." That theme recurred in the ServiceNow discussions below.
2. Key debates
Is SaaS dead? "Crowding out the losers" versus "overblown and underblown at the same time." One camp (Eisman's guests) sees a genuine culling: unimportant software gets cut first, and Salesforce is named as a prime candidate. The more nuanced camp pushed back hard. Mark Nelson (Madrona): "This is the SaaSpocalypse, I can both say it's overblown and underblown all at once." His point: categories like CRM and ERP won't disappear, but "the rules of what it means to be a CRM vendor are going to change radically." On Swimming with Allocators, "AI, SaaS and the Next Private Markets Shakeout" (July 29), the allocator guest (Chris, of Axia) was explicit: "It's not legacy SaaS goes boom, that's way too simplistic." His framework: software with "proprietary data, an installed customer base, sensitive customer data, and the intersection of functional workflows with industry specifics" is highly embedded and fine; "more bland horizontal software that's simple workflow stuff… probably doesn't have a right to exist anymore."
Which vintage of software is actually at risk? Old versus new. The same allocator gave the cleanest numbers of the week: roughly 20 to 25% of all private-equity money over the last 8 or 9 years went into software, and the 2019 to 2021 "vintage" is the most exposed, bought at peak-zero-rate valuations, often leveraged against recurring revenue, and now facing resets (a company "expecting to be borrowing at 6% wound up borrowing at 11%") plus pre-ChatGPT business models. By contrast, "a lot of 2024 [software businesses] we see pretty strong EBITDA growth… because those are AI-native."
Who wins the AI agent economy, incumbents or startups? Mark Nelson's answer: "It's execution." Incumbents like Salesforce have huge customer bases, revenue, brand and data, but also "hundreds of thousands of customers who expect them to do what they did yesterday. It is the classic innovator's dilemma." Startups can "build whatever they want" but "have no customers and no money and no data." Sai Dhanak added a durability test: defensible software needs one of three things, "hardcore, difficult integrations" (his example: "you can't vibe code integrations with a thousand global payment infrastructure banking providers," like Stripe), "a ledger or system of record," or "a human liability backstop."
Does open source shrink the frontier AI labs' market? On 20VC, host Harry Stebbings pressed Matt Murphy: if open-source models can do "96% of enterprise workflows," doesn't that gut the paid model companies? Murphy's answer was a firm no on displacement: companies find that "if I use Anthropic, it actually increases my customer retention, I generate more revenue," so the future is multi-model, "50% this, 30% that, 20% this," not open-source-takes-all.
Can private equity still fix legacy software? "Blood from a stone" versus "can they pick the gems?" On 20VC's markets roundtable, "Google Cloud Grows 82% But The Market Tanks" (July 30), Jason Calacanis was scathing about the PE turnaround playbook for mature software: "We're five years into no net new customers and price increases… I don't think there's another five years of those knobs and dials left." His example: Marketo "raised our prices from $22,000 to $80,000 since 2020," so his company left. Rory (of Liberty) was more open-minded, the real question being "can you buy a Monday or a Wix at a price where you can make a return from leverage, operational efficiency, and slight AI growth?" noting Wix now trades "at less than 1x revenues." (Rory also mentioned buying the whole SaaS index during a prior "SaaS apocalypse" discussion and being "up 35%.")
ServiceNow and Intuit: value trap or generational buy? The market has voted "loser" on both; the podcasts this week were largely the other side of that trade (details below).
3. Specific names, with the bull and bear case as stated
Public companies
ServiceNow (NOW), the week's most-discussed name; largely bullish. On Chip Stock Investor, "ServiceNow (NOW): Down 60%, But Now a Cybersecurity Player" (July 31), Nicholas and Kasey Rossolillo made the bull case: the market is "pricing ServiceNow as an AI loser," but CEO Bill McDermott says it is "already a $1 billion-plus cybersecurity business" and "the fastest growing of the top 10 cyber companies," now the "eighth largest cybersecurity business in enterprise." Recent acquisitions, Vesa (mapping human, machine and AI identities across "2.2 billion agents entering the enterprise") and Armis (7 billion devices tracked), plus its new "AI Control Tower" product position it for "edge AI." The numbers: Q2 2026 (June quarter) revenue hit $4 billion, up about 25% year over year, extending a 20%-plus growth streak; full-year 2026 subscription revenue guidance was raised to about $16 billion. Their reverse-DCF implies the stock only needs "6 to 7% free cash flow per share growth" over ten years, "an ultra-low bar." Bear points they concede: GAAP operating income fell "55% year over year" and free cash flow "down 10%," squeezed by AI investment and acquisition amortization; the whole thesis rests on AI moving "to the edge," which the market doesn't yet believe. Morningstar's Dave Sekera (The Morning Filter) is also bullish: $165 fair value, 4-star, roughly 40% discount, with deals involving "five or more AI products" up "5.5x year over year," over "$1 billion in annual contract value," and sequential growth accelerating to "40%." His analyst's line: "one of the best blends of growth and margins in enterprise software, and results continue to show AI is not hurting the firm's fundamentals."
Microsoft (MSFT), bull. On the Eisman episode, Gil called Microsoft the clear winner: "accelerating growth right now… AI is a tailwind not only to the Azure business but to the Office business and the infrastructure software business," and yet it trades at a low multiple. He called its Outlook and Office franchise "insurmountable."
Salesforce (CRM), bear, with a nuanced counterpoint. Eisman's Gil: Salesforce "is in the category of software they're trying to cut… it has not been adding value in years and it keeps charging them more and more for that less value… a bad business that's been declining regardless of AI." The nuance came from Mark Nelson, who worked adjacent to it: he does not think CRM dies, calls the move to "headless 360" (letting customers put their own AI-built interface on top instead of Salesforce's screens) "the right move," but frames Salesforce's fate as an execution and innovator's-dilemma question.
Palantir (PLTR), bull on the business, bear on the price. On Wall Street Wildlife, "Badger's 23-Year Investing Playbook + $PLTR + $AMPG" (July 26), the bull case rested on explosive growth, government revenue up "76% year over year," commercial up "95%," across its three products (Gotham for defense and intelligence, Foundry for commercial, and the AIP artificial-intelligence-platform layer). The bear case is valuation: the host flagged Palantir at "113x free cash flow, the most expensive among 200 tracked tech companies." Eisman's Dan volunteered Palantir as a favorite ("especially Palantir"). On The Aboard Podcast, "Karp Carping: Decoding Palantir's Whole Deal" (July 28), Paul Ford and Rich Ziade argued Palantir's 20-year history wrangling "messy data" pairs naturally with modern AI, but flagged an existential risk, its dependence on third-party models (OpenAI, Anthropic, Google), which is why CEO Alex Karp is pushing partnerships (for example with Nvidia) to build independent infrastructure.
CrowdStrike (CRWD) and Palo Alto Networks (PANW), bull. Eisman's Dan named both as best-positioned in cybersecurity: "George at CrowdStrike and Nikesh at Palo Alto… are able to see around corners," riding budgets he expects to "double the next two or three years." (The ServiceNow discussion likened its Vesa deal to "Palo Alto's acquisition of CyberArk.")
Intuit (INTU), bull, as a beaten-down quality compounder. On The Investor's Podcast TIP835, "Intuit (INTU): The S&P 500's Biggest Loser" (August 2), Shawn O'Malley and Kyle Grieve laid out the setup: Intuit compounded "nearly 24% a year for a decade from 2015 through 2025," traded at a "median price-to-earnings ratio of almost 50x," and has now "sold off 60%… the worst performer in the S&P 500 this year," with its multiple collapsing "from 60x earnings to 16x." Bear case: TurboTax is "potentially as much as two-thirds of the company's operating income," and AI plus the DIY-tax narrative threatens it. Bull case: taxes are "high stakes," people want a "trusted brand" and a human backstop, not a general chatbot; QuickBooks' bank-linked financial integrations create heavy switching costs ("if you rely on Intuit for financing to pay your bills," you won't switch); and Intuit's "GenOS" plus AI agents move it toward "done-for-you" autonomous workflows. Their memorable framing of agentic AI: "the drill grows legs, walks over to the wall, and builds the shelf for you while you have a coffee." They grouped Intuit with Adobe (ADBE) and CoStar as quality names "thrown out with the bathwater."
Adobe (ADBE), mixed to bearish. Beyond the TIP framing above, Eisman's Dan put Adobe in the loser bucket with Salesforce, saying it "essentially miscalculated what AI is going to do to the business model," and said "you could say the same thing for names like Intuit."
Nvidia (NVDA), bull on infrastructure and valuation. Morningstar's Sekera called it a "GARP" (growth at a reasonable price) pick: 4-star, wide moat, fair value $280 versus a roughly $208 price, forward P/E of 22x against a 35% five-year earnings growth estimate, a PEG ratio of only 0.6. Eisman's Dan: "one chip in the world fueling the AI revolution," and "for every dollar spent on an NVIDIA chip, we estimate there's an $8 to $10 multiplier across the rest of tech." (Competitive note: startup Etched raised $300M "to take on Nvidia.")
Broadcom (AVGO), bull. Sekera's other GARP pick: 5-star, "40% discount," maker of custom AI accelerators ("XPUs"), forward P/E of 33x against a 46% growth estimate, a PEG of 0.7, with the market "underestimating the growth dynamics for these XPUs."
Micron, AMD and Intel, valuation dislocation. Eisman's Gil highlighted an inconsistency: Micron trades "at 6x earnings" as if the memory cycle is already over, while "AMD [is at] 50x… Intel 100x, as if the cycle's continuing for five more years," his argument being that the memory market is arguably stronger than CPUs right now.
IBM (IBM), the incumbent's defense, against a weak print. On WSJ Tech News Briefing, "At 115 Years Old, Can IBM Survive the AI Era?" (July 28), CEO Arvind Krishna made the bull case: IBM benefits as AI "democratizes" across "70% of the market" via smaller and edge models, positioning IBM as a "neutral infrastructure provider" for enterprises without in-house AI skills, with "80% of revenue from mission-critical" software and hardware. The context is bearish, though: this came right after a disappointing earnings pre-announcement that Eisman separately called "stunning."
Wix and Monday.com (MNDY), skepticism. In the 20VC roundtable's PE debate, these were the archetypes of mature software growing mainly on price increases; Wix was singled out as trading "at less than 1x revenues," implying the market is paying almost nothing for the core business.
Private companies
Anthropic, bull, and the anchor of Menlo's franchise. Matt Murphy detailed leading Anthropic's first round when it was "pre-revenue" at "a $4 billion-plus valuation," a "little over $10 million" first check, followed by a "$500-million-plus" special-purpose vehicle. His conviction: it's "the most innovative company around," the natural "number two" to OpenAI, and open source "just can't be powerful enough to really displace it." He noted Anthropic's own tiered model family (Sonnet, Opus, Fable) as evidence that even the labs are moving to "not one size fits all." (Separately, on Everyday AI, "Claude Opus 5" was flagged as taking the model-quality "crown.")
Lovable, bull. Murphy invested at a "$6.2 billion" valuation after it went "from zero to something like $300 million [ARR] in a year." The thesis: an "iconic entrepreneur" with a vision to turn "99% of people… who were never coders" into "creators." He also flagged the flip side, that its margins depend heavily on model costs, which open source could improve.
OpenRouter, the "Stripe for AI models." Murphy (a board member) described it as doing "for LLMs what Stripe does for money." The 20VC roundtable noted a rumored "$10 billion" deal and that routing is "a very commoditized market very quickly," with Cursor and others (Merge.dev) launching their own routers.
Cursor / Anysphere, resilient despite competition. Murphy: Cursor "was about as in the crosshairs of Anthropic as possible, and I think they still had a pretty darn good outcome."
Stripe, bull. On the 20VC roundtable, Stripe was praised as having hit a "sweet spot": premium pricing ("2.75%"), high efficiency after the Collison brothers' cost focus, and a surge from signing up "all the AI companies that are selling [online]… printing 2.75% of that money." Compared favorably to Adyen, and posed as an own-it question against fintech Revolut.
Gamma, bull, bootstrapped and AI-native scaling. Its founder detailed reaching "$100 million in ARR profitably… with a team of 50" and "600,000 paying subscribers" and "50 million users," almost entirely through product-led and creator-led growth ("word of mouth is the only form of distribution that can help amplify everything else"), famously with "zero marketing, zero sales," and only now "growing a sales team" to reach enterprises. A launch tweet ("the most valuable skill in business is about to become obsolete") went viral after "Paul Graham" responded.
Legora and "Max," hyper-fast AI legal and enterprise startups. Murphy used them to illustrate how fast the market moves now: the window where an early-growth company sits at $3 to 10M ARR "used to last a year, year and a half. Now it lasts a week, or in the case of Max and Legora, that's what they do in a day."
Etched, bull on challenger silicon. Raised "$300 million to take on Nvidia" with custom AI chips (20VC roundtable).
Deduction (Sai Dhanak), the "human backstop" model. An AI tax service where "behind every message, every filing is a licensed tax professional… who signs off on everything." Built on the thesis that pure SaaS is vulnerable but a "human liability backstop" is a durable moat, "until an AI can sign [a tax return], a human must sign it."
Auctor / Octar, bull on an agentic services platform. On Venture with Grace (July 30), the founder pitched an "operating system" for software-implementation and system-integrator firms, using "hybrid platform plus usage-based pricing" to fix scope creep in enterprise implementations; Sequoia-backed on early revenue traction.
Vertical and niche SaaS (Zenoti, Millennium, Vagaro). On Acquiring Minds (July 27), an operator described buying a salon point-of-sale software company "at 5-6x EBITDA" and converting it from on-premise licensing to full SaaS subscription, naming Zenoti (Accel, Advent, TPG and Tiger-backed), Millennium and Vagaro as competitors, a reminder that the vertical-SaaS roll-up game continues beneath the AI headlines.