How do you read the current, noisy moment for artificial intelligence—especially with so much public worry and excitement?
We have lived through several big waves—the browser, the dot-com boom and bust, the mobile revolution, even fads that fizzled such as early chat-bots and the metaverse. None of them let humans engage technology entirely on our own terms. Generative AI changes that contract: anyone who can speak or type can now direct sophisticated systems and receive answers in plain language. That single shift creates a line in time—before AI, interaction meant adapting to machines; after AI, machines adapt to us.
The other breakthrough is AI’s role as connective tissue. Today a model can watch you hail a ride, extract the receipt, file an expense report and notify finance—all without copy-and-paste gymnastics. By removing friction between applications as well as between people and applications, the technology delivers immediate, durable value while we continue debating heavier topics such as employment or metaphysics. Even if AI did nothing more than those two things, it would still represent an historic pivot.
If natural-language AI can simplify everyday apps, what happens when the same idea is applied to enterprise data and the cloud?
Corporate information has long been stranded—first in on-prem software, then in cloud stores that still required translators armed with SQL and BI dashboards. An agentic model layered on Snowflake can traverse all that data directly. Ask, “How did prepaid churn trend after our last price change?” and the answer arrives in prose, charts or recommended actions without an analyst in the middle.
We already use this internally. Every sales record in Salesforce streams into Snowflake; an agent can assemble a two-page brief on Fidelity’s usage patterns, risks and upsell targets in minutes. Extend that pattern across finance, supply-chain or customer-support data and you give every knowledge worker the equivalent of a dedicated research assistant—something that once belonged only to companies with vast analyst teams.
Snowflake is known for governance, observability and reliability. How will you preserve those traits as customers embed large language models more deeply?
From day one we enforced role-based access control: rights attached to a role, not a person, so when an employee shifts departments or leaves, their privileges update automatically. That same mechanism protects petabyte-scale tables with billions of rows where, for example, a regional manager in Spain can see only her territory. The policy engine runs inline with the query planner, so performance is unaffected even under complex rules.
Every AI feature is built on the identical substrate. A chatbot generated in two commands still inherits the exact row-level rules. Beyond access, we push engineering discipline: evaluation sets, regression tests and change logs for prompts as well as code. Enterprise customers need answers that are consistently correct, not “roughly right,” and a platform rooted in data governance makes that possible without bolt-on tools or manual audits.
You spent pivotal years at Google Search and later founded Neeva. Which lessons from those chapters now guide Snowflake’s product culture?
Google taught me the power of intensity and iteration. When PC growth stalled in 2009 and mobile monetised at a tenth the desktop rate, the Search team had to reinvent ads, ranking and infrastructure without ever compromising latency or relevance. That experience embedded a reflex: disrupt yourself before someone else does, and always hold the craft to a consumer-grade standard no matter how technical the product may be.
At Snowflake we keep that ethos alive. The core analytic engine continues to “just work,” but we have re-tooled teams, processes and release cadences so AI moves at consumer speed. Engineers dog-food every feature; if a workflow slows or a label misaligns, the bug report often comes from me. Maintaining that feedback loop ensures new capabilities integrate seamlessly instead of accumulating as separate, harder-to-maintain silos.
Hyperscalers and Databricks push hard into your territory. How do you translate philosophy into day-to-day competitive advantage?
First, we respect rivals—they are smart, motivated and well funded—so complacency is never an option. Our counterweight is focus: a single, integrated platform that scales elastically yet hides most of the plumbing. Customers consistently tell us the decisive factor is simplicity. If spinning up an AI prototype on Snowflake takes two commands and an hour of elapsed time, a competitor offering six services and five security models feels heavy before the proof of concept even starts.
Second, we obsess over cohesion. Governance, performance and developer experience travel together rather than as à-la-carte add-ons. That coherence is not easy to duplicate; it is the product of countless micro-decisions, from how we version UDFs to how we expose metadata. The moment we treat new features as shiny bolt-ons we surrender our advantage, so cultural vigilance is as important as any algorithmic leap.
Revenue is still growing above 30 percent, yet costs and platform complexity also rise. How will you maintain both growth and attractive margins?
Growth comes from two streams. The first is the still-massive migration of analytic workloads from on-prem systems to the cloud, where our engine remains the benchmark for cost-performance. The second stream is newer: data engineering, open data sharing and, of course, AI workloads that now arrive routinely with seven-figure budgets. Together they sustain high net-revenue retention—most recently in the upper-120 percent range—while adding fresh logos.
Margin outlook is equally positive. AI agents already generate tests, boilerplate code and synthetic data, reducing the need for incremental head-count. We are also hiring earlier-career talent fluent in modern tooling and therefore both cost-efficient and innovative. Publicly we have committed to bringing stock-based compensation into the mid-teens within five years, and internal models show that is reachable without starving R&D.
Why do you believe relentless innovation is necessary—and who bears responsibility for the broader societal consequences?
Technology itself is amoral; humans steer its trajectory. Well-applied, AI can help a pharmaceutical firm shorten time-to-cure or a bank detect fraud hours earlier, outcomes that feel unambiguously good. That sense of purpose keeps our teams energized through late nights and daunting problems. Stagnation, by contrast, rarely serves customers or society.
Yet commercial ambition must coexist with civic oversight. Governments and regulators have the charter to weigh issues such as displacement, misinformation and privacy across the whole population. Instruments like the EU’s AI Act will need constant refinement, but the goal is sound: encourage progress without turning a blind eye to unintended harm. Our role is to provide transparent tools and verifiable controls so policymakers have the data they need to legislate wisely.
On a personal level, what was it like to see Neeva acquired and then to step into Snowflake’s CEO role so quickly afterward?
The acquisition was both satisfying and bittersweet. Neeva began as an audacious attempt to reimagine search; winding down that dream stung. My first priority was to land the team safely, and Snowflake proved an excellent home. The episode underscored a lesson I share with my sons: take risks, but cultivate resilience because even brilliant ideas sometimes meet immovable realities.
I joined Snowflake to integrate Neeva’s tech and talent, not to occupy the corner office. Yet the broader AI opportunity emerged, the board asked, and the mandate expanded. That pivot illustrates why I remain optimistic: in technology, apparent endings often seed better beginnings. Staying open to those twists is, in its own way, a strategy.
