Synopsys AgentEngineer Brings Long-Horizon AI Agents to Chip Design

Synopsys bets on long-horizon agents for the chip design lifecycle

Synopsys announced on September 28 that it is moving its electronic design automation business from AI-assisted tools to autonomous engineering, unveiling a portfolio of seven long-horizon AI agents called Synopsys AgentEngineer alongside a new orchestration foundation, the Synopsys Autopilot Platform.

The Synopsys AgentEngineer announcement positions the company, Nasdaq-listed SNPS, at the center of the industry's agentic shift. Each agent is designed to reason, plan, and execute complete engineering workflows across hundreds or thousands of reasoning steps, revising its own plan when an intermediate result falls short rather than waiting for a human to trigger the next stage.

General availability of Synopsys AgentEngineer is planned for the end of 2026. The company said more than 30 customer engagements are already underway, and the launch carries endorsements from Intel, NVIDIA, Samsung, MediaTek, Fujitsu, and AheadComputing.

Seven agents, one open platform

The Synopsys AgentEngineer portfolio spans coverage closure, PPA closure, analog design, mask synthesis, and finite element, computational fluid dynamics, and electromagnetic coding agents. An eighth lane is reserved for customer-built agents in an open ecosystem.

These agents cover verification, system validation, implementation, analog and mixed-signal design, manufacturing, and simulation and analysis. Task-level agents handle bounded work such as autonomous coverage closure, software bring-up and validation, multi-die 3DIC assembly, analog layout synthesis and design migration, and mask synthesis.

Underneath sits the Synopsys Autopilot Platform, which Synopsys describes as a comprehensive, open, and secure foundation for autonomous engineering. It supplies context intelligence, persistent memory, telemetry, security, and governance, and connects agents to Synopsys EDA and simulation ground-truth engines.

Synopsys confirmed that Synopsys AgentEngineer runs on leading commercial and open-source language models rather than a proprietary reasoning model. The differentiation, the company says, concentrates on context intelligence and on privileged tool APIs that let agents run Synopsys engines more efficiently than public interfaces allow. That is where its claimed 2x token efficiency comes from.

From AI-assisted design to autonomous engineering

"Our customers are re-engineering their engineering workflows across semiconductor and systems products to keep pace with increasing system complexity and tight market windows," said Ravi Subramanian, Chief Product Management Officer at Synopsys, in the company's press release distributed via PRNewswire. "Synopsys' portfolio of AgentEngineers and the Autopilot Platform enable customers to accelerate their shift from AI-assisted design to autonomous engineering, built on trusted EDA and simulation and analysis engines."

The performance claims come from vendor-reported customer engagements. Synopsys cites up to 50x faster verification closure and 20% higher coverage. Fujitsu reported a 10% to 30% RTL productivity boost, according to the release.

Two weeks before the launch, Synopsys sketched the destination. Thomas Andersen, Vice President of AI and Machine Learning at Synopsys, told the Futurum Group at the AI Infra Summit that "the silos will go away. There will just be a chip design engineer." In the end state Andersen described, the tools humans operate today become engines, engineers write specifications and keep the creative decisions, and agents launch tools, fix errors, and close timing, congestion, and verification issues behind the scenes.

The Futurum Group, which analyzed the Synopsys AgentEngineer launch in detail, called it the furthest any vendor has gone toward lifecycle agents that let a single team design across every stage. Its analysis, published on September 28, argues the architecture could consolidate procurement around whichever stack the agents live in, putting integration pressure on point-tool rivals.

Futurum also flagged the link to NVIDIA's OpenShell secure runtime, an open-source layer that isolates autonomous agents in sandboxes and enforces policy at the infrastructure layer before actions reach the host environment. NVIDIA's Tim Costa named Nemotron models, the NVIDIA Agent Toolkit, and OpenShell as ingredients in the collaboration, making Synopsys the most prominent engineering-domain application of that runtime. The move echoes the safety-first direction explored in earlier coverage of NVIDIA's own agent safety platform, which puts agent controls on separate hardware outside the agent's reach.

A duopoly race in agentic chip design

Four days before the Synopsys AgentEngineer announcement, rival Cadence added an RTL Generation Agent to its ChipStack AI Super Agent, converting natural language specifications into PPA-optimized RTL with claimed 24% area and 18% power improvements over foundation model code generation. Early access is planned for the fourth quarter of 2026.

The contest now runs on breadth rather than concept. Cadence's agentic footprint concentrates on spec-to-RTL, while Synopsys claims the span from there through signoff and system physics. Both architectures stack long-horizon agents over task agents over trusted engines.

The partnership dimension adds another front. Synopsys is collaborating with Microsoft to offer autonomous workflows on the Microsoft Discovery platform, where early benchmark testing showed 25% to 40% reductions in debug cycle time. TMTPost reported that AMD is evaluating the technology for next-generation products.

Monetization remains a 2027 story. Synopsys expects no 2026 revenue contribution from its integrated AI solutions, and on its Q3 FY2026 earnings call management framed agents driving EDA tools at higher duty cycles as an incremental growth opportunity. An agent iterating to coverage closure runs simulation, emulation, and formal engines at a duty cycle no human team sustains, and every additional design start compounds the effect.

That economic logic also explains why the industry is watching token efficiency so closely. Futurum noted it has heard of a major design lab that ran up a seven-figure token bill on internally built agents before bringing in outside help to restructure the spend. The Synopsys AgentEngineer pitch is that privileged APIs and context intelligence make the self-built path look wasteful.

Whether production teams actually hand over the reins is the open question. Futurum's framing is that capability and operating autonomy will diverge for years, with verification crossing the trust threshold first because its ground-truth engines score every intermediate result, while analog and manufacturing lag because a bad layout or mask decision surfaces expensively and late.

Still, the direction is unmistakable. The same pattern of agents moving from bounded tasks to full workflows is playing out across the agent economy, from software testing to agent-to-agent commerce infrastructure. Synopsys AgentEngineer simply brings it to the place where the chips that run every other agent get designed.