Daniel Nenni's preview of roughly 30 companies heading to Long Beach shows a show floor that has tilted, almost entirely, toward startups building agentic, verification, and analog AI tooling.
At DAC 2026, the real story isn't any single startup. It's that the entire chip-design supply chain is being rebuilt around AI agents, and the roughly 30 exhibitors Daniel Nenni has previewed on SemiWiki may be the densest signal yet that this transition has crossed a threshold.
That's a claim worth tagging. Nenni's curated list is a pre-event preview, not a verified floor report. He explicitly tells readers he intends to walk the floor, visit booths, and blog more once the show opens July 27 in Long Beach, California. What his article offers is a roster snapshot, and the shape of that roster is what's interesting.
The clusters are tight. Agentic and AI tooling for end-to-end design shows up across Agentrys, AIDAchip, Architect Labs, Verkor, VerifAIX, VoltAI, and MooresLab. AI for verification pulls in Caspia, Mach42, Normal Computing, Silimate, GoldsmithAI, SIgmanticAI, and InstaDeep. AI for analog, RF, and custom design is represented by GenAlpha, Move Silicon, Future Intelligence Lab, and Mach42 (which spans two clusters). AI for PCB design is split between Quilter.AI and Cadstrom, while physical verification gets a parallel-AI entry in Ramtera.
That's a lot of companies to fit under one DAC tent, and Nenni flags it directly: this may be the largest collection of startup companies at one DAC, and AI technology dominates the list. The framing is his, not a quantified measurement. The exhibitor count is his own tally from a curated preview. But the pattern, even allowing for marketing language in the capsule blurbs, is hard to miss.
The interesting second question is what the clusters actually do. The verification group is the largest and most uniform. Caspia, Mach42, Normal Computing, Silimate, and GoldsmithAI all pitch AI for the formal and dynamic verification problems that have grown unwieldy at advanced nodes. Silimate stands out for naming Tenstorrent and SiFive as customers in Nenni's preview, a rare move for a startup at this stage, and one that suggests real silicon is already running through their tools.
The agentic-design cluster is more varied. Some companies on Nenni's list pitch AI for design space exploration, others for IP derivation, regression triage, or coverage closure. Future Intelligence Lab, in particular, claims a five-person design team can do what a fifty-person team used to do, leaning on agents for DV regression, IP adoption, and PPA exploration. That figure is the company's own framing, not a measured industry outcome, and the gap between aspiration and shipping product is exactly the kind of claim that needs floor follow-up.
The analog and custom-design cluster is where the AI thesis gets tested in a particularly unforgiving way. Analog doesn't reward brute force. Tools like GenAlpha, Move Silicon, and Mach42 are betting that the patterns hidden in mature analog IP can be turned into something generative. The fact that three companies are making similar bets at the same show suggests the bet has at least graduated from contrarian to consensus.
The supporting categories round out the picture. RF, wireless, and specialty IP are covered by Fermionic, FILPAL, and Suitera. Security gets Amida and Caspia. Supply chain has Cofactor and AI-TechSales. Prototyping and emulation are split between Correva, Oboe, and Flexcompute, with Flexcompute extending its optical and EM simulation work into AI-driven aerodynamic optimization and photonic IC design.
There are also some non-AI outliers worth noting. Aras PLM is showing up with a customer list that includes Toyota and Renesas, which puts it into direct competition with Siemens, Dassault, and PTC in the chip-and-system PLM space. AI-TechSales positions itself as a services firm for the AI-native semiconductor stack, naming a portfolio that includes InPSY, Quaxys, Rise Design Automation, ModelCat, CraftiFAI, softweb solutions, yieldWerx, Glide Systems, Axiomise, Tuple Tech, and MachineWare. Ricursive, the most heavily capitalized company in the preview, is described by Nenni as a frontier AI lab with a self-improving system for chip design, with stated $335M in startup funding, and a talent pedigree that includes Google DeepMind, Anthropic, NVIDIA, Cadence, Apple, xAI, Stanford, MIT, and Harvard.
What does the roster say about the next two to three years? Three things, with appropriate hedging. First, the verification and analog-AI clusters are dense enough that incumbents like Cadence, Synopsys, Siemens EDA, and Ansys will face real, if still small, competitive pressure on product roadmaps. Second, the willingness of startups to name customers (Silimate with Tenstorrent and SiFive, Aras with Toyota and Renesas) suggests that the buyer side of AI-native design tooling is no longer hypothetical. Third, the sheer number of agentic-AI design startups at a single DAC, combined with the funding and pedigree behind Ricursive, suggests the floor has become the announcement channel for a category transition rather than a place to show incremental EDA point tools.
The watch items for the show itself: which companies bring working silicon, which bring pilots with named customers beyond their PR, and which rely on the aspirational framing that Nenni flags in his preview. The roster is a thesis. DAC 2026 will be the first real test of how much of that thesis has shipped.