The Twenty Minute VCAnthropic Raises $30BN at $900BN Price | SpaceX Files S1: How Does it Trade | Cerebras Smashes Day 1
CHAPTERS
- 0:00 – 1:17
AI backlash brewing: scientists vs. politics in the age of layoffs
Rory frames the episode with a warning: AI leaders may be technically brilliant but politically naïve, and public sentiment is turning. The group tees up a week packed with mega-valuations, public market moves, and blockbuster IPO chatter.
- •Public opinion outside Silicon Valley is increasingly hostile to AI hype
- •Tech leaders may be underestimating political consequences
- •Preview of the episode’s major stories: Anthropic, public SaaS rebound, Cerebras, SpaceX
- 1:17 – 2:09
Anthropic targets $900B+ valuation as Andrej Karpathy joins
Harry introduces two headline items: Anthropic’s enormous prospective raise and Karpathy’s move to the company. The panel separates the fundraising mechanics from the talent signal and discusses what each implies.
- •Anthropic reportedly in talks to raise $30B at a $900B+ valuation
- •Andrej Karpathy joining is treated as a meaningful talent/momentum signal
- •Investors listed span major growth funds and mega-check writers
- 2:09 – 4:56
Does valuation math still matter? ARR multiples vs. DCF in AI mega-rounds
Jason and Rory debate whether traditional valuation frameworks still apply and why Anthropic’s rounds can look “cheap” relative to earlier-stage venture. They argue Anthropic carries less existential risk than typical private rounds, shifting the risk profile toward valuation and execution.
- •ARR multiples still drive many growth rounds despite shifting macro conditions
- •Later-stage AI leaders may have “post-IPO” characteristics with lower existential risk
- •If ARR multiples apply, Anthropic can look like a relative bargain vs. Series A/B pricing
- •Key risk shifts from “will it exist?” to “what’s the right multiple?”
- 4:56 – 8:37
Why raise $30B if demand is overwhelming? Balance-sheet war and compute commitments
Rory argues the raise is rational because it’s small dilution for massive de-risking amid extraordinary burn and compute commitments. The conversation turns to how frequently Anthropic may need to raise and how CapEx needs define the growth narrative.
- •Raising $30B on $900B implies only ~3% dilution to fund another year of burn
- •Compute commitments (gigawatts) imply staggering implied capital needs
- •Hyperscalers may fund CapEx, but AI labs still need fortress balance sheets
- •A slowdown in CapEx need can signal slowing growth—raising is part of the story
- 8:37 – 14:25
The true cost of AI tokens: Salesforce’s $300M spend and the token-to-wages conversion
Using Benioff’s disclosure, they benchmark token spend per engineer and debate whether current usage supports trillion-dollar revenue projections for model companies. Rory proposes a core heuristic: tokens must consume a meaningful share of knowledge-worker and engineering payroll to justify the largest valuations.
- •Salesforce token spend translates to roughly $15–$20K per developer per year
- •Token spend is “small per engineer” yet massive as a new vendor line item
- •To justify projected scale, tokens may need to reach ~20% of engineering payroll
- •If spend plateaus, AI-lab TAM assumptions may be overstated; if it rises, labor displacement accelerates
- 14:25 – 17:58
Agent economics and the bear case: models get better, workflows get cheaper
Jason shares examples from SaaStr and Klaviyo suggesting agentic systems can be run far more cheaply than people assume. This introduces a counterpoint: efficiency gains may reduce token consumption, limiting revenue upside for model vendors even as adoption grows.
- •Jason’s team runs many agents for relatively low monthly spend
- •Klaviyo’s CEO claims agentic operations can be cost-efficient with good harnesses
- •Per-agent costs can be surprisingly low (hundreds, not thousands) in some setups
- •Bear case: ubiquitous usage but lower-than-expected token intensity due to efficiency
- 17:58 – 23:12
Public SaaS rebounds—Figma, Datadog, and the reality of never returning to 2021 multiples
They interpret strong quarters as evidence of re-acceleration but emphasize that the era of 2021-style valuation exuberance is gone. The discussion reframes success as becoming “good normal companies” priced on revenue growth and cash flow rather than pure future narrative.
- •Datadog posts a standout quarter and hits major revenue/ARR milestones
- •Figma re-accelerates with strong NDR; Atlassian and others show renewed momentum
- •2021 multiples are described as a once-in-a-lifetime ‘prospects’ era for SaaS
- •AI captures investor attention, leaving SaaS to win on fundamentals
- 23:12 – 27:26
Figma’s AI strategy: design-to-production gap and threats from vibe-coding tools
Jason revisits what he thinks Figma missed (shipping faster into production workflows) while acknowledging it benefits from a broader ‘software explosion.’ Rory highlights a strategic risk: users could bypass Figma by prototyping directly in tools like Lovable and similar systems.
- •Figma benefits from increased software creation even if it lagged on ‘vibe coding’
- •New AI features can enhance and iterate on designs at scale
- •Big missed opportunity: one-click path from Figma design to production prototype
- •Emerging tools (e.g., Lovable/Replit-style flows) threaten to route around Figma
- 27:26 – 38:51
Wix and Squarespace under pressure: terminal narratives, Shopify’s squeeze, and buyback mistakes
The panel argues web builders face a two-pronged attack: vibe-coded alternatives and Shopify’s dominance in commerce. They also dissect why buybacks can backfire when businesses are deteriorating, and why cash flexibility may matter more during platform transitions.
- •Core web-builder growth looks challenged; markets price in terminal decline
- •Shopify’s expansion squeezed low-end commerce competitors
- •Vibe coding makes it easier for non-tech-phobic users to build custom sites fast
- •Buybacks may appease activists but can destroy flexibility if fundamentals worsen
- 38:51 – 48:32
Compute starvation and the infrastructure boom: Nebius/CoreWeave, CapEx risk, and scarcity as a moat
Nebius’s explosive growth sparks a debate on whether we’re in a sustainable compute shortage or a cyclical bubble. Rory argues permitting and data-center build friction may ironically prevent catastrophic oversupply, while Jason notes almost everything outside ‘old software’ is inflating.
- •Infrastructure firms win if compute remains scarce; lose if it commoditizes
- •Permitting/build delays may prevent extreme oversupply and stabilize economics
- •Hyperscaler + model-provider CapEx drives semis, networking, power, and infra stocks
- •Venture question: is it too late (or too capital-intensive) to start new compute providers?
- 48:32 – 52:04
Cerebras’ blockbuster IPO: what it signals—and what it doesn’t—for the IPO window
Cerebras’ first-day performance is treated as an ‘N of one’ driven by perfect positioning: semis, inference, and marquee AI customers. They debate whether this meaningfully reopens IPO markets for smaller companies or mainly helps the very best, most narrative-rich names.
- •Cerebras prices up and pops strongly on Day 1
- •Demand reflects hunger for public exposure to AI infrastructure and lab-adjacent growth
- •Rory calls it exceptional timing: product maturity meets demand spike + key customer proof
- •Base-rate caution: buying after an IPO pop has historically negative expected returns
- 52:04 – 1:02:29
SpaceX’s mega-IPO setup: S-1 limitations, AI acquisitions, and meme-versus-fundamentals trading
They anticipate SpaceX’s filing and argue the S-1 may be backward-looking relative to recent AI-driven moves, forcing the real story into the roadshow. The group debates how retail momentum could drive price action at trillion-dollar scale, while Rory warns that valuation and information gaps can still bite.
- •S-1 may not fully capture recent AI-related deal activity and financial impact
- •Market appetite for ‘excitement’ is high; cash-flow focus would change the tone
- •Retail could be a major force, but trillion-dollar market caps limit ‘GameStop-style’ multiples
- •Historical analogy: even iconic IPOs (e.g., Facebook) can trade down if priced too aggressively
- 1:02:29 – 1:11:04
YC’s ‘token for equity’ offer: how OpenAI’s $2M credits could reshape startup pricing and go-to-market
Sam Altman’s offer to YC startups is framed as a strategic move to win developer mindshare and distribution. They discuss how non-transferable credits still function like capital, potentially anchoring higher valuations, shrinking round sizes, and enabling aggressive freemium/token-as-marketing strategies.
- •$2M in tokens can meaningfully de-risk early-stage execution even if non-transferable
- •Optics and anchoring effects can push valuations upward and reduce VC ownership
- •Tokens can become ‘marketing budget’ for freemium and rapid iteration
- •Conclusion implied: OpenAI likely has surplus capacity if it can underwrite this at scale
- 1:11:04 – 1:27:16
Elon vs. OpenAI aftermath and the looming political backlash: layoffs, resentment, and calls for ‘re-inflation’ hiring
They revisit the Musk/OpenAI legal outcome, then widen to the broader reputational and political risks facing AI leaders. As layoffs mount across tech, Rory argues the public is turning and that industry messaging has been politically reckless; Jason suggests companies may need proactive social policies to avoid unrest.
- •Musk/OpenAI dispute seen as driven by leverage and disruption more than legal merit
- •Complex governance/financial structures create vulnerability to scrutiny and attacks
- •Public sentiment shifts from early excitement to booing AI leaders amid job fears
- •Mass layoffs plus rising costs (e.g., power) amplify political risk; debate over whether AI impacts are overstated or imminent
