The Twenty Minute VCSridhar Ramaswamy, CEO @Snowflake: Deepseek is Not a Threat to OpenAI & OpenAI Beats Anthropic|E1258
CHAPTERS
- 0:00 – 1:06
OpenAI vs DeepSeek: why the product layer beats the model layer
Sridhar argues that DeepSeek isn’t a real consumer threat to ChatGPT because users stick with product experiences, not raw models. He explains why building startups “on top of OpenAI” is risky due to blurred lines between infrastructure and applications.
- •Infrastructure vs application boundaries in AI are increasingly blurry
- •Consumers don’t switch models; they stick with integrated products (ChatGPT)
- •Anthropic’s weaker consumer traction as a model-first approach
- •Platform providers can copy fast-growing apps (coding, legal, etc.)
- •Why this creates anxiety for startups depending on foundation-model vendors
- 1:06 – 2:09
From PhD aspirations to CEO: “aspire big, take little steps”
Harry prompts Sridhar’s personal journey: he didn’t set out to be a CEO and initially preferred an academic path. The shift came when he moved from research into software engineering and advanced through incremental steps.
- •Never expected to become a CEO during college/PhD years
- •Early dream: being a professor; later lost interest in research
- •Transition into software engineering as a practical pivot
- •Career philosophy: take small steps while aiming high
- •Right place/right time can amplify prepared people
- 2:09 – 4:07
Advice to graduates entering an AI-shaped workforce
Sridhar’s guidance centers on passion plus market demand, then staying nimble amid rapid change. He frames adaptability and drive as the durable traits, regardless of role seniority.
- •Choose work you love that society truly values (not just ‘noble’ on paper)
- •Embrace change rather than deny it (AI is real and accelerating)
- •Develop drive + malleability as a career advantage
- •Stay open to new opportunities and shifting skill needs
- •These traits matter even when hiring senior executives
- 4:07 – 8:29
Will AI hollow out software engineering? Impact on knowledge work
He predicts AI will transform many knowledge professions, including software engineering, but avoids declaring that the profession will disappear. Using analogies (journalism/music post-internet), he explains how technology can narrow or reshape fields while still expanding total opportunity.
- •AI as a ‘translation layer’ across structured and unstructured knowledge
- •Most knowledge work will be materially impacted, not just coding
- •Historical analogy: internet narrowed some professions (journalism, music)
- •Hard to predict whether engineering becomes ‘thinned out’ or redefined
- •Opportunity remains: software + AI will expand into more domains
- 8:29 – 10:50
Intensity as a leader: setting expectations without burning the team
Harry asks how intense CEOs bring teams along. Sridhar emphasizes clarity on mission, the fleeting nature of opportunity, and aligning people to high expectations—while accepting it’s not for everyone.
- •Intensity works best when paired with a clear ‘big picture’ mission
- •High expectations are justified when returns/opportunity are high
- •Explicitly communicate what success demands at the company
- •Not everyone wants (or should) operate at that level—and that’s okay
- •Snowflake’s ambition: be the data engine for every enterprise
- 10:50 – 13:00
Scaling leaders and the ‘doubling your team’ reinvention problem
Sridhar explains why leaders sometimes fail to scale: incentives shift, life circumstances change, and what made someone great at one stage can block them at the next. He describes a duty to give people chances, while staying decisive if fit doesn’t improve.
- •People change—wealth and time horizons can reduce drive
- •When teams double, prior strengths can become new constraints
- •Scaling requires reinvention, not just doing ‘more of the same’
- •Leaders must provide opportunities but also make hard calls
- •Why Google’s IPO-era wealth shifts affected retention and leadership depth
- 13:00 – 16:59
How to have hard conversations—and why demotion can be real leadership
He shares a practical mindset for difficult conversations: conflicts don’t resolve themselves, delaying makes it worse, and avoidance harms the other person. On demotions, he argues fast-growing contexts can outgrow a role, and reshaping scope can help someone succeed rather than firing reflexively.
- •Delay amplifies conflict; address issues early
- •Hard feedback is often a service, not punishment
- •Use respect, straightforwardness, and humility in delivery
- •Demotion/re-scoping can match capability to a fast-changing role size
- •Example: refocusing a leader to set them up for long-term success
- 16:59 – 18:28
Does wealth make leaders better? Risk tolerance, empathy, and constituents
Sridhar rejects the idea that richer leaders are better. He warns that wealth can produce callousness or excessive risk-taking and stresses leaders must balance employees, customers, and shareholders—especially during tough company periods.
- •Wealth can increase tolerance for risk in unhealthy ways
- •Leadership must account for multiple constituents beyond the CEO
- •Not every situation calls for ‘swinging for the fences’
- •Difficult company periods highlight real-world consequences
- •Being personally insulated financially isn’t automatically beneficial
- 18:28 – 21:58
Where durable value accrues in AI: relationships, data platforms, and product experience
Sridhar describes why value creation feels murky: big platform companies can copy new apps quickly. He argues durable advantage comes from owning customer relationships and delivering clear utility, and he positions Snowflake as a data platform accelerated by AI.
- •Big players can rapidly enter promising application categories
- •Sustainable value: customer relationships + clear delivered outcomes
- •Snowflake’s advantage: data lifecycle platform + AI as an accelerant
- •Examples of incumbents self-disrupting (e.g., Salesforce Agentforce)
- •OpenAI’s edge: product experience + massive loyal user base
- 21:58 – 24:41
Why DeepSeek isn’t a consumer ‘switching’ event—and what OpenAI would do next
Pressed on DeepSeek’s rise, Sridhar reiterates that ChatGPT’s feature-complete product reduces churn. He also notes OpenAI could incorporate external models if it improves the product, and he highlights OpenAI’s growth as extraordinary beyond pure model quality.
- •ChatGPT vs DeepSeek: users choose the overall product bundle
- •Features matter: uploads, images, code execution, integrated UX
- •OpenAI would ‘shamelessly’ use others’ models if it helps the product
- •OpenAI’s scale (hundreds of millions of users) is historically rare
- •Closed-model mystique, misdirection, and myths being challenged by open entrants
- 24:41 – 29:02
Snowflake vs giants and Databricks: defending position in a stacked ecosystem
Harry probes threats from cloud incumbents and NVIDIA, plus Databricks’ perceived lead in AI workloads. Sridhar argues product-market fit and sustained innovation matter more than money, and he differentiates classic ML maturity from the newer generative/agentic AI wave where Snowflake is investing aggressively.
- •Competing as a ‘mouse among giants’: AWS, Microsoft, Google, Oracle pressures
- •Money doesn’t automatically create a Snowflake-quality data product
- •Databricks strong in ML; gen-AI is newer and more open to leadership shifts
- •Snowflake’s AI focus areas: transformations, unstructured data, reliability to structured data
- •Agentic direction: Snowflake Intelligence as a unifying framework
- 29:02 – 33:12
Innovation under constraints: public vs private, spending discipline, and clarity
Sridhar argues innovation is mandatory regardless of being public, though public markets add scrutiny and second-order effects. He claims constraints can sharpen focus, while private companies can ‘buy’ growth, and he defends staying public for transparency, liquidity, and self-calibration.
- •Public-company scrutiny creates constraints and externalities (employees’ financial exposure)
- •Constraints can drive clarity and prevent chasing unscalable business
- •Private firms can overspend without immediate cash-flow pressure
- •Wouldn’t take Snowflake private: transparency and feedback loops matter
- •Public markets reduce the ability to ‘three-card monte’ narratives
- 33:12 – 36:46
Enterprise AI adoption: ROI skepticism, Davos learnings, and ‘utility’ as the driver
He predicts adoption may be gentler than hype suggests, but insists AI is already creating real value. Using Davos examples, he emphasizes practical utility—summarization, internal copilots over structured data, and combining structured/unstructured data for agentic workflows like underwriting.
- •Adoption curve: not pure hype; value is already real and compounding
- •If leaders see ‘no value,’ they may not be using AI well
- •Personal productivity examples: transcription + summarization at scale
- •Enterprise copilots over structured data reduce dashboard complexity
- •Davos sentiment: ‘show us utility’ and what’s possible with agentic systems
- 36:46 – 39:10
Why incumbents are suddenly moving fast: lessons from past platform shifts
Sridhar explains incumbent speed as learned behavior from multiple disruption cycles. He points to mobile as a prior shift where major tech adapted well, and argues today’s massive bets reflect both fear of disruption and the resources to invest aggressively.
- •Historical memory: nobody wants to be the next DEC/SGI story
- •Mobile shift taught incumbents to respond decisively (apps, monetization)
- •Big tech now makes ‘crazy investments’ when it spots platform shifts
- •Meta example: pivoting from AR disappointment to AI with speed
- •Incumbents have both motivation and capital to move quickly
- 39:10 – 41:12
The AI capex arms race: bubble dynamics, what’s productive vs wasteful spend
Harry asks where trillion-scale commitments end; Sridhar expects a bubble to burst but notes some bubbles build lasting infrastructure (like 90s fiber). The key uncertainty is whether spending goes into enduring assets (power/buildings) or rapidly depreciating hardware, while niches still remain for innovation.
- •Arms-race capex likely produces a bubble and eventual correction
- •‘Good bubble’ vs ‘dumb bubble’ framing (telecom fiber vs Webvan)
- •Enduring value depends on where money goes: infrastructure vs depreciating hardware
- •Innovation still possible: incumbents can’t cover every workflow
- •Investing thesis: find defensible niches (e.g., Harvey in legal)
- 41:12 – 44:36
Snowflake growth playbook: widening the product aperture, selective M&A, and new revenue engines
Sridhar responds to growth pressure by describing a broader Snowflake mandate: ingestion, data engineering, analytics, ML, and AI-driven end-user access. He supports targeted acquisitions that strengthen product-led innovation, and he predicts AI plus customer-built data applications on Snowflake will become major future revenue streams.
- •Growth strategy: expand from ‘warehouse + analytics’ to full data lifecycle platform
- •Product-led innovation over PE-style rollups
- •Selective acquisitions that fit the platform (example: small strategic buys)
- •Future revenue: AI capabilities as a major line item
- •Big unlock: customers building data apps on Snowflake—moving from cost center to top-line enabler
- 44:36 – 50:28
Model landscape and Google distribution lessons: entry points, defaults, and fragmentation
Sridhar compares AI’s future to Google Search’s rise: distribution defaults and a central entry point enabled Google to conquer verticals. He believes ChatGPT is becoming that consumer entry point, while enterprise remains fragmented and specialized, making a single ‘winner-take-all’ less certain.
- •Google’s dominance was partly engineered via default distribution deals
- •Universal Search let Google absorb verticals even when rivals had better point products
- •Consumer AI likely consolidates around a primary entry point (ChatGPT)
- •Enterprise AI more likely to remain specialized and multi-entry
- •Strategic takeaway: distribution can beat purely technical superiority
- 50:28 – 55:38
Quick-fire: founder mode, parenting, change, and what he’s proud of
In rapid Q&A, Sridhar reframes ‘founder mode’ as simply being effective, shares a minimalist philosophy of parenting, and emphasizes belief in personal change. He ends on pride in being a present, loving father and the challenge of balancing humanity with performance as CEO.
- •‘Founder mode’ as a shorthand for effectiveness—anyone can do it
- •Parenting: ‘90% presence, 10% luck’ and the primacy of showing up
- •Core contrarian belief: more is changeable than people assume
- •No major regrets: acts on instinct, fails, learns, continues
- •Proudest achievement: being a good dad; hardest CEO part: people and being humane while driving outcomes
