A $60 million funding round underscores a growing belief in Silicon Valley that artificial intelligence may not only run on advanced semiconductors, but also help engineers create them faster and at lower cost.
Artificial intelligence has spent the past several years driving a global race for more powerful semiconductors, pushing chipmakers and cloud companies to pour billions of dollars into the hardware needed to train and deploy large models. Now, a new startup is making a more provocative argument: AI should not only consume chips, it should help design them.
That idea moved closer to the mainstream this month after startup Cognichip announced it had raised $60 million in Series A funding to build AI systems aimed at assisting semiconductor engineers. The financing, led by Seligman Ventures with participation from SBI Investment and existing backers, brings the company’s total funding to $93 million and gives fresh visibility to a corner of the AI industry that is trying to tackle one of hardware’s most stubborn problems.
For decades, designing a cutting-edge chip has been among the most expensive, time-consuming and specialized tasks in technology. Modern processors contain extraordinary numbers of transistors, must satisfy exacting requirements on power consumption and heat, and need to function reliably across a web of manufacturing and packaging constraints. Industry veterans often describe chip design as a process that remains highly serial, with progress in one stage dependent on work completed in another. That structure makes development cycles long and expensive even before a design reaches the factory.
Cognichip is betting that AI can change that equation.
The company says it is building what it calls “Artificial Chip Intelligence,” a physics-informed foundation model tailored for semiconductor design. In practical terms, the goal is to create software that can work alongside human engineers on tasks such as navigating design trade-offs, exploring potential architectures, and reducing the time needed to move from concept to manufacturable layout. The ambition is not merely to automate isolated chores but to compress a development timeline that in advanced chips can stretch across years.
The pitch arrives at a moment when demand for new silicon has become more urgent. AI companies want specialized accelerators. Cloud providers are designing in-house chips to reduce dependence on outside suppliers. Automakers, industrial firms and telecom companies continue to demand more customized semiconductors. Yet as strategic interest in chips has risen, the complexity of designing them has also grown. The newest generation of AI processors packs in vast numbers of transistors and requires increasingly sophisticated engineering across digital, analog and mixed-signal domains.
That widening gap between demand and the difficulty of supply has created an opening for startups that promise better design tools.
Faraj Aalaei, Cognichip’s founder and chief executive, has framed the opportunity in terms familiar to software developers who have watched AI coding assistants become more capable. In software, generative AI tools have been adopted to speed repetitive work, suggest solutions and allow engineers to focus more on higher-level decisions. Cognichip argues that semiconductor design, despite being far more constrained by physics and manufacturing realities, could see a similar shift if AI models are trained on the right technical data and embedded deeply enough into engineering workflows.
The comparison is appealing, but it also highlights the challenge. Writing software and designing chips are not the same problem. A bug in code can often be patched after release. Errors in silicon can cost enormous sums, trigger delays measured in quarters, or force entire product plans to be rewritten. Semiconductor design is governed not just by logic, but by the physical behavior of circuits, manufacturing tolerances, verification bottlenecks and the demands of production at advanced process nodes. Any AI system operating in that environment has to do more than generate plausible outputs. It must deliver results engineers can trust.
That is why Cognichip’s emphasis on physics-informed models matters to investors and industry observers. Rather than portraying chip design as just another language problem, the company is trying to position its technology as grounded in the constraints that define real semiconductor work. It says it has built curated datasets and production-ready integrations intended to support this approach. The company also claims its tools can cut development costs by more than 75% and reduce design cycles by more than half, figures that, if proven in commercial deployments, would amount to a major shift in the economics of chip creation.
For now, though, those claims remain largely a promise about the future. Cognichip has not publicly pointed to a fully realized chip created through its system, and it has not disclosed the customers it says it has been working with since September. That leaves the startup in a familiar position for frontier AI companies: long on ambition, newly rich in capital, but still needing to prove that its technology can perform in production under the scrutiny of conservative and highly risk-sensitive customers.
Even so, the list of people backing the company suggests that the semiconductor industry is paying close attention. Intel Chief Executive Lip-Bu Tan participated in the round and is joining Cognichip’s board, alongside Seligman Ventures Managing Partner Umesh Padval. Tan’s involvement is particularly notable. He is one of the most respected figures in the chip industry, with decades of influence in design software, semiconductor investing and corporate leadership. His decision to back Cognichip signals that at least some senior industry leaders believe the existing model for chip design is reaching its limits.
That belief is not hard to understand. The economics of chipmaking have become more punishing as Moore’s Law has slowed and the cost of advanced manufacturing has climbed. Only a small number of companies can afford to build the most advanced processors. If AI tools can reduce the time and labor required to design complex chips, they could lower barriers for a broader set of hardware developers. In theory, that would not only help elite chip companies move faster but also allow more startups and industrial players to experiment with custom silicon.
The implications stretch beyond Silicon Valley. Governments in the United States, Europe and Asia have all treated semiconductors as strategic infrastructure, pouring money into manufacturing capacity and supply-chain resilience. But design remains just as critical as fabrication. Owning fabs is one part of technological power; being able to create novel, optimized chips efficiently is another. A meaningful improvement in design productivity could reshape competitive dynamics across AI, cloud computing, consumer electronics and defense.
Still, caution is warranted. The semiconductor sector has a long history of grand claims about automation. Electronic design automation software has already transformed the industry over several decades, enabling levels of complexity that would have once been impossible. The question is whether AI represents the next layer of genuine productivity or simply another tool that will help at the margins while human expertise continues to do the essential work.
The most likely near-term outcome may be somewhere in between. Rather than replacing chip engineers, AI systems may first act as specialized collaborators — accelerating simulation, surfacing design options, flagging likely issues and helping teams move more quickly through well-defined bottlenecks. In that scenario, the role of the engineer becomes even more central, not less. Human judgment would still determine architecture, trade-offs and sign-off, while AI compresses the cycle between idea and decision.
That may be enough to make Cognichip significant even if its most ambitious vision takes years to realize.
The startup’s rise captures a broader transition in the AI economy. For much of the recent boom, chips were discussed mainly as the scarce infrastructure on which AI depends. Nvidia’s dominance, the scramble for data-center capacity and the search for alternatives all reflected a world in which AI demand was reshaping the semiconductor business from the outside. Cognichip represents the reverse flow of influence: AI moving inward, into the engineering process that creates the very hardware sustaining the boom.
Whether that effort succeeds will depend on evidence, not enthusiasm. Cognichip will need to show that its software can survive the unforgiving standards of semiconductor development, where mistakes are expensive and hype is cheap. But the company’s new funding suggests investors believe the potential reward is large enough to justify the risk.
If they are right, one of the most important uses of artificial intelligence may turn out not to be what happens after a chip is made, but what happens long before it reaches the factory floor.
