Newsletter · · Ashutosh Agarwal
Salesforce and Oracle Slide as Vibe Coding and Outcome Pricing Reset SaaS - Weekly SaaS / Software Podcast Recap - Week of July 19, 2026
SaaS and software podcast intelligence for the week of July 12 to 19, 2026. Operators and investors dug into a roughly $1.4 trillion software valuation reset, the shift from seat-based to outcome and usage pricing, the vibe-coding threat of firms building their own tools, cheap commoditized AI models, and the AI data-center capex question, with named calls on Salesforce, Oracle, Microsoft, CrowdStrike, Palo Alto Networks and more.
Weekly SaaS / Software Podcast Recap
Week of July 19, 2026: Salesforce and Oracle Slide as Vibe Coding and Outcome Pricing Reset SaaS
The one-sentence version: This was the week the "is software dead?" question stopped being a scary abstraction and turned into a set of hard numbers, real quotes, and named winners and losers. Almost every serious software conversation on the podcasts circled the same handful of ideas, a roughly $1.4 trillion wipeout in software market value since late 2025, a fast shift away from charging "per login" toward charging for results, a wave of people building their own software instead of buying it ("vibe coding"), and a growing fear that the trillion-dollar AI data-center build may not earn its keep. The mood was not "software is finished." It was closer to what one veteran investor put it: software "isn't dead. It's gotten harder."
Below: the themes everyone kept returning to, the arguments where smart people flatly disagreed, and a company-by-company rundown of who got talked up and who got talked down.
1. Dominant Themes
Theme 1, The "SaaSpocalypse" and the great multiple reset
The single loudest topic was the crash in software stock valuations and whether it's a fair repricing or an overreaction. The clearest framing came from Ahmad AlZaini, founder of the $13B restaurant-software company Foodics, speaking on with Loulou (July 16). He tied the sell-off to a specific trigger: "The moment Anthropic launched that model [in February], the sell-off in the U.S. market is around $300 billion... in software, in SaaS." Adding it up from October 2025, he said, "the number of market cap lost is around $1.4 trillion... Figma's lost 70, 80% of their value, Salesforce, 30, 40%." His diagnosis of why: "before, investors were telling themselves that SaaS growth will be tied with [seats] growth", the more people you hire, the more software seats you buy, and "they discover that this is not anymore true. Because companies are not hiring more people necessarily."
But, and this is the important nuance, he insisted the underlying business is fine: "software is actually not dying. AI is not killing software. Actually, this year, the total revenue from software businesses is $1.5 trillion... 15% growth versus last year." The stocks fell; demand did not.
That "it's the growth rate, not the death of software" point was echoed on The Official SaaStr Podcast (July 15), where Scale Venture Partners' Rory O'Driscoll gave the episode its title, "Software Isn't Dead. It's Gotten Harder." His line to founders: "the collapse in the SaaS multiples has happened because of the collapse in the SaaS growth rate... the average company used to grow at 30% and now it's what? It's sub-10... So they're not dying for nothing. Wall Street's not just being a bitch, Wall Street's on their case 'cause they're not growing."
On the ground the drop is brutal. On Money School Elite (July 16) a wealth manager to high-net-worth families said software-as-a-service equities are "down 50%," with private credit portfolios "down 25-50%," and that his clients are asking the obvious question: "is this overblown?"
Theme 2, From "paying for logins" to paying for outcomes
If seats are dead as a pricing model, what replaces them? Nearly everyone landed on the same answer: usage and outcomes. AlZaini again: "In 2030, people will start selling outcomes, results and usage... they will not pay per seat." Instead of logging into a portal to change a menu price, "you will just talk with your agent."
The most concrete version came from Bret Taylor, OpenAI's chairman and CEO of the AI-agent startup Sierra, and former co-CEO of Salesforce, on Squawk on the Street (July 16). His pitch for a new pricing model: "With a Sierra agent, you're not paying for tokens. You're paying for business outcomes delivered. You're paying for that appointment being booked." He was blunt about what that means for traditional software: "I would think that would instill fear in a lot of software companies. If they are used to the subscription model, if we are starting to pay for outcomes, that might be a longer conversation." Sierra's new "Horizon" product, he said, chases "long horizon goals", one health-system customer has "100,000 referrals... that go unfulfilled every year," and a Sierra agent will "call the specialist on the phone to get that appointment scheduled" over days or weeks to recover that lost revenue.
Pricing expert Dan Balcauski, on Navigating the Customer Experience (July 14), put a striking figure on how much upheaval this is causing: of all application-software companies that have shipped AI features, "100% of them have revised pricing and packaging within 18 months." His practical advice is a new "early access" stage between beta and general release, where a company "announces our limits... put a price on these features, but we're not going to enforce those limits or meter that", a way to get customers hooked before the meter starts, so "the CFO doesn't have a stroke" when the token bill arrives. A related discussion on AI to ROI (July 16) laid out the migration path: AI-native services companies "start with labor-based pricing because that's what buyers understand," then move to outcome-based pricing "as their AI capabilities mature."
Theme 3, "Vibe coding": building your own software instead of buying it
A genuinely new competitive threat kept surfacing, companies using AI to build in-house what they used to buy from software vendors. The Accounting Podcast (July 17) had the most vivid examples: Starbucks "wants to eliminate or slash their software spend, which is now at $400 million a year." A top-100 accounting firm needed a financial-reporting tool that "would have cost around $200,000 annually" from a vendor, "instead, his team used AI, Word, and Excel, and they built their own." Another CEO was quoted $300,000 a year for audit software and "built an app that cost under $30,000 using vibe coding" (Claude plus Retool). The hosts flagged the catch honestly: "he's the only person who knows how it works", a single-point-of-failure risk that vendors are betting will limit the damage. Xero, meanwhile, is leaning in, telling users "we're not going after traditional developers anymore. We're going after... accountants... to vibe code your own code."
The venture world sees the same wave from the supply side. Wix founder Avishai Abrahami, on 20VC (July 13), was refreshingly candid that this is partly a market-shrinking force ("maybe that's 20%... the TAM that was 100% is now 80%"), but argued most small businesses won't do it themselves: "your business in life is making pizza." Wix's hedge is Base44, the vibe-coding startup it bought, famously a one-person company, for "roughly $80 million," now "probably double the revenue that you bought it for."
Theme 4, The model layer is being commoditized, and that may be good for software
A major storyline was a price war among the underlying AI "models," and what it means for everyone built on top. On Big Technology Podcast (July 17), the hosts walked through investor Gavin Baker's much-shared analysis of China's new Kimi K3 model. His core point: a world with "only 2 to 3 dominant frontier labs with 90% inference margins is net negative for every other layer." So "anything that lowers the margins and increases competition at the model layer is good for every other AI layer: power, semiconductors, hyperscalers... and yes, even software." Kimi K3 is "40% cheaper than GPT-5.6, 70% cheaper than Fable" while being competitive on quality, different from earlier Chinese models that competed purely on being cheap.
The practical read for software investors: if the raw intelligence gets cheap and interchangeable, value shifts to whoever owns the customer relationship, the data, and the workflow. Bret Taylor made exactly this argument: "The AI itself will become more commoditized... What do you have? You have your relationships with your customers and you have the memory of all those interactions." Microsoft's Satya Nadella coined a phrase for why buyers should be wary of feeding their data straight to a model maker, the "reverse information paradox": "you essentially pay for intelligence twice, once with money and again with... the proprietary knowledge you must reveal to make that intelligence useful."
Theme 5, The AI data-center build: who pays, and does the math work?
Even on software-focused shows, the elephant was the trillion-dollar infrastructure build behind AI. Estimates flew around: Facts vs Feelings (July 15) cited hyperscaler capex rising "from initially projected $470 billion in 2026 to $740 billion," and 2027 "from $530 billion to nearly $900 billion", a jump from "1.7% to 2.5% of US GDP." Squawk Pod (July 17) cited 2026 estimates rising "from $500 billion to $700 billion."
The bearish counterweight was Jim Chanos on RiskReversal Pod (July 17). His key metric: the hyperscalers' return on incremental invested capital "has gone from, as a group, 40% a year and a half ago to about 20% today. And if the spend keeps up... it's going to be moving toward 10%." His warning most relevant to software investors: "No one that is dependent upon NVIDIA to exist should trade at a higher valuation than NVIDIA itself. And yet many, many companies do." He also flagged that the whole build rests on short contracts financing long assets: "people are committing long-term capital projects based on near-term spot pricing... that's a terrifying thing."
The counterparty angle showed up on The Dividend Cafe (July 17), quoting S&P's credit downgrade of Oracle: OpenAI "makes up roughly half of the $638 billion in remaining performance obligations. If OpenAI were unable to pay Oracle... Oracle would be left with massive data center leases that it might be unable to exit." The host added that hyperscalers "have already spent over 100% of their free cash flow" on AI, forcing new debt or share issuance.
Theme 6, Cybersecurity: the one software corner where AI is a clear tailwind
Cyber was the sub-sector where the AI story is read as unambiguously good for demand, offset by very rich valuations and a frenzy of dealmaking. More on the names below, but the through-line (from Wall Street Wildlife, July 12) is that AI makes attacks cheaper and more dangerous, so "the world needs like 10 times as much cyber security as it does today."
Theme 7, Infrastructure software: databases built for agents, not humans
On the data-platform side, the pitch is that the user of software is becoming an AI agent rather than a person. Databricks co-founder Reynold Xin, on DataFramed (July 13), described new products, LTAP (a platform unifying transactional and analytics data), sub-100-millisecond real-time querying, and a governance layer for agents, arguing "you need to completely rethink how databases and data warehouses are built" now that agents, not analysts, are the main users.
2. Key Debates
Debate 1, Is SaaS dead, or just repriced? (Bears vs. "it's harder, not dead")
On one side, the market has voted brutally, a ~$1.4T value loss, SaaS equities down ~50%. On the other, operators and investors argue the fear outran the fundamentals. Rory O'Driscoll (SaaStr): "Software isn't dead. It's gotten harder... it's tricky, but it isn't dead." Canva co-founder Cameron Adams, on Masters of Scale (July 14), flatly rejected the "SaaSpocalypse": "this notion that there's going to be one place that the entire world goes to to do all of their work, I don't think is going to be a reality. Tools like Canva that focus on particular areas... will be able to create an experience that is far deeper." The value case for beaten-down software was made concretely on Yet Another Value Podcast (July 14) around accounting roll-up CBIZ ($CBZ): the guest argued the stock fell because "this used to trade at 18x because we thought this was recession-resistant" and now "everybody... fears AI automation will disrupt the industry", but well-capitalized incumbents "in an age of AI disruption... are positioned to massively benefit" by consolidating weaker rivals (the stock trades near "9x free cash flow"). The wedge: whether AI permanently caps software's growth rate, or whether the strongest franchises re-accelerate and today's multiples are a gift.
Debate 2, Seat-based vs. outcome/usage pricing
Almost everyone agrees seats are under pressure, but there's real disagreement on how fast and how cleanly the switch happens. The outcome-pricing camp (Bret Taylor/Sierra, Foodics, AI to ROI) sees it as inevitable and imminent. The skeptics point to how messy it is in practice, Dan Balcauski's "100% revised pricing within 18 months" is a story of chaos, not clean transition, and he warns that "where we see a lot of pricing changes blow up is not that companies are changing it, but that they're communicating it poorly." The wedge: outcome-based pricing aligns vendor and customer, but it also makes revenue harder to forecast, a real problem for public-market software multiples that were built on predictable recurring revenue.
Debate 3, Does cheap, commoditized AI help software or hurt it?
Gavin Baker's framing (via Big Technology) is that model commoditization is "a godsend for software", cheaper inputs, more margin for the app layer. The bear reply, articulated on Marketing in the Age of AI (July 13): if all you are is "a clever prompt... you do not have a business. You have a wrapper," and "the foundation model ships your feature for free." The wedge: cheap intelligence rewards software companies with real moats (proprietary data, sticky workflows, owned distribution) and annihilates the thin wrappers in between.
Debate 4, Is the AI data-center build a bubble? (Chanos bear vs. "you can't stop the train" bull)
Chanos's deteriorating-ROIC case (40% → 20% → 10%) versus the bull view on TFTC (July 18): "the gears are going and you can't stop the train," and the hyperscalers "own the compute infrastructure so they benefit regardless of whether OpenAI or Anthropic succeed." Peter Schiff, on The Peter Schiff Show (July 15), took the harshest line, the gear "becomes obsolete in 2-3 years rather than lasting 20-30 years, making this really operating expense not capital investment." The wedge: whether the current spend will ever earn an acceptable return, or whether late-2026/2027 brings the moment when, in Chanos's words, "people are going to start putting pencil to paper... and say, wait a minute, does this next trillion dollars make sense if you're only going to earn 50 billion on it? We can earn that on treasuries."
Debate 5, Can incumbents like Salesforce and ServiceNow actually monetize AI agents?
The optimistic case came straight from inside Salesforce on Everyday AI (July 14): its Slack bot is "the fastest growing feature in Salesforce history," with "1.8 million Slack bot messages... sent inside of sales by Salesforce employees just last week," positioned as "the 2% of your AI spend that unlocks the value from the other 90%." The skeptical case came from an enterprise-software analyst panel on WBSRocks (July 14): Salesforce's Agentforce lives in "a walled garden" without "full business context" (most companies also run ServiceNow, custom apps, etc.), the pricing "packaging hasn't quite got worked out yet," and "if agentforce DevOps capabilities mature, this Salesforce-only positioning almost becomes a single point of failure." The wedge: incumbents own the data and the customer, but AI-native challengers (Sierra, Monaco, Legora) are unburdened by legacy pricing models and moving faster.
Debate 6, Build vs. buy (vibe coding)
Covered above, the risk to vendors (Starbucks slashing a $400M software bill; firms replacing $200–300K tools for under $30K) versus the "software vendors aren't losing sleep yet" rebuttal that home-built tools create single-person key-man risk and lack support. The wedge: whether "documented-by-the-AI" code and better tools eventually neutralize the maintainability problem that today protects incumbents.
Debate 7, Where does value accrue, models, infra, or apps? (Coatue's talent-and-capital view)
On How I Invest with David Weisburd (July 13), a Coatue investor argued the durable advantage sits with the frontier labs that have "talent and capital," noting OpenAI and Anthropic combined have "over 80 billion" in rumored revenue, "adding $10 billion a month", and that in this era "talent... matters way more than it did in a SaaS world where all SaaS companies tasted like chicken." That's a direct challenge to the "value flows to the app layer" thesis.
3. Specific Names, who got talked up and who got talked down
Public software & infrastructure
Salesforce (CRM), Mixed-to-bearish on the stock, bullish on the product engagement. Named as a top casualty of the reset (Foodics: "Salesforce, 30, 40%" off). Bull data point: the Slack bot's explosive internal usage (1.8M messages/week). Bear/skeptic: Agentforce's walled-garden limits and unclear pricing (WBSRocks).
Oracle (ORCL), Bearish tilt on credit risk. The S&P downgrade and the OpenAI counterparty exposure (~half of $638B in obligations) featured on The Dividend Cafe. Chanos: Oracle "has the worst of those ROIIC metrics." On The a16z Show (July 15), noted as a share gainer in cloud, with "Azure losing a lot of share to Oracle and CoreWeave and Google."
Microsoft (MSFT), Sharply criticized on execution. On the a16z show, the guest was scathing: "Azure is losing a lot of share... Their internal chip effort is by far the worst of any hyperscaler... GitHub Copilot is failing. Microsoft Copilot is like still crap... they have the best B2B relationship with so many enterprises... but they end up not having the actual product to sell them." GitHub Copilot's desktop app relaunch was covered more neutrally on MacBreak Weekly (July 15).
CrowdStrike (CRWD), Bullish thesis, but "nosebleed" valuation. On Wall Street Wildlife: ARR "$5.5 billion... up 24% year over year," record free cash flow of "$469 million with a 34% free cash flow margin," a "$174 billion market cap," trading at "33 times enterprise [value to] sales." The bull case: AI makes cyberattacks far more dangerous, driving demand; the "Falcon Flex" flexible-credit model supercharges land-and-expand; and CEO George Kurtz's handling of the 2024 outage built long-run trust. The bear case: "price for perfection... I don't know how well I'd sleep buying at today's valuations."
Palo Alto Networks (PANW), Bullish but "Frankenstein," and cheaper than CRWD for a reason. Next-gen security ARR "over $8 billion," "$228 billion" market cap, "21 times trailing sales", expensive but "objectively cheaper than CrowdStrike." Headline growth "31%... the true number is closer to 14, 15%" because much of it is acquired revenue: it bought CyberArk for "$25 billion" (identity security) and observability player Chronosphere. The knock: a "Frankenstein... platformization" of bolted-together products versus CrowdStrike's single-agent design.
Databricks (private, but a bellwether), Strongly bullish. Coatue: CEO Ali Ghodsi has hopped multiple technology waves, and "Databricks' growth rate's over 80% now" versus Snowflake's "mid double digits." New agent-era products (LTAP, real-time lakehouse) via DataFramed. Reached a "$188 billion" private valuation (Practical News, July 18). Also moving into cybersecurity (bought Panther) per Resilient Cyber (July 16).
Snowflake (SNOW), Bearish-relative. Coatue's read was unflattering: its slower growth "reflects inferior talent and strategic decisions" versus Databricks.
NVIDIA (NVDA), The valuation anchor. Chanos: nothing dependent on NVIDIA should trade above it. The a16z show flagged the real threat as custom silicon, Google "making millions of TPUs," Amazon "millions of Trainium", with the provocative suggestion that Google's TPUs might be worth more sold on the open market than kept internal.
Micron (MU), Bearish caution from Chanos, who noted a company "that had negative gross margins three years ago" now being awarded huge market cap on long-term memory contracts (LTAs) built on today's spot prices.
IBM, Cautionary tale. Its "worst day for that stock in history" (per Squawk on the Street) came after redirecting capex toward supply-constrained servers, storage, and memory ahead of price hikes.
Wix (WIX), Founder-endorsed, market-punished. Abrahami said the stock "trades on other companies' news... mostly on what OpenAI or Anthropic or Google are saying," admitted the Base44 buyback was badly timed, but defended the ~$80M Base44 acquisition (now ~$160M revenue) and buybacks generally. Generates "about $400 million a year in free cash flow," investing "$200 and something million into Base44."
CBIZ (CBZ), Contrarian value long (Yet Another Value): trading near 9x free cash flow, positioned to consolidate AI-threatened small accounting firms; debate centered on whether to restart its M&A "flywheel" versus keep buying back stock. Caveat raised: 3.5x leverage into an "existential" AI risk, and an aging board.
Canva (private), Bullish self-portrait. Now "a quarter of a billion people... every single month," targeting 1 billion; built "the world's first foundational design model"; rejects the SaaSpocalypse. Also flagged (by Foodics) as having a slice of the value narrative shift.
Private AI-native disruptors (the names to watch)
Sierra (Bret Taylor), outcome-based AI agents; new "Horizon" long-horizon product; explicitly aiming to replace subscription software pricing.
Legora, AI for legal, on All-In (July 13). Growth: "50% quarter over quarter for the last seven quarters," and it just became one of the fastest enterprise companies "to go from one to 150 [million], beating Sierra." The opportunity: legal is "a trillion dollars every year in legal services... but the software spend is about 40 billion. So it means there's 4% software, 96% service." Kirkland Ellis alone "turns around $10 billion a year" with partners billing up to "$4,000 an hour." Legora uses "forward-deployed lawyers" (à la Palantir's forward-deployed engineers). Legacy incumbents LexisNexis and Westlaw were described as "getting crushed" by the AI uncertainty. Competitor Harvey was named alongside it.
Cursor / Anysphere, the standout financial story, via Marketing in the Age of AI: "crossed roughly $2 billion in annualized revenue with a small team," valuation "$50 to $60 billion range... among the highest revenue per employee companies in the history of tech." On a16z, noted as having (with Claude Code) blown past GitHub Copilot.
Lovable, "about $500 million in ARR, more than 8 million users, and a valuation around $6.6 billion," with "more than 100,000 new projects... every single day." (Its CEO also appeared on All-In this week.)
Base44, the one-person company Wix bought for ~$80M, now ~$160M revenue.
FieldGuide, AI-native audit/accounting (AI to ROI): used by "almost half, 50% of the top 100 US accounting firms," including KPMG; Goldman Sachs "led a $75 million Series C at a $700 million valuation." Peer Luminance trained on "150 million legal documents."
OpenAI & Anthropic, the frontier labs at the center of everything. Bull case (Coatue): combined ~$80B revenue, adding ~$10B/month, durable talent-and-capital moats. Bear case (Big Technology, Motley Fool Hidden Gems, Tom Bilyeu's Impact Theory): the 90%-margin model business is being commoditized by Kimi K3, Grok and Meta; OpenAI reportedly "burned $20.9 billion in 2025." Anthropic's answer has been to shift toward products, its revenue "10x'd" over the past year, and Claude Code was singled out repeatedly as a genuinely great product. On IPO timing, PowerLaw's CEO on Brew Markets (July 17) said an OpenAI listing is now "definitely a 27... not a 26."
Stripe, Ramp, AlphaSense (private fundings as signals), Marketing in the Age of AI flagged the money trail: Ramp raised "$750 million at a $44 billion valuation" (spend/AI-cost management), AlphaSense "$350 million at a $7.5 billion valuation" with "$650 million in ARR" (AI market intelligence). Brew Markets discussed Stripe's reported bid (with Advent) to take PayPal private "for just over $50 billion," and noted Stripe and Databricks now "operate like public companies" with no need to IPO.