Newsletter · · Ashutosh Agarwal
Death of Seat Based Pricing as SpaceX Moves to Buy Cursor for 60 Billion Dollars - Weekly SaaS / Software Podcast Recap - Week of June 28, 2026
SaaS and software podcast recap for the week of June 21-28, 2026. The week's loudest signal was the death of seat-based pricing and the 'LLM lift' moat-destruction thesis, layered over a CIO 'token maxing with no ROI' reality check, a Chinese open-source ceiling on US labs, and SpaceX's reported $60B move on Cursor.
Weekly SaaS / Software Podcast Recap
Week of June 28, 2026: Death of Seat Based Pricing as SpaceX Moves to Buy Cursor for 60 Billion Dollars
Window: June 21 – June 28, 2026 | ~40 substantively relevant episodes across 20VC, All-In, Lenny's, SaaStr, Latent Space, Big Technology, Tech Disruptors (Bloomberg), Equity (TechCrunch), Run the Numbers, Hunters & Unicorns, In Good Company, a16z Show, Market Maker, The Investor's Podcast, and others.
Top of Mind
The week's loudest signal was the death of seat-based SaaS pricing and the parallel "LLM lift" moat-destruction thesis. Jason Lamkin and Databricks both argued that AI-assisted migration is collapsing the switching costs that defined SaaS. Underneath that, a sharper enterprise-AI reality check emerged: CIOs are "token maxing" with no measurable ROI, setting up 2027 as the "show me the return" year, while Chinese open-source models (GLM-5.2) put a ceiling on US frontier-lab pricing power. M&A and capital-markets drama ran hot, SpaceX's reported $60B Cursor deal, Salesforce's record 14-day losing streak, Alphabet's worst single day on record, and a Micron blowout, alongside the recurring infra-vs-application value debate.
1. Dominant Themes
1. The seat-based pricing extinction event. The most consistent cross-show theme. Confluent CFO Rohan Savaram (Run the Numbers, 6/22) called consumption pricing "a one-way door. I don't think it was reversible," adding that for AI-native businesses usage-based pricing "is needed. Otherwise you're probably going to lose money." Jason Lamkin (20VC, 6/25) framed it as binary: "Everyone selling seats for the most part is getting crushed. Everyone selling variably one way or the other is winning." Figma's Dan Barrett (Hunters & Unicorns, 6/24) confirmed Figma is shifting from seat-based to "more of a consumption based model or a blend of both." Anthropic's Mike Krieger (Big Technology, 6/24) flagged the catch, internal analysis found "not a lot of correlation between the person who's using the most tokens and the person that is most productive," warning outcome pricing is "very gameable."
2. Enterprise AI reality check, token maxing without ROI. Databricks co-founder Arsalan (SaaStr 863, 6/24): "Everybody's like, okay, all my employees are token maxing, my spend on tokens is going up, but I have no idea what I'm getting out for it." Nikesh Arora (20VC, 6/22) quantified the stakes: "Benioff said the other day that he spends $300 million on Anthropic a year for his devs… about 3.8% of developer salary spend. If it stays there, the valuations of Anthropic and OpenAI are grossly overvalued. And if it moves to 20%, they're actually very undervalued." Rory O'Driscoll (20VC, 6/25): "The big story of 2027… is show me the ROI." Big Technology (6/27) reported Uber "blew through their entire year's AI budget in a quarter," and an audit (Vaudit) of 60 companies found $1.7M in "mistaken overcharges" from Claude Code billing, "Enterprise AI billing has become increasingly opaque."
3. Frontier model "flabby middle" plus open-source ceiling. Lamkin's new framework (20VC, 6/25): "The flabby middle in AI is at risk… OpenAI and Anthropic don't have this middle product… just when they're all ready to go IPO, they have a brand new existential risk." O'Driscoll: a 5x-cheaper open-source alternative "is gonna grind everyone down." GLM-5.2 (Zhipu AI) dominated the news cycle. David Sacks (All-In, 6/26): "GLM-5.2 is now the best open weight model for coding, software engineering, and long context agent work… right up there with GPT-5.5" (744B params, MIT-licensed, ~85% cheaper API, trained on Huawei Ascend). Gavin Baker (Atreides): heavy distillation has occurred via "harvesting reasoning traces."
4. Databricks' "LLM lift" as moat destroyer. Lamkin (20VC, 6/25): "Databricks claims they can do that lift in 30 days… It's underdiscussed and it's a moat destroyer. Your moat can be LLM lifted away." His anecdote: "We've been trying to get off Marketo for 5 years… Salesforce used their LLM thing and moved us to their product in a couple weeks. They just lifted it with no humans." Arsalan (SaaStr, 6/24): migrations that "used to take six months… can be done in six weeks," and "any industry that is highly valuable, that has a monopoly today, will not have a monopoly 12 to 24 months from now."
5. AI agents as enterprise infrastructure, but pipelines still broken. Databricks' Matei Zaharia and Reynold Xin (Latent Space, 6/24) unveiled Omnigen (open-source agent harness, 400+ merge requests in days) and "LTAP / Dream Engine" unifying OLTP+OLAP. Xin: "vector databases should have never been a separate category." Scott Brinker (Humans of Martech, 6/23): if data is stale "all we've done is give machines a faster way to be wrong… the warehouse-first architecture is almost a prerequisite for AI."
6. Anthropic/OpenAI regulatory drama and IPO timing. Anthropic's Claude "Fable/Mythos" model was pulled under White House pressure (SK Telecom access) then re-cleared by 6/27 for 100+ US institutions, with Commerce Secretary Lutnick's "stamp of approval." Sacks (All-In, 6/26) suggested Anthropic wanted federal oversight: "Dario posted a blog saying he wants an FAA for AI… it could be a FAFO situation." OpenAI's IPO reportedly slips to next year, it "can't hit the $1 trillion valuation Sam Altman has been seeking."
7. SpaceX-Cursor and coding-tool consolidation. Market Maker (6/22): SpaceX option to acquire Cursor for ~$60B (~15x ~$4B ARR; $4B break fee), giving Cursor "access to data centers currently tied up with Google and Anthropic agreements." OpenAI's Broadcom-built "Jalapeno chip" launched. Calacanis: "They are saying… F you to Jensen and NVIDIA."
8. Infrastructure vs. application layer value. Chi-Hua Chien (Goodwater, Equity, 6/24): in the web era infra created $400B of market cap vs. $3.1T for apps (88%); in mobile $700B vs. $3.7T (85%). His call: "2025, 2026 is going to be the peak of infrastructure value creation," then value shifts to apps. Philippe Laffont (Coatue, Squawk Pod, 6/23) is more infra-tilted: "Intelligence is going to become this utility for $50 or $100 a month."
9. Vertical AI TAM expansion. Nick Tippmann (TipTop, Investing in Startups, 6/24): vertical SaaS was priced as a "$300–350 billion market"; vertical AI now competes against "a $4 trillion plus services and labor budget." His framing: the moat has moved from models (22-23) to data (24-25) to "the harness" in 2026.
10. Developer productivity inflection. Fiona Fung (Anthropic, Lenny's, 6/21): "Anthropic engineers on average ship 8 times as much code per quarter as they did compared to 2025… Coding is no longer the bottleneck." Kevin Weil (OpenAI, a16z Show, 6/26): running "three or four things in parallel across different work trees" with Codex.
2. Key Debates
Debate 1, are SaaS moats dead? (LLM lift vs. switching-cost persistence). Moats are being lifted (Lamkin / Databricks Arsalan): "Your moat can be LLM lifted away"; price compression is "the new greenfield" at 30-40% of incumbent pricing. Moats persist via data, network effects, brand, context (Tippmann / Brodersen): the real edge is "the harness," workflow orchestration and domain RL loops; Adobe-style creative quality endures. Wedge: which segment, enterprise contracts with security/compliance and deep integration look durable; SMB/mid-market look highly exposed.
Debate 2, who wins the AI stack: infrastructure vs. application? Apps win long-term (Chien): infra gets "increasingly commoditized"; apps captured 85-88% of new market cap in prior platform shifts. Infra dominates near-term (Laffont / Baker / Arora): "Infra's making money… that's where you're seeing trillion-dollar market caps." Wedge: pure timeframe, Chien is a 5-10yr call; the infra bulls are 1-3yr. Both can be right.
Debate 3, is frontier-model pricing sustainable? Under threat (Big Technology / O'Driscoll): "the moment people realize they don't need the Ferrari, the Honda Accord is great"; the "flabby middle" is an existential IPO-eve risk. Worth more, not less (Baker / Laffont): Baker, "Anthropic is worth $3 trillion today… 85% gross margins on inference"; Laffont, frontier providers as "$10 trillion" intelligence utilities. Wedge: demand is agreed; the fight is whether enterprises keep paying Ferrari prices vs. routing to cheaper Chinese open-source.
Debate 4, Salesforce: executing or in structural decline? Structural bear (Closing Bell / Jeffries' Brent Dale / martech practitioners): record 14-day losing streak; "headwind of AI impact and… departures from the senior management team"; today's bounce is "a dead cat bounce"; "if anything is going to displace Salesforce, it might actually be a data warehouse." Executing on AgentForce (Fast Slow Motion / Agile Brand): AgentForce GA in Setup with analytics dashboard; Simulate acquisition accelerates AI-search roadmap "by three years." Wedge: technical feature execution vs. cultural/organizational drift and capital-allocation worries (the $3.6B Fin AI deal).
Debate 5, is enterprise AI "ready"? Not ready (Arora, 20VC, 6/22): "More than half the enterprises are still not getting it right… The software is not intelligent." The agentic enterprise is here now (Snowflake / Databricks / Anthropic): Craig Kerstiens (Snowflake, What the Dev, 6/23), "the agentic enterprise is here now… we're here today to go from prototype to actual production." Wedge: task type, data/analytics workflows are in production; zero-false-positive decision workflows are still nascent.
3. Specific Names, Bull/Bear
Public
Snowflake (SNOW), net bull. Sonu Chawla (Times Square Capital, Pitch The PM, 6/26): trades "7x revenue vs. Datadog at mid-teens" despite "30%+ growth"; Coco hit 50% customer adoption in 9 weeks (not in guidance), enabling him to "double the position size… caught that 87% move"; mid-20s FCF margins. CEO Sridhar Ramaswamy (In Good Company, 6/22): ~50% of Global 2000 addressable; concedes coding-agent model companies are his "biggest competition" but leans on data gravity.
Salesforce (CRM), net bear. Record 14-day losing streak; $3.6B Fin AI acquisition raising integration concerns; IGV software ETF -17% YTD (Closing Bell, 6/22). Brent Dale (Jeffries): "continuing to have the headwind of AI impact and… departures." Practitioner risk of demotion to "the sales-led part of the business." Offsetting bull: AgentForce GA plus analytics dashboard; Simulate accelerates AI search "by three years."
Databricks (private), strong bull. "$6 billion, growing 80%, accelerating" (Lamkin); 50%+ of Fortune 500; Genie agent powering a 70,000-user deployment; open data format "had won" vs. Snowflake's historically proprietary format (Zaharia/Xin, Latent Space, 6/24).
Palo Alto Networks (PANW), bull (CEO). Arora (20VC, 6/22): from "<2% market share" to "8 or 9%" with room to "20 or 30 or 40"; platformization, "people are realizing they can't have 40 to 60 cyber security companies." Warns of being "model captive." Bear note: SentinelOne's Weingarten skeptical of PANW's CyberArk deal.
CrowdStrike (CRWD), bull. George Kurtz (Closing Bell, 6/22): "security can actually be the accelerator"; AIDR "extremely well-received"; shadow AI driving demand. Stock "up 65 percent over the past three months."
SentinelOne (S), bull (CEO). Weingarten (Tech Disruptors, 6/24): ">50% of bookings revenue… doesn't come from endpoint protection"; Purple AI cuts triage "from hours and days to literally minutes"; questions "the role of the Windows operating system in a year or two."
Alphabet/Google (GOOGL), bear. "$250B+ single-day market cap loss, worst single day on record" (Closing Bell, 6/22); researcher departures (Jumper to Anthropic, Shazir to OpenAI); one model has Alphabet flipping "cash flow negative next year"; hyperscaler capex topping 90% of cash flow, "highest since dot-com."
Microsoft (MSFT), neutral/conflicted. Brent Dale (Jeffries): "a source of funds… worst month since December of 2000," but "your number one dodgeball draft pick. Nadella and Amy Hood"; ~18x potentially attractive. Weingarten on Security Copilot: "table stakes… by far not groundbreaking."
Meta (META), bull with capex caveat. Hari Ramachandra (Investor's Podcast, 6/21): 2026 ad revenue ~$243B (>$Google), 41% op margin, $46B FCF (2025); "Distribution becomes more advantageous… I see 46% upside." Caveat: $135B capex with unclear incremental revenue. TIP fair value ~775 vs. ~600.
Adobe (ADBE), cautious bull. Stig Brodersen (Investor's Podcast, 6/21): near a "seven-year low"; "96% of $23B revenue is subscription"; buy zone "around 200"; biggest risk is top-of-funnel ("what if you don't go there in the first place?").
Micron (MU), strong bull. Revenue up 4x YoY ($9B to $42B), beat by 16%, Q4 guide $50B vs $43B expected, 2026 HBM sold out (All-In, 6/26). Baker: "HBM DRAM will be 30-40% of all hyperscaler CapEx next year." Matt Bryson (Wedbush): "only at 10 times."
Accenture (ACN), bear. "Accenture plummets 19%… down about 40% for the year" (20VC, 6/25). O'Driscoll: "There's only one thing worse than a seat-based model… and that's a model that's based on bodies." AI SIs undercutting SAP/Salesforce implementation work.
NVIDIA (NVDA), bull (even from ex-holder). Laffont (Squawk Pod, 6/23): "13 times earnings next year… a very cheap stock"; only worry is "other people making similar chips." Watch OpenAI's Jalapeno chip as a competitive wedge.
Cerebras (public), mixed. Q1 rev $191M, GM 47%; Q2 GM guide 36-38% (G42 capacity step-down); FY26 guide $855-865M; stock -5% post-print. Bryson (Wedbush, $270 PT, Outperform): "hardware sales were better than I thought"; bottleneck is "wafers from TSM… stupidly tight." ~$20-25B OpenAI contract signed Dec 2025 (Baker).
Booking Holdings (BKNG), contested. Carlisle (Investor's Podcast, 6/21): AI-disintermediation risk vs. resilience ("Google has tried… Booking.com has continued to grow"); DCF ~$220 vs ~$167. Ramachandra: "gradually losing the mindshare," risk of becoming "just a plugin for Cloud or OpenAI."
SpaceX (public), bull, valuation contested. Laffont long, 10-15yr thesis ("Mining is going to move to space… Data centers… to space"). Market Maker (6/22): IPO $135 to peak $220 to ~$185, $2.4T cap, $19B revenue, $5B net loss; Ackman: "what makes SpaceX so valuable is how valuable it is… an extraordinarily cheap source of funding."
MongoDB (MDB), bull (historical). Dan Barrett (Hunters & Unicorns, 6/24): scaled $35M to $2B ARR; Atlas PLG plus consumption transition; CS team 35 to 350 people, "improved retention by ~14 points in two years, saved $197 million in ARR."
Confluent (CFLT), neutral/data point. CFO Savaram (Run the Numbers, 6/22) detailed the consumption-model operational playbook; no explicit bull/bear stance.
Private / AI Startups
Anthropic, complex/contested. Krieger (Big Technology, 6/24): Fable/Mythos (Claude 5) "the best model I've ever used," 8x internal eng productivity; ~$965B per episode. Baker bullish at "$3 trillion today, 85% inference gross margins." Bear: Fable regulatory fumble, "if companies or countries cannot rely on… frontier AI from US labs, they will rush toward open source"; $1.7M Claude Code billing overcharges found; accused Alibaba of a distillation attack.
OpenAI, bull on product, cautious on IPO timing. Weil (a16z Show, 6/26): GPT-5.6 series, "10 or 12 open mathematics problems solved… beyond the frontier of human knowledge," Prism (AI-native scientific writing), Jalapeno custom chip. IPO reportedly delayed (can't hit $1T this year).
Cursor / Anysphere, mega-bull on execution. ~$60B SpaceX deal, $4B ARR in ~3 years, 15x ARR, valuation $50M to $60B (~1,200x). Big Technology meme: "Cursor would have sold for 300 billion if it wasn't for" Claude Code.
Figma, bull. Barrett (6/24): "north of a billion dollars growing north of 40%"; multiplayer canvas as durable network-effect moat; shifting to consumption pricing.
Harvey (legal AI), vulnerable. Chien (Equity, 6/24): "Anthropic caused some concern for companies like Harvey when it launched a bunch of products for the legal arena."
Zhipu AI / GLM-5.2, competitive threat to US labs. 744B params, MIT-licensed, beats GPT-5.5 on SWE coding, ~85% cheaper (All-In, 6/26).
Deepseek, sovereignty play. $7.4B Series A at $50B; only Chinese entities get voting rights. Lamkin (20VC, 6/25): "China basically saying we do not want to be subservient to the United States… an existential sovereignty thing."