All-In PodcastWhy Cursor chose xAI's compute over Anthropic and OpenAI
Competing foundation model vendors threatened Cursor's independence; SpaceX structured a $60B acquisition or $10B breakup fee to lock in xAI compute alignment.
CHAPTERS
Cold open: Besties banter, DC flexing, and media narratives
The episode kicks off with playful ribbing about being “SPLC-adjacent,” personal safety jokes, and Sacks arriving late due to a White House meeting. Sacks contrasts his positive experience with the president against negative media portrayals, setting a tone of skepticism toward mainstream narratives.
- •Jokes about SPLC files and “adjacent” associations
- •Sacks explains delay: meeting at the White House
- •Discussion of how media characterizes the president vs. personal experience
- •AI policy context: leadership during the AI revolution
- •Tone-setting theme: distrust of institutional narratives
AI policy and the White House: data centers, regulation, and Anthropic
Sacks argues the administration is pragmatic and pro-building, especially on data centers and energy, and criticizes “doomer” approaches to AI. They touch on Anthropic’s relationship with government, framing it as strong product capability mixed with political friction.
- •Pro-data-center stance: “behind-the-meter” power generation idea
- •Critique of broad fear-based AI regulation vs. targeted rules
- •White House interest in AI and engagement with industry leaders
- •Anthropic praised as smart/product-strong but described as “left wing”
- •Tensions around defense relationships and who ‘tells the Pentagon what to do’
SpaceX–Cursor mega-deal explained: compute as leverage
Jason outlines a blockbuster structure: SpaceX either acquires Cursor for $60B by 2026 or pays a $10B collaboration fee, effectively a breakup fee. The group frames it as a synergy between Cursor’s coding product and SpaceX/xAI’s massive GPU capacity.
- •Deal structure: $60B acquisition option vs. $10B collaboration/breakup fee
- •Cursor financials: ~$2B run rate (Feb), targeting ~$6B by 2026
- •SpaceX revenue projections and IPO/valuation implications
- •Cursor product context: IDE + proprietary models and LLM flexibility
- •Core thesis: compute constraints + infrastructure advantage shift bargaining power
Why the partnership works: vertical integration pressure in AI coding
Sacks and Chamath argue Cursor needed independence from foundation model vendors that are now competing directly (Codex, Claude Code). Aligning with xAI provides a foundation model, compute, and a strategic moat while giving xAI enterprise distribution and coding training data.
- •Cursor’s vulnerability: dependency on OpenAI/Anthropic as they vertically integrate
- •xAI contribution: compute scale + foundation-model roadmap
- •Cursor contribution: enterprise customers, UX-leading IDE, coding workflows/data
- •Question of model stack: will Cursor move fully to Grok vs. keep multi-model choice
- •Expectation of rapid product improvement and leaderboard movement
Enterprise agents reality check: IDEs, sprawl, and token cost optimization
Friedberg reframes coding as the near-term locus of AI value and warns that agent proliferation creates inefficiency and governance problems. The besties discuss the coming need for orchestration/middleware that routes tasks to cheaper or open models to control exploding token bills.
- •Agents = fast-spun apps; enterprises risk “agent sprawl” and wasted spend
- •Strong IDE and software engineering discipline become critical infrastructure
- •Token spending is surging; incentives to optimize are lagging
- •Middleware idea: route tasks to appropriate models by cost/complexity
- •Cybersecurity flagged as a hot sub-area: specialized “cyber models” likely
Polymarket odds, IPO timing, and M&A incentives
The panel cites prediction markets indicating high odds of the SpaceX–Cursor acquisition and a near-term SpaceX IPO. They note the deal structure keeps IPO paperwork from going stale and motivates Cursor’s team to execute, while also highlighting typical post-acquisition talent/retention risks.
- •Polymarket: high probability signals for acquisition and IPO timing
- •IPO process efficiency: avoid reworking S-1/disclosures
- •Incentive alignment: huge upside to ‘make it work’ before close
- •Compute costs imply collaboration fees can flow back via usage
- •M&A risk: retention challenges after employees feel ‘set’ via stock windfalls
SaaS debt bomb: Medallia, PE leverage, and the “SaaSpocalypse”
They dissect reports that Thoma Bravo may hand Medallia to creditors after rising debt service costs and poor performance, using it as a case study for leveraged SaaS fragility. The group connects this to AI-driven substitution: enterprises can increasingly build internal tools instead of buying pricey SaaS.
- •Medallia case: $6.4B purchase (2021), ~$3B debt, refinancing/interest shock
- •Theory: PE may have de-risked via recaps, but equity impairment is real
- •Why SaaS is vulnerable: net new sales slowing + higher attrition risk
- •AI substitution: ‘spin up an agent’ vs. buy a vertical SaaS tool
- •Macro lens: AI as deflationary force lowers cost to run businesses
Private equity opportunity vs. unpredictability of cash flows
Sacks argues SaaS multiples collapsing could create buyout opportunities—if cash flows remain stable. But the core PE model breaks if retention and renewal economics become volatile due to headless/agentic usage, renegotiated contracts, or seat-based pricing disruption.
- •Historic PE value-add: a third exit path beyond IPO/M&A
- •Bull case: SaaS valuations/multiples compressed dramatically
- •Bear case: leverage requires predictability; AI introduces sudden churn/NRR drops
- •Seat-based pricing threatened as agents replace human seats
- •Key question: where is the ‘clearing price’ for software value in an AI era?
Winners’ playbook: go headless, founder-led reinvention, and Salesforce debate
The panel praises Benioff’s “headless” direction as a pragmatic adaptation, contrasting it with toll/lock-in strategies. They argue founder-led companies may be better positioned to burn the boats and reinvent, while manager-led incumbents may cling to legacy monetization.
- •Salesforce valuation shock discussed as potential bargain vs. value trap
- •‘Headless’ strategy framed as aligning with agent-driven workflows
- •Founder advantage: willingness to make radical shifts and absorb short-term pain
- •Risk: legacy cash flows may compress as customers demand value-based pricing
- •Broader thesis: AI transformation may reorder incumbents by adaptability
Debt as fragility: venture debt takedown and “optionalities” of free cash flow
They strongly criticize venture debt, arguing it reduces maneuverability, adds covenants, and can lead to predatory renegotiations during downturns. In contrast, companies with strong free cash flow gain strategic flexibility to weather disruption and invest through cycles.
- •Debt creates brittleness; reduces ability to pivot aggressively
- •Banks can’t tolerate zeros the way VCs can; incentives diverge
- •Venture debt often ‘works’ only when it’s never drawn
- •War chest optionality: survive storms, buy assets, or pivot fast
- •Personal anecdote: leverage can spiral during market disruptions
Apple succession: John Ternus, innovation gaps, and the AI-native OS layer
They debate Apple’s reported CEO transition from Tim Cook to hardware exec John Ternus, framing it as a shift from operational excellence to product-led innovation. The panel argues Apple’s biggest imperative is an AI-native assistant layer across devices, while also acknowledging Cook’s extraordinary financial stewardship.
- •Cook legacy: massive market cap growth, services mix, privacy positioning
- •Critiques: missed/slow bets on glasses, AI Siri, search, car, broader hardware leaps
- •Proposals: personalized, ubiquitous Siri-like agent integrated with user data
- •Idea: allow users to choose underlying model providers (OpenAI/xAI/Anthropic)
- •Strategic risk: over-reliance on iPhone margins in a heterogeneous device future
SPLC indictment and NGO accountability: incentives, secrecy, and 501(c) reform
The besties react to allegations that the Southern Poverty Law Center used hidden accounts to fund informants and extremist-group activity, calling it emblematic of NGO grift and misaligned incentives. They broaden the critique to nonprofit governance, arguing many 501(c) organizations operate outside the spirit of tax exemption and require audits, transparency, and structural reform.
- •Allegations: wire fraud/money laundering tied to concealed payments and informants
- •Argument: incentives can reward ‘creating the problem’ to fundraise against it
- •Nonprofit critique: fundraising replaces market feedback as success metric
- •Call for transparency/auditing; question of why NGOs act like law enforcement
- •Broader point: many nonprofits don’t fit IRS “exempt purpose” definition
Science Corner: pesticide link to rising early-onset colon cancer
Friedberg summarizes a study comparing gene-expression/epigenomic signatures in colorectal tumors of younger vs. older patients, identifying a strong association with the herbicide picloram. He argues the finding should trigger modernized EPA review processes that incorporate epigenomic methods to catch long-latency harms.
- •Early-onset colorectal cancer up ~80% in under-50s over two decades
- •Method: tumor gene-expression differences used to infer environmental triggers
- •Top signal: picloram (persistent herbicide, introduced in 1963)
- •County-level correlation: higher usage linked to higher cancer rates (strong odds ratio)
- •Policy implication: update chemical safety evaluation using epigenomic evidence
