Speaker
Nikesh Arora
Appearances over time
2 episodes
Episodes
2
20VC: Airtable Sold for $1.285BN | Leo Achenbrenner's Situational Awareness Blows Up | Moonshot AI Raises $3.5B at $35B | Anthropic Model Breaches Three Companies' Security | Big Tech Earnings: Why Palantir Beat The Rest
Nikesh Arora: Mythos is Real, Analytical SaaS is Dead, and Google can be a $10T company
Podcasts
Quotes & moments
Claude's Mythos model found vulnerabilities in Palo Alto's codebase in 6 weeks that would have normally taken 5 to 7 years to discover through conventional methods.
Running Claude's Mythos model on Palo Alto's codebase for 6 weeks cost only a few million dollars in compute, with costs expected to fall further.
Nikesh Arora noted the average time to patch a zero-day vulnerability found in the wild is 55 days, while AI models can find and exploit vulnerabilities in split seconds.
Nikesh Arora estimates Mythos-level AI vulnerability-finding capabilities will be available in open-source models within 3 months.
According to Nikesh Arora, 89% of cybersecurity breaches occur not from sophisticated exploits but from stolen username and password credentials.
The average enterprise time to detect and respond to a cybersecurity breach is 4 days; Nikesh Arora's team currently achieves 1 minute with machine learning for 1,200 customers.
Claude's Mythos model had a 30% false positive rate when used for vulnerability detection — great for offense, but problematic for defense.
Palo Alto Networks found 14,000 vulnerabilities in open-source packages over 14 weeks of testing, illustrating how pervasive security gaps are in enterprise infrastructure.
Nikesh Arora argues enterprises will need to store 10 times the data they currently hold over the next 3 years to properly defend against AI-powered cyber threats.
Nikesh Arora believes Google is underrated and will become the first $10 trillion company in our lifetime, citing its full-stack AI assets and massive sales force.
Nikesh Arora estimated that $1 trillion in CapEx has been committed across major hyperscalers for the next year, with Big Tech cloud revenues validating the spend.
Palo Alto Networks acquired a $25 billion identity security company that closed 3 months prior to the episode, pivoting toward agentic security infrastructure.
A CEO of a leading AI model company told Nikesh Arora that the entire weights of their newest frontier model fit on a single USB stick, distillable in under 48 hours.
Palo Alto's strategy of buying product companies and integrating them into its go-to-market engine drove the company's valuation from $17 billion to north of $150 billion.
Palo Alto Networks' largest acquisition was $28 billion — Nikesh Arora identified it as CyberArk — which he said is now worth north of $50 billion and validates their identity-security-for-agents thesis.
AI is supercharging attackers faster than defenders can respond. Every vendor is now showing up at CIOs' doors asking them to patch newly discovered vulnerabilities — while CIOs are simultaneously trying to find and fix their own — and open source remains an unsolved nightmare.
While analytical SaaS dies, infrastructure software is quietly undervalued. Enterprises will need to store 10 times their current data over the next three years to train and defend AI systems — making databases, data platforms, and cloud storage essential picks-and-shovels plays.
Enterprise UIs were always just a workaround to let humans interact with data. Once agents can do that interaction directly, the entire trillion-dollar investment in enterprise UI becomes unnecessary. The system-of-work software layer must be fully reinvented in the next five years.
The real AI cyberattack threat isn't cracking power grids or defense systems — those are well-protected. The danger is the ransomware attack on a healthcare clearing house that shuts down every physician's office in the country, like Change Healthcare. Small offices with old software are the real attack surface.
AI models will commoditize into a utility layer where you buy intelligence on demand at different price-performance tiers. The profit pools are in the application layer — which is why OpenAI is pushing Codex and Anthropic is pushing Claude Code. They know that's where the money will be.
A major AI model CEO told Nikesh Arora that the full weights of their newest frontier model fit on a single USB stick and can be distilled in under 48 hours. That fact alone makes a 6-month export control regime essentially meaningless — the IP walks out the door.
Mythos had a 30% false positive rate — meaning nearly 1 in 3 vulnerabilities it found didn't exist. For offense that's fine; for defense it's catastrophic. The real challenge of deploying AI in enterprise isn't getting the newest model, it's driving false positives from 20% down to 0.01%.
For years, Palo Alto bought product companies and plugged them into its sales engine — doubling revenue per customer at negligible incremental cost. That playbook ran to $150 billion. The next phase is using AI to drive operating margins far above industry norms, making any acquisition accretive regardless of the target.
The fastest path to revenue in AI is replacement — taking existing software budget and replacing the incumbent with something better. You don't have to create demand; it's already there. Replacement TAMs are beautiful because the customer already knows they need to pay.
Palo Alto Networks ran Claude's Mythos model against its own codebase for 6 weeks and found vulnerabilities that would have taken 5 to 7 years using conventional methods. The cost was in the low millions — and the capability will be in the wild within 3 months.
Any SaaS company whose core value proposition is collecting and analyzing data for you is finished. You can now run an LLM directly against your raw data and get better, faster answers — no middleware needed. The entire incremental-module business model is gone.
Google is underrated. It has the models, the data, and critically — the largest enterprise sales force of any hyperscaler. Most people forget that selling AI at scale requires an army of salespeople, and Google already has that army. That's why it wins.
The AI race isn't about model quality anymore — it's about who can secure land, permits, energy, and compute. Chicken manure methane companies and small modular reactors are both raising billions because anyone who can produce energy in any form is now a premium asset.
Palo Alto Networks is building organizational context — every customer case, every resolution — into vector databases so it can swap any model underneath and maintain performance. Nikesh Arora's thesis: in 5 years, domain context will be as important as model intelligence, and that context is yours to own.
DroneDeploy sold to Procore for $900 million after never raising at a price above what it sold for — a rare feat of capital discipline that made the exit straightforward. Being on the right side of physical-world AI trends helped too: drones and robots are only starting to explode.
Analysis
What they talk about
- Technology 70%
- Business 25%
- Society & Culture 5%
Connections
Shows they appear on and people they share episodes with. Drag to explore.
