Anthropic, Gamma, and Clay walk onto TechCrunch Disrupt and put a number on enterprise AI: $300M+ of ARR, 50 employees, and the unbundled path has already won the first wave

TechCrunch broke the lineup on Sep 28: Cat de Jong, Anthropic's Head of Applied AI, will share patterns the company sees across its enterprise Claude deployments; Grant Lee, Gamma's CEO, will speak from inside a50-person team that hit $100M ARR without a sales force; Kareem Amin, Clay's CEO, will speak from inside a company that crossed $200M of run-rate revenue in September with 17,000 paying customers. The shared session title — \"What Anthropic Sees When Enterprises Actually Deploy Claude\" — frames the entire 2026 enterprise AI debate as a deployment problem, not a model problem.

Gamma runs on roughly 50 people and crossed $100M ARR in six years while staying profitable for fifteen consecutive months. Clay reached its first $1M ARR in roughly four years, then compounded to $100M by December 2025, with Sacra tracking about $150M ARR by May 2026 and the New York Times reporting a ~$200M run rate in September. Both grew primarily through product-led adoption — Gamma with no paid marketing and a free tier that converts, Clay with a go-to-market angle its CEO openly calls \"upstream from Salesforce\" — and both now count AI-native buyers (Anthropic, OpenAI, Google) among their reference customers.

Salesforce reported Q2 FY27 results on Aug 26, 2026: total revenue of $11.35B (+11% YoY), with Agentforce ARR above $1.5B (up more than 240% YoY) and the combined Agentforce + Data 360 ARR approaching $3.9B. CEO Marc Benioff said on the call that the AI-and-data ARR line is \"about to cross $4 billion,\" and Salesforce raised full-year FY27 guidance to $46.1–$46.4B. ServiceNow reported on Jul 22, 2026: Q2 2026 subscription revenue of $3.877B (+24.5% YoY), ServiceNow AI crossing $1B of annual contract value in the quarter, and a stated path to $1.5B by year-end 2026. Microsoft reported FY26 Q4 on Jul 29, 2026: revenue of $90.0B, Azure annualised past $100B, and Microsoft 365 Copilot at 30M paid seats — up 50% from 20M the prior quarter.

The cracks are not in growth — they are in the deployment pattern. Industry reporting through2026 has repeatedly documented that Salesforce and ServiceNow customers stall at the pilot stage because their agents live inside suites that were never designed for AI-led workflows. The Anthropic–Gamma–Clay panel is the first time all three sides — model vendor, unbundled tool, enterprise deployment view — share the same stage, and the implicit thesis is that the unbundled players will keep eating the workflow-specific spend while the suites defend the procurement interface.

None of the Disrupt panel's growth numbers are feasible without the silicon underneath. On Apr 20, 2026, Anthropic and Amazon jointly announced an expansion that commits Anthropic to more than $100B of AWS spend over the next decade and secures up to 5 gigawatts of Trainium2 through Trainium4 capacity — with nearly 1 gigawatt of Trainium2 plus Trainium3 expected to come online by year-end 2026. Amazon committed an additional $5B of equity to Anthropic on the same day, on top of the prior $8B, with up to $20B more optional. For context: Anthropic's annualized run rate moved from $1B (Jan 2025) to $9B (year-end 2025) to $30B (Apr 2026) to $47B (May 2026) to $65B (end of July 2026) — a curve that physically requires a multi-gigawatt silicon roadmap to stay linear.

The Disrupt session itself is on Oct 13–15, but the catalysts that move the listed names sit around it. Salesforce and ServiceNow will print Q3 numbers in October and November — the test is whether Agentforce's 240% YoY growth and Now Assist's $1B ACV pace hold as the unbundled path takes more workflow share. Microsoft reports FY27 Q1 in late October and will be judged on whether Copilot seat additions (10M in Q4 FY26) reaccelerate or plateau, and on whether Azure growth holds above 40%. Amazon's Q3 print (late October) will show how much of the Trainium ramp is already in AWS revenue. NVIDIA's Q2 FY27 print in late August already reset expectations; the next check is the data-center revenue mix.

Long-term, the unbundled path and the integrated stack converge on the same procurement interface — the place where a Fortune-500 CIO signs the contract. Kareem Amin's framing of Clay as \"upstream from Salesforce\" is the working hypothesis: AI agents handle data, the suite handles the system of record, and the unbundled tool captures the workflow that the suite cannot price per seat. For investors, that argues for holding the infrastructure layer (NVIDIA, Amazon) unconditionally, treating Microsoft as a Copilot-distribution play rather than a model play, and treating Salesforce and ServiceNow as incumbent-yield stories whose AI growth is the cushion against unbundling — not the moat.