The Twenty Minute VCThe Impact of H1B Visas on Startups in the US & NVIDIA Invests $100BN Into OpenAI
CHAPTERS
- 0:00 – 5:45
NVIDIA’s $100B into OpenAI: capital flywheel, scaling laws, and “no timeout” financing
The group unpacks the headline NVIDIA investment and what it implies: an unusually large, reinforcing loop where capital and compute enable bigger bets. They debate whether scaling laws are stalling or whether this financing ensures OpenAI can keep pushing until reality forces a stop.
- •The apparent “money-printing loop”: NVIDIA equity → OpenAI spend → more chips and infrastructure
- •Rory’s framing: the deal ensures OpenAI can keep doubling down until returns clearly fail
- •Harry challenges the premise by citing GPT-5 efficiency focus as a sign scaling may be slowing
- •Jason argues Sam Altman’s stated compute needs (3 orders of magnitude) signal continued ambition
- •Core takeaway: markets are letting the experiment run rather than calling time early
- 5:45 – 9:54
Anthropic vs OpenAI after the deal: chips, capital, and monopoly optics
Conversation shifts to whether Anthropic is newly disadvantaged. They argue Anthropic likely isn’t capital-constrained, but GPU access and narrative momentum matter—along with OpenAI’s incentives to avoid being viewed as an outright monopoly.
- •Anthropic can likely raise capital easily; the differentiator may be preferential GPU access
- •Jason: OpenAI and NVIDIA must manage monopoly perception and regulatory risk
- •Consumer chatbot dominance: ChatGPT as “Standard Oil of tech” vs competition from Gemini/Perplexity
- •NVIDIA’s own strategic tension: selling to customers who may build competing chips
- •Speculation: deal terms might implicitly discourage OpenAI’s in-house chip ambitions
- 9:54 – 13:28
NVIDIA’s concentration risk: $4T value, six customers, and a CapEx boom far ahead of revenue
They dig into NVIDIA’s customer concentration and how AI CapEx has run far ahead of current AI revenue. The group highlights why suppliers and adjacent services (like data labeling) are thriving: the big buyers are racing, not optimizing costs.
- •NVIDIA revenue concentration: a handful of customers drive the majority of sales
- •AI CapEx boom vs AI revenue reality: massive infrastructure spend relative to market revenue
- •Data labeling providers see similar concentration and benefit from buyer urgency
- •Cost optimization is secondary when buyers are in an arms race to win AI dominance
- •Dynamic: concentration is risky, but the top buyers currently show no sign of slowing
- 13:28 – 16:44
Déjà vu of 1999: vendor-financing echoes, unlimited belief, and balance-sheet blind spots
Rory compares the moment to late-1990s exuberance—less a perfect repeat, more a rhyme—citing vendor-financing analogies. They stress how booms collapse quickly when expectations overshoot timelines, and how in bull markets people ignore balance sheets until conditions tighten.
- •Rory: feels more like ’99 than any time in 20 years; vendor-financing parallels (Nortel/Lucent)
- •Jason: today differs because there’s vastly more capital and mutual “guaranteeing” of capacity
- •NVIDIA free cash flow expansion underscores how much internal capital exists to reinvest
- •Bull markets overweight income statements; downturns suddenly make cash and solvency paramount
- •The “time element” matters: truths can be right long-term yet disastrous short-term
- 16:44 – 19:54
Buybacks, dilution, and capital allocation: why companies repurchase at peaks
The panel debates NVIDIA’s aggressive buybacks and the common practice of offsetting RSU dilution with repurchases. Rory argues it’s irrational to tie buybacks to dilution rather than valuation, while Jason notes the practice keeps stock-based compensation economically honest.
- •NVIDIA buybacks: large repurchase authorization despite cyclical uncertainty
- •Jason: many companies buy back to offset RSU/option dilution and maintain EPS
- •Rory: buy back when stock is cheap—linking to dilution is “dumb as rocks”
- •Macro pattern: corporations tend to repurchase at peaks and stop when shares are cheap
- •Larry Ellison cited as a rare example of exceptional timing and capital redeployment
- 19:54 – 22:25
Are we at the top? Personal asset allocation, cash buffers, and froth signals
Harry asks how to think about selling in frothy markets; Jason admits he’s fully invested with no cash. Rory offers a framework: short-term valuation predicts little, but long-term starting valuations matter—and holding cash can be the price of sleeping at night.
- •Jason: 0% cash; recalls the pain of having to sell at large losses in 2008
- •Rory: valuation weakly predicts 1-year returns but strongly informs 10-year outcomes
- •Asset allocation is about risk tolerance, not just maximizing returns in a bull market
- •Froth signal: LPs bragging about returns publicly (echoes of 2021 behavior)
- •Practical stance: accept some underperformance to maintain liquidity and resilience
- 22:25 – 29:50
Is ‘triple, triple, double, double’ dead? VC concentration and the unpredictable middle tier
They address claims that classic SaaS growth benchmarks no longer guarantee funding and discuss the increasing concentration of venture dollars. Rory argues the core venture engine is intact; what’s changed is the rise of ultra-late-stage private investing lumped into “VC.”
- •Data point: large share of VC dollars going to a small number of companies
- •Rory: early-stage volume is relatively stable; the ‘new’ business is ultra-late-stage private public-style rounds
- •Jason: top-tier ‘super hot’ is obvious; the murky zone is just below S-tier where opinions diverge
- •Benchmarks still matter—outlier growth gets funded—but meeting bars takes more effort than before
- •Category bias matters: “unloved” markets face higher skepticism despite strong growth metrics
- 29:50 – 34:29
Exits still matter: Iconiq’s wins, the ‘asterisk’ problem, and why 7X is still 7X
Jason questions whether $1B–$8B outcomes “count” in the era of mega-winners like OpenAI. Rory defends the fundamentals: multiples are multiples, money is fungible, and mid-sized IPOs can still generate excellent returns—though fund size changes what moves the needle.
- •Iconiq examples (e.g., Netskope, DX) spark debate on meaningful outcomes vs mega-deals
- •Jason: in big-return funds, smaller wins can feel like ‘asterisks’ in DPI tables
- •Rory: OpenAI’s blended multiples may resemble other strong outcomes; absolute sums differ
- •Fund size and ability to deploy capital drive the need for fewer, larger winners
- •Key tension: significance and status vs actual financial returns and liquidity
- 34:29 – 42:00
When to sell a monster winner: risk aversion, fund dynamics, and the psychology of holding
They explore whether investors should sell stakes in OpenAI-like winners and why few do. The conversation spans taxes, incentives, and portfolio theory—plus the idea that early success increases willingness to take risk, which can compound future success.
- •Jason: emotionally hard to sell when a position might double with no effort and taxes are deferred
- •Rory: decision requires both fair-value view and overlay of institutional/personal constraints
- •Discussion of ‘fund returner’ pressure: converting 1X to 2X is huge for outcomes and carry
- •Book reference (The Missing Billionaires): wealth often lost through bad bet sizing, not stock picking
- •Success begets success: risk tolerance and referral effects increase future access and conviction
- 42:00 – 54:11
Navan’s S-1: timing the IPO window, comps (Brex/Ramp), and concentration in a single bet
Navan’s filing triggers a discussion on IPO strategy and whether it’s smart to go out before adjacent peers. They examine the business mix (travel booking vs payments/card), growth and retention metrics, and why IPO timing may matter more than reaching profitability first.
- •Navan metrics: ~$613M revenue, ~32% growth, ~110% NDR; notable but not ‘best in class’
- •Debate: Navan’s category differs from Brex/Ramp (travel booking vs card/payments), but markets may comp them anyway
- •IPO game theory: better to be early than last if peers could set the benchmark
- •Profitability trade-off: slowing OpEx suggests push toward profit, but timeline may be long
- •Concentration investing: benefits and stomach required; distinction between GP-level vs fund-level risk
- 54:11 – 1:02:17
IPO liquidity mechanics: lockups, secondaries, distributions, and board trading constraints
Rory explains how liquidity actually happens post-IPO: typical lockups, occasional early releases, structured secondaries, and gradual selling over many months. They also cover why “paper billionaire” headlines are misleading and how board roles constrain trading windows while offering informational advantages.
- •Lockup basics: typically ~6 months; sometimes waived or performance-triggered
- •Liquidity paths: registered secondaries during lockup vs open-market selling after expiry
- •Why IPO ‘pops’ can help enable secondaries; down-trading makes them hard
- •Board dynamics: restricted windows; holding isn’t illegal even with inside knowledge, but selling is constrained
- •Typical exit timeline for large holders: ~12–18 months (sometimes longer) post-IPO
- 1:02:17 – 1:07:41
H-1B visa fee hike: pragmatic workarounds (O-1), startup impact, and policy trade-offs
They discuss the announced higher cost for new H-1B visas and how it could affect early-stage hiring. While agreeing it’s negative at the margin, they argue founders often find alternatives (especially O-1), big tech can pay up, and the broader issue is that skilled immigration gets swept into wider political tensions.
- •Consensus: skilled immigration has been strongly positive for US tech; changes are directionally harmful
- •Materiality debated: likely modest immediate impact if changes don’t broaden further
- •Jason’s experience: early teams relied on H-1B talent; such hires can be existential for startups
- •Workarounds: O-1 pathways often used, especially for founders, though stressful and complex
- •Policy critique: dollar thresholds are a crude proxy for skills-based immigration systems
- 1:07:41 – 1:25:48
Notion at $500M ARR and the hangover from 2021: re-acceleration, pricing on fundamentals, and ‘founder-friendly’ myths
Closing topics cover Notion’s scale and re-acceleration, how mature SaaS companies can regain growth by leaning into AI, and why 2021 valuations must be mentally written off. They also argue that “founder friendly” has become performative—real founder support shows up in hard moments and in concrete actions, not slogans.
- •Notion’s re-acceleration at scale: 30–40% growth at $500M ARR is meaningfully impressive
- •Public market reality: older stories get priced on fundamentals (multiples, cash flow), not sizzle
- •2021 valuation overhang: they advocate marking down and moving on rather than anchoring
- •AI deal frenzy: shrinking diligence windows, term sheets used to ‘lock’ deals then pulled
- •‘Founder-friendly’ vs ‘founder-honest’: true support is shown in crises, recruiting help, and staying engaged when times are tough
