The Twenty Minute VCThe SaaS Massacre: Public Market Collapse |Microsoft Lost $360B & NVIDIA’s $100B Dispute with OpenAI
CHAPTERS
- 0:00 – 7:13
SpaceX acquires xAI: portfolio “load balancing,” dilution, and the industrial logic of the merger
The panel reacts to SpaceX completing the acquisition of xAI at a reported $1.25T combined valuation. They debate who wins economically (SpaceX holders facing dilution vs. xAI/Twitter holders rolling into a larger entity), and why Elon repeatedly recombines assets when capital needs shift.
- •Elon’s pattern: rescuing/boosting related entities via consolidation (SolarCity→Tesla analogy)
- •How dilution is partially “softened” by markup and liquidity (secondary) for holders
- •Who benefits most: SpaceX investors vs. xAI/Twitter investors in the combined cap table
- •Industrial rationale teased: data centers and compute—potentially even “in space”
- •Elon’s style: ‘no investor left behind’ framing
- 7:13 – 11:41
AI capital intensity forces a return to IPOs: the end of “stay private forever”
They argue that even the most celebrated AI companies are running into the limits of private capital. That pressure is pushing firms to get serious about IPOs, shifting the cultural narrative from avoiding public markets to needing them for massive compute-driven expansion.
- •Signals: Anthropic IPO planning, OpenAI slowing hiring, mega-deals reshuffling access to capital
- •‘Rehabilitation of the IPO’ as a structural outcome of compute capex needs
- •Private capital is large but insufficient for the scale implied by AI roadmaps
- •Public markets become the only funding pool deep enough for sustained capex
- •IPO readiness becomes strategy, not stigma
- 11:41 – 14:06
Compute-to-revenue “perpetual motion”: inference as the new sales & marketing
Jason frames AI economics as a near 1:1 relationship between compute and revenue—encouraging companies to raise and spend aggressively. They compare this to the old SaaS era when sales & marketing could efficiently convert spend into ARR, arguing inference now plays that role at larger scale.
- •Compute→revenue correlation drives rational ‘raise everything’ behavior
- •Microsoft and others echo compute scarcity and allocation tradeoffs
- •“Inference is the new sales and marketing” as a venture underwriting heuristic
- •AI products must be viral/ROI-obvious so usage becomes the growth engine
- •Traditional playbooks (Rule of 40, etc.) feel broken in this regime
- 14:06 – 18:33
Durability crisis and the ‘SaaS Massacre’: why public software keeps decelerating
The conversation pivots from AI optimism to public SaaS pessimism: markets are repricing recurring revenue durability. Even where churn hasn’t spiked, new logo growth is slowing, budgets are reallocated to AI, and public investors punish deceleration harshly.
- •Public software growth has slowed quarter-after-quarter since Q1’22 (basket effect)
- •Durability vs. growth: churn may be stable for systems of record, but new customer growth slows
- •CIO attention and budget shift toward AI crowds out legacy upsells
- •Saturation: ‘anyone who needs a CRM at scale has one’ dynamic
- •Market narrative: SaaS terminal value questioned → rapid multiple compression
- 18:33 – 20:52
Winners vs. losers in SaaS: systems of record, “seats” pressure, and valuation bottoms
They build a framework for which SaaS categories are most at risk: systems of record (finance, core backends) vs. lighter workflow/task tools. Rory argues true bottoms come when stocks are priced on free cash flow multiples net of dilution, reflecting a full regime shift away from pure revenue multiples.
- •Systems of record (SAP/Oracle-like) are harder to rip out than task apps
- •Seat contraction: renewals asking for fewer seats is a key leading indicator
- •SMB and workflow tools face weaker stickiness and faster disruption
- •Valuation reset: from revenue multiples ignoring SBC to FCF multiples net of dilution
- •Bottom-finding heuristic: when priced like cash-flow assets rather than growth stories
- 20:52 – 30:15
Venture strategy under extreme growth dispersion: when to double down vs. give up
They discuss how the gap between ‘insane growers’ and everyone else is forcing venture behavior changes. Harry and Jason note investors triage earlier due to opportunity cost, while Rory warns against reflexively abandoning companies and emphasizes compounding, economics, and founder relationships.
- •Two classes: ‘growing like a beast’ vs. ‘unfundable’ (AI label matters less)
- •Career incentives: partners feel they have ~22 months to prove themselves
- •Earlier “give up” behavior increases—but can be costly if the next deal doesn’t save the fund
- •Rory’s dot-com lesson: hypergrowth can mask broken unit economics; slower compounders can win
- •Competitive position decay matters as much as absolute growth
- 30:15 – 37:28
Next-gen CRM war: HubSpot valuation collapse vs agentic “CRM” startups
They reconcile how incumbents like HubSpot can trade at low multiples while VC-backed ‘next-gen CRM’ companies raise at huge valuations. The key distinction: many startups aren’t rebuilding CRM; they’re selling agentic customer acquisition and sales execution that can replace human labor and command high budgets—often with churn risk.
- •“Dad VC” dynamic: VCs invest in familiar categories like CRM, sometimes without true differentiation
- •Agentic sales tools sell outcomes (pipeline/bookings) vs seat-based pricing
- •SMB buyers face seat pressure, making incumbents more vulnerable if growth depends on seat expansion
- •Churn risk: promises of automated pipeline often fail due to TAM/sales-cycle constraints
- •Qualified/Artisan examples: strong qualification and refusal to onboard bad-fit customers
- 37:28 – 48:09
Own the CRM stack or build on Salesforce? SMB vs enterprise dynamics
They debate whether the agentic layer must own the full CRM stack or can sit atop Salesforce. The conclusion is nuanced: enterprise deals can support data cleanup and integration, making ‘agents on top’ viable; SMB often needs an integrated platform because messy data and small contract sizes can’t carry implementation costs.
- •Counterpoint to ‘must own full stack’: agents can succeed atop Salesforce in enterprise contexts
- •Economic constraint: data cleanliness and onboarding/FDE cost determine feasibility
- •Vertical/SB vs enterprise: platforms like Salesforce attract rich ecosystems; SMB stacks may need tighter integration
- •Shopify analogy: markets may price in platform owners absorbing partner ecosystems
- •Incumbent advantage = distribution; startup advantage = better product—race depends on stickiness
- 48:09 – 55:37
Microsoft’s $360B market-cap hit: narrative reversal around OpenAI, Azure, and product execution
They unpack Microsoft’s sharp selloff despite a small Azure growth miss. The panel argues investor concern centers on OpenAI-linked RPO quality, GPU allocation tradeoffs, and a broader narrative shift: Microsoft executed brilliantly in corp dev but lacks clearly dominant, owned AI products at the model/app layer.
- •Azure miss framed as within margin, but markets react to narrative breaks
- •RPO concentration: large share tied to OpenAI sparks skepticism about conversion to cash
- •GPU allocation: internal product use vs Azure supply affects reported growth
- •Strategic critique: Microsoft owns a big stake in OpenAI but lacks its own standout LLM/app suite
- •Mega-cap valuation context: sustaining long-run multiple requires AI relevance and execution
- 55:37 – 1:00:17
NVIDIA’s “up to $100B” OpenAI investment dispute: negotiation optics and systemic ripple effects
They analyze Jensen Huang’s walk-back and how the original joint language (“intends to invest up to $100B”) created expectations. The discussion expands into second-order impacts: if OpenAI growth assumptions are revised downward, the ripple hits suppliers and partners—even if nothing ‘fails’ in an absolute sense.
- •Press release semantics vs market expectations: ‘up to $100B’ still implied serious intent
- •Negotiation dynamics: public friction signals high-stakes bargaining and misalignment
- •Ripple risk is about derivatives (growth rate and re-acceleration), not existential collapse
- •NVIDIA’s incentives: maintain ecosystem demand while avoiding overly circular financing
- •OpenAI ‘missing’ could mean 5x vs 10x growth—still huge, but disruptive to spend plans
- 1:00:17 – 1:08:00
Could governments backstop AI data centers? 0% financing debate and ROI reality check
Jason proposes the possibility of government-supported financing for data centers to sustain AI-driven growth, citing precedent-like interventions and geopolitical competition. Rory pushes back: capital access isn’t the bottleneck; ROI is—if returns don’t pencil out, guarantees won’t fix overbuild.
- •Hypothesis: low/zero-cost loans or guarantees to keep AI capex expanding
- •Motivations: macro stability, retirement accounts exposure, and ‘China is doing it’ argument
- •Rory’s rebuttal: hyperscalers/public markets already provide capital; the risk is poor returns
- •Bubble crash modes: ‘money runs out’ vs ‘business case fails’—AI more likely the latter
- •Political feasibility skepticism, plus longer-term toxicity concerns
- 1:08:00 – 1:18:27
Waymo raises $16B at $110B: growth dispersion, Tesla comparison, and cost-structure challenges
They argue Waymo’s huge round is not a contradiction to SaaS multiple compression but the other side of it: markets reward perceived future dominance and explosive growth. Rory frames Waymo as ‘cheap’ relative to implied Tesla autonomy valuation, while noting execution risks largely sit in cost structure rather than core tech feasibility.
- •High dispersion regime: ‘old and boring’ reprices down; ‘new and exciting’ reprices up
- •Waymo market size: autonomy targets massive driver displacement and transport spend
- •Relative valuation: Tesla autonomy implied value vs Waymo’s actual operating service
- •Capacity constraints: current revenue is constrained by fleet/coverage, not demand
- •Key risks: vehicle and sensor costs, teleop, peak utilization, and long path to strong margins
- 1:18:27 – 1:23:01
The Elon premium and key-person risk: why traditional valuation breaks on Musk-led assets
They discuss how much value in Tesla/SpaceX is attributed to ‘Elon will figure it out’ optionality, making conventional modeling difficult. This introduces extreme key-person risk: an unexpected departure could cause outsized value destruction, even if operational benches are strong in the short term.
- •Rory’s model: a large share of value is ‘Elon premium’ vs current fundamentals
- •Entry valuation matters: unlike Apple’s low multiple at Cook transition, these assets price in genius
- •Key-person risk: retirement/accident scenario could trigger sharp repricing
- •Bench strength may sustain existing roadmaps temporarily, but long-run innovation premium is fragile
- •Investor challenge: paying for optionality at massive scale
- 1:23:01 – 1:37:14
OpenClaw + Moltbook: agents with tools, agent-to-agent networks, and security blowback
They close on OpenClaw (a tool-using local agent) and Moltbook (a social network for agents), which drew ~1.5M agents quickly. While much of the content is ‘punked’ or synthetic, they see the real significance as early mass agent connectivity—and the dangers of granting agents broad permissions in an adversarial networked environment.
- •OpenClaw: local tool-use agents that can manipulate files, email, and workflows
- •Moltbook: agents join a network and interact socially; rapid scaling to ~1.5M agents
- •Not sentience: many posts are prompted or synthetic, but it previews agent-to-agent communication
- •Security risks: breaches, leaked credentials, silent updates, hidden DMs, and tool-access escalation
- •Practical takeaway: tightly scope permissions; never give networked agents full access to sensitive systems
