a16zFrom the Dot-Com Crash to the AI Era: How Builders Survive Waves of Disruption
CHAPTERS
- 0:00 – 1:31
VMware’s two-decade arc: from disruptor to disrupted
Raghu frames VMware’s 20-year story as two distinct decades: first disrupting the market, then being disrupted by new waves. He introduces the idea that big infra companies must repeatedly adapt as the industry’s “rules of the game” change.
- •VMware’s first decade: rapid growth through disruption; second decade: defending against new disruptors
- •How market transitions force incumbents to rethink identity and operating model
- •Why infrastructure layers are especially prone to wave-driven reshuffles
- 1:31 – 2:55
“Weapons of mass disruption”: abstraction, usage models, new users, business models
Raghu outlines the recurring patterns that create outsized disruption in infrastructure. He explains why new abstractions and new usage models can quickly become the center of gravity for ecosystems—and why those shifts are hard to reverse once behaviors change.
- •Disruption drivers: new abstraction layers, new usage models, new user classes, and new business models
- •VMware’s core disruption: software virtualization plus software-first business model
- •Behavioral lock-in: once practitioner workflows change, displacement becomes extremely difficult
- 2:55 – 4:52
Cloud and containers as VMware’s hardest transition tests
The conversation turns to the two biggest threats VMware faced: cloud and the container/Kubernetes wave. Raghu emphasizes that cloud’s real disruption wasn’t just technology—it was opening infrastructure access to developers outside traditional IT channels.
- •AWS changed the business model and made infra accessible directly to developers
- •VMware lacked the playbook for developer-first engagement early on
- •Containers/Kubernetes changed packaging and abstraction, even if value capture was diffuse
- •Cloud was the most fundamental competitive disruption because it brought a new user base
- 4:52 – 6:34
Cisco’s missed cloud wave—and why big companies lose the “soul” of innovation
Jeetu candidly states Cisco missed the cloud wave and diagnoses a common scaling failure: companies optimize financial math but lose urgency and proximity to the front lines. He describes a deliberate “reset button” approach to restore founder-like velocity inside a large org.
- •Large companies drift from customer front lines and innovation speed
- •A “reset” requires cultural and leadership changes, not just new strategy decks
- •Founder’s mentality as an operating principle—across levels, not only executives
- •Balancing ex-CEO acquisition leaders with internal operators who know the machine
- 6:34 – 8:15
Operating like the world’s largest startup: 0→1, 1→100, 100→1,000
Jeetu lays out a practical scaling model: launch quickly, then scale aggressively to meaningful revenue outcomes. He contrasts large-company strengths (many experiments) with a frequent failure (not doubling down hard enough when something works).
- •Mantra: “speed with scale” and repeatable product creation cycles
- •Target cadence: zero-to-market in ~9 months; to $1B in 3–4 years (when applicable)
- •Incumbents run many experiments but often fail to commit to the winners
- •Route-to-market advantage only works if products are designed to leverage it
- 8:15 – 10:20
Go-to-market reality: disruption rarely comes from your best customers
Martin and Raghu dig into why incumbents struggle when the new wave emerges outside their core enterprise accounts. Raghu explains two classic responses—ring-fence a team or acquire—and how the right choice depends on adjacency and buyer/practitioner differences.
- •Deep relationships with top customers can blind incumbents to disruptive footholds
- •Two playbooks: fence off a new team or buy capabilities (M&A)
- •Adjacency matters: same user can use existing sales force; different user often needs separation
- •Example: Nicira/networking required a different approach than closer adjacencies
- 10:20 – 12:34
Organic vs. inorganic innovation: vSAN and Nicira lessons + the “10X” bar
Raghu explains when building internally can work versus when acquisitions are the pragmatic path. The discussion highlights the importance of matching the go-to-market motion to the actual buyer/practitioner and holding products to a true step-function improvement threshold.
- •vSAN succeeded after shifting from ‘storage buyer’ to expanding ‘compute buyer’ scope
- •Winning established categories often requires being meaningfully (10X) better, not slightly better
- •Large orgs risk self-deception: “10X” is often only 15% in reality
- •Nicira’s value was a new operational model (programmable/virtualized networking), not just incremental performance
- 12:34 – 13:58
Competing in brownfield markets: insertion points, coexistence, and open ecosystems
Jeetu argues most markets are brownfield, not greenfield, so entrants must find a specific insertion point rather than attempting a full-platform replacement. He stresses the counterintuitive need to coexist first, and to embrace ecosystem openness even when it’s uncomfortable.
- •Define clear market insertion points instead of “boil the ocean” platform plays
- •In brownfield, you often must coexist before you can displace incumbents
- •Ecosystem openness becomes a competitive requirement, not a nice-to-have
- •Large companies often resist this because it challenges traditional control instincts
- 13:58 – 18:05
Structuring disruption inside a giant: two-pizza teams, air cover, overlays, and ICP expansion
Jeetu details how Cisco incubates new initiatives: start with a small, protected team and avoid being crushed by organizational “antibodies.” He explains why version-one products need tight ICP focus, often supported by overlay sales, before broad field enablement is feasible.
- •Start with a two-pizza team with top-level air cover and high agency
- •Antibodies will argue to fold it into the core too early—protection is essential
- •Use overlay sales and a narrowly defined ICP to get early adoption and repeatability
- •Expand ICP stepwise only after clear product-market fit and saturation signals
- •Clarify ICP as “ideal practitioner profile,” not just big-name enterprise logos
- 18:05 – 19:28
Storytelling as execution: “the story is the strategy”
Jeetu shares advice from a board member: don’t delegate storytelling—one consistent narrative is required to align a massive organization. Raghu extends this further, arguing the story isn’t just messaging; it becomes the strategy that galvanizes tens of thousands of people.
- •Large orgs lose momentum when the narrative fragments across teams
- •A single owner telling the end-to-end story preserves clarity and control
- •Apple as a benchmark for narrative-driven alignment
- •“The story is the strategy”: humans align to narratives more than bullet points
- 19:28 – 22:36
The consumer/prosumer AI wave: breaking old enterprise selling assumptions
Martin introduces AI’s unusual entry path through consumers and prosumers, even for enterprise-centric companies. Raghu and Jeetu argue that AI is big enough to demand first-principles thinking and that traditional “sell-to-IT” chains are increasingly broken.
- •AI adoption often starts with individuals using credit cards, not centralized IT procurement
- •Past ‘AI’ analogies mislead; OpenAI broke prior rules and changed expectations
- •AI further disrupts the seller→IT buyer chain that SaaS already weakened
- •Product teams must design for direct end-user reach and new usage patterns
- 22:36 – 24:55
Cisco’s AI-era thesis: networking, security, and data as foundational constraints
Jeetu explains Cisco’s positioning as “critical infrastructure for the AI era,” focusing on where AI is constrained today. He highlights network latency to GPUs, sustained inference demand from agents, and the need to secure AI systems—plus Splunk/data as a platform pillar.
- •AI constraints: power, compute, and especially networking (GPU idle time is costly)
- •Training requires low-latency, high-performance, energy-efficient packet flow
- •Agent-driven inference creates persistent (not just spiky) infrastructure demand
- •Security becomes foundational for AI systems; data/observability strengthened via Splunk
- •Infrastructure winners ride workload/usage model shifts
- 24:55 – 35:41
Regaining innovation momentum: perception, product-led culture, and truth-seeking
Jeetu describes Cisco’s recent innovation surge—and the harder challenge of changing market perception at scale. The conversation expands into leadership mechanics: product as the company’s “soul,” moving from sales-led to product-led, and creating truth-seeking cultures where the best idea wins.
- •Innovation output can rebound faster than customer perception; storytelling must scale
- •Product-centric leadership: market pull reduces reliance on enablement arguments
- •Truth-seeking is harder in large companies; avoid “versions of the truth”
- •Use direct, binary language about failures and existential risks when needed
- •Create safe, title-free review spaces (design/code reviews) so “best idea wins,” not rank
- 35:41 – 37:29
AI rebuilds infrastructure end-to-end: power-to-tokens, new bottlenecks, 10X+ market shifts
Raghu argues AI fundamentally reshapes every infrastructure layer, following changing workloads as prior waves did (e.g., browsers → internet infra). He frames the end-to-end system as “power to tokens,” implying transformations from energy through compute, memory, networking, and storage.
- •Infrastructure follows workload shifts; AI is a step-change similar to early internet eras
- •Bottlenecks move across the stack: memory, networking, storage, and power
- •“Power to tokens” as the guiding abstraction for AI-era systems thinking
- •AI expands markets because it replaces labor and boosts GDP-scale productivity potential
- 37:29 – 40:36
Vertical integration vs. horizontal openness: partnering with competitors in the AI era
Jeetu emphasizes vertical integration (silicon to security to data) while staying “horizontally friendly” to ecosystems—sometimes even integrating deeply with competitors. Raghu and Martin debate the historical swing from disaggregation to reintegration, noting AI’s structure is still evolving quickly.
- •Cisco investing in vertically integrated capabilities (ASICs, platforms, security, data)
- •Models may become smaller and more bespoke; integration can improve performance and safety
- •Ecosystem openness is mandatory—sometimes requiring partnership with major competitors
- •Rule of thumb: if someone has >20% share and you don’t integrate, you exclude yourself
- •AI wave may start horizontal and become more vertical over time; outcome not settled
- 40:36 – 44:32
Founder-grade advice: timing, market, team, product, brand, distribution—and running toward disruption
In closing, Jeetu offers a six-part founder framework with timing as the most important factor, urging builders to ride AI tailwinds rather than fight them. Raghu adds that brand and distribution look very different today, and Martin emphasizes that disruption usually creates more opportunity than danger if you run toward it.
- •Six-part startup formula: timing, market, team, product, brand, distribution (in that order)
- •Attack large markets stepwise; ideally create new TAMs rather than chase old ones
- •Product must earn love, adoption/retention, and commercial relevance
- •Brand requires a consistent message; modern channels reward authenticity over scripting
- •Disruption is an opportunity—“run towards the fire” rather than retreat
