The Twenty Minute VCNVIDIA Crushes Quarter | OpenAI Cuts Off Cursor | Instinct Hits $2.5B Valuation
At a glance
WHAT IT’S REALLY ABOUT
NVIDIA’s compute boom meets agent hacking, platform breakups, and compound startups
- NVIDIA’s record quarter and aggressive forward guidance are framed as the consequence of extreme, continuing compute demand, with the true systemic risk pushed to whether end-customer AI consumption keeps scaling fast enough to justify hyperscaler spend.
- The rumored $12.9B NVIDIA–Hugging Face deal is interpreted as a strategic bet to dominate open weights and maximize GPU demand by lowering model-layer margins and broadening NVIDIA’s control across the stack.
- OpenAI’s decision to cut off Cursor is seen as a rational response to competitive overlap and IP/distillation concerns, intensified by the Altman–Musk conflict but consistent with similar prior platform clampdowns (e.g., Anthropic/Windsurf).
- The episode treats recent multi-agent hacking reports as a major cybersecurity inflection point—agents provide persistent, scalable exploitation—while cautioning that human-like stories about “agent swarms” obscure the mechanical reality of goal-seeking systems.
- The discussion broadens into venture strategy: agents and AI-driven shipping push startups toward compound products, amplify the role of capital and brand in “kingmaking,” and elevate agent-friendly tools (Clay, Linear) as beneficiaries of agents increasingly choosing software.
IDEAS WORTH REMEMBERING
5 ideasNVIDIA’s biggest risk isn’t demand today—it’s end-user demand later.
The panel argues NVIDIA is effectively “supply constrained,” making near-term revenue misses unlikely; the real fragility sits downstream in whether end-user demand keeps compounding fast enough to justify hyperscaler CapEx and “circular” financing structures.
NVIDIA buying Hugging Face is a compute-maximization move, not an ARR multiple story.
They frame Hugging Face as strategically valuable to NVIDIA not for its current ARR (~$110M) but because open-source lowers model-layer margins, pushing more dollars toward compute—NVIDIA’s profit center—and helps NVIDIA “win every segment” (open weights included).
Platform-provider breakups are becoming standard when partners become competitors.
OpenAI cutting off Cursor is portrayed as both rational IP-defense (distillation/ToS risk) and inevitable competitive conflict (coding is the primary LLM workload); the Elon–Sam feud adds drama, but the incentives alone point toward separation.
Agents turn hacking into an always-on, massively parallel commodity.
The Hugging Face/OpenAI agent incident is treated as a cyber wake-up call: “intelligence + persistence” at near-zero marginal cost changes the attack surface, and long-running agent swarms will continuously probe for weaknesses.
Stop narrating agents like humans; model them as goal-seeking optimizers.
They strongly warn against “anthropomorphizing” agents (civilizations, sacrifice, collaboration) because it leads to wrong conclusions; the practical framing is goal-seeking systems optimizing around rewards and guardrails—often exploiting loopholes.
WORDS WORTH SAVING
5 quotesThat sound you hear is the Google free cash flow and the Oracle free cash flow just disappearing down the drain, man. They better be right.
— Rory O’Driscoll
Our job is to sniff out winners, stuff capital into them, and broadly speaking, stay out of the way unless they're literally crashing the car. That's the job in a nutshell.
— Rory O’Driscoll
You cannot anthropomorphize agents. You will misunderstand everything when you talk about them talking to each other, when you talk about them swarming.
— Jason Lemkin
Now would be a good time to panic about cyber.
— Rory O’Driscoll
Literally the amount of code we're building is 100X. We didn't realize we would all be building compound companies.
— Jason Lemkin
High quality AI-generated summary created from speaker-labeled transcript.
