AI ‘Existential Risk' Is Close to Zero: Databricks CEO

Databricks CEO Ali Ghodsi said the existential risk from AI is “close to zero,” but warned that AI is creating a growing cybersecurity threat by sharply reducing the time attackers need to exploit vulnerabilities. Databricks is preparing for these risks, while Ghodsi said rapid technological change strengthens his preference to keep the company private for now.

Analysis

The investable implication is not a broad AI de-rating but a shift in enterprise AI budgets from experimentation toward security, identity and data-governance controls. As exploit-to-production cycles compress, the cost of delayed remediation rises nonlinearly: enterprises will prioritize tools that prevent credential misuse, secure cloud workloads and validate code before deployment. This favors PANW, CRWD, ZS, OKTA and CYBR, while increasing pressure on point-solution vendors whose products cannot demonstrate measurable reduction in breach probability or response time.

Near term (1-3 months), cybersecurity multiples may rerate modestly if management teams cite AI-driven attacks as an incremental budget source rather than merely a feature-marketing opportunity. The key distinction is whether spending comes from new security allocations versus displacement of legacy vendors; the latter produces winners but not sector-wide revenue upside. Watch forthcoming earnings commentary on net-new platform consolidation, remaining performance obligations, large-deal duration and federal/regulated-industry demand.

The second-order risk sits with software companies accelerating AI-enabled coding and agent deployment without mature permissions, observability and data-loss controls. Higher breach frequency would create episodic downside for cloud/SaaS names with concentrated enterprise data exposure, while cyber insurers could face adverse loss trends before premiums reprice. The contrarian view is that public cyber has already priced a sizable AI-security tailwind; absent evidence of budget expansion, valuation dispersion will matter more than the thematic narrative.

Over 6-18 months, secure data-layer architecture becomes a gating factor for AI adoption, favoring platforms integrated into identity, endpoint, cloud and data workflows rather than stand-alone generative-AI applications. A material cyber incident tied to an autonomous agent or AI-generated exploit would accelerate procurement, but also invite regulation that could slow enterprise AI deployment and reduce near-term consumption growth for AI infrastructure beneficiaries.

Key Decisions for Investors

• Prefer a 3-6 month basket long PANW and CRWD versus short IGV or a high-multiple application-software basket: the trade captures security-budget priority while hedging broad software-duration risk. Reassess if either company reports slowing platform adoption or material billings deceleration.
• Add ZS selectively on weakness rather than chase: zero-trust architecture is a direct beneficiary of accelerated credential and access attacks, but the thesis requires sustained large-enterprise bookings growth and stable billings guidance at the next report.
• Maintain a watch alert on OKTA rather than a fresh directional long: AI-driven identity risk is structurally supportive, but the position needs independently visible evidence of improved retention, enterprise expansion and no renewed security-related reputational drag.
• Avoid treating private-market AI platform valuations as a read-through for public AI infrastructure. A private-company preference can constrain near-term public comparables and liquidity, while a future IPO would be a sentiment catalyst only if it discloses durable consumption growth and credible security/governance monetization.
• Use any confirmed AI-enabled breach involving a major enterprise SaaS or cloud workflow as a tactical catalyst to add cyber exposure; take profits if the event drives indiscriminate sector spikes without corresponding upward revisions to bookings or free-cash-flow estimates.