Exclusive: Seltz, a startup rebuilding web search for AI agents, raises $12.5 million in seed funding

AI search startup Seltz raised $12.5 million in seed funding, led by Speedinvest and B Capital, to build an independent search stack for AI agents and chatbots. The company says its crawler already indexes hundreds of millions of pages per day and returns results in under 200 milliseconds, positioning it against Google and other AI search rivals. The news is supportive for the AI infrastructure and private markets space, but it is unlikely to move broader markets.

Analysis

This is less about another “AI search startup” and more about the market beginning to price a structural break in who controls retrieval margins. If agents become the primary consumers of web data, the economic value shifts from click-generating ranking to low-latency, machine-readable extraction, which favors full-stack infrastructure vendors and hurts any intermediary that depends on legacy search APIs or scrapes without proprietary indexing. The first-order implication is modest, but the second-order effect is a proliferation of vertical, task-specific search layers inside enterprise workflows, which could compress the moat around generalist search over the next 12-24 months.

For GOOGL, the risk is not immediate query share loss; it is that AI-native retrieval reduces the strategic scarcity of its index over time by training customers to expect structured answers rather than ad inventory exposure. That matters because the monetization model is still built around human attention, while agents optimize for utility and cost per successful task. In the medium term, the more dangerous pressure is not direct competition from Seltz but the normalization of alternative indexes that eventually weaken Google’s pricing power in search distribution and API access.

NBIS is the cleaner listed beneficiary because any enterprise or model-company outsourcing search stack components becomes a demand tailwind for infra platforms that can package adjacent AI workflow services. RAMP is a softer read-through: if enterprise buyers adopt agentic research and procurement workflows, spend controls, vendor discovery, and reconciliation become higher-value modules, but that benefit is several quarters behind and depends on conversion into actual enterprise deployments. The contrarian miss is that the market may be overestimating how quickly AI search can monetize; technically impressive retrieval does not equal durable demand if it cannot lower total task cost enough to beat incumbent tools by a wide margin.

Key Decisions for Investors

• Reduce tactical long exposure to GOOGL on strength; the near-term legal and competitive overhang is not about volume today but about multi-year index commoditization. Better entry to re-add would be a 5-8% drawdown or evidence of AI search monetization bridging to ad economics.
• Go long NBIS vs short GOOGL as a 3-6 month pair trade: NBIS has cleaner upside from AI stack fragmentation, while GOOGL faces increasing risk of margin dilution from defensive investment and ecosystem leakage. Target 1.5-2.0x gross exposure with a hard stop if Google discloses accelerating enterprise AI search adoption.
• Small long RAMP as an indirect beneficiary of agentic enterprise workflows, but size it as a call option rather than a core position. Use a 6-12 month horizon; upside depends on procurement/task automation expanding beyond pilot programs, so take profits into any multiple re-rating before revenue proof catches up.
• Avoid chasing venture-style winners in public comps until enterprise sales conversion becomes visible. The fastest money here is in infrastructure adjacency, not in assuming every AI search startup becomes a category leader.
• If you want expression with convexity, consider short-dated GOOGL puts as a hedge against headline risk from further search-stack competition or litigation escalation; the thesis is not collapse, but gradual multiple compression if investors start assigning a higher probability to index disintermediation.