Cutting-edge microprocessors have propelled the advancement of artificial intelligence. The question now is, can AI reciprocate?
Cognichip is developing a sophisticated deep learning framework intended to collaborate with engineers in crafting novel computer chips. The challenge it aims to tackle has plagued the sector for many years: the creation of chips is incredibly intricate, prohibitively costly, and protracted. State-of-the-art chips require three to five years to progress from initial concept to widespread manufacturing; the planning stage itself can consume up to two years before any physical arrangement commences. Bear in mind that Nvidia’s newest GPU series, Blackwell, incorporates 104 billion transistors — an immense number to meticulously arrange.
Faraj Aalaei, Cognichip’s CEO and founder, states that within the duration required to produce a novel chip, the market landscape can shift, potentially rendering the entire investment futile. Aalaei’s objective is to introduce the types of AI utilities that software developers have utilized to accelerate their processes into the realm of semiconductor blueprinting.
“These platforms have now attained sufficient intelligence that simply by directing them and specifying the desired outcome, they are capable of generating elegant code,” Aalaei informed TechCrunch.
He asserts that the company’s technology has the potential to diminish chip development expenses by over 75% and shorten the overall schedule by more than 50%.
The enterprise debuted publicly last year and announced on Wednesday that it had secured $60 million in fresh capital, spearheaded by Seligman Ventures. Prominent involvement came from Intel CEO Lip-Bu Tan, who contributed via his venture capital entity, Walden Catalyst Ventures, and will be taking a seat on Cognichip’s board of directors. Umesh Padval, a managing partner at Seligman, is also slated to join the board. Since its establishment in 2024, Cognichip has now accumulated a total of $93 million.
Nonetheless, Cognichip is not yet able to showcase a new chip conceived using its platform, nor has it revealed any of the clients it claims to have been working with since September.
A Techcrunch gathering
San Francisco, California
|
October 13th to 15th, 2026
The firm asserts its unique selling proposition lies in employing its proprietary model, refined with specialized chip design information, instead of commencing with a generic large language model. This necessitated obtaining entry to particular domain-specific instructional data, an undertaking of considerable difficulty. In contrast to software creators, who openly disseminate extensive volumes of code, chip architects meticulously protect their intellectual property, thereby rendering the sort of open-source repository commonly utilized to educate AI coding aids mostly inaccessible.
Cognichip has been compelled to construct its unique data collections, encompassing simulated data, and to acquire licenses for data from collaborators. Furthermore, the company has devised protocols to enable chip manufacturers to safely refine Cognichip’s models using their confidential data without divulging it.
In instances where exclusive data is not obtainable, Cognichip has relied on open-source substitutes. During a demonstration last year, Cognichip extended an invitation to electrical engineering scholars at San Jose State University to test the model during a hackathon. The participating groups successfully utilized the model to conceive CPUs founded on the RISC-V open-source chip framework — a publicly accessible blueprint that anyone can develop upon.
Cognichip faces rivalry from established competitors such as Synopsys and Cadence Design Systems, alongside a multitude of amply financed nascent companies. Notable among these are Alpha Design AI, which secured $21 million in Series A funding in October 2025, and ChipAgentsAI, which concluded an expanded Series A round totaling $74 million in February.
Padval remarked that the present surge of investment into AI infrastructure represents the most substantial he has witnessed in four decades of capital deployment.
“Should this prove to be a robust growth phase for semiconductors and hardware, then it is similarly a robust growth phase for enterprises akin to [Cognichip],” he stated.
{content}
Source: {feed_title}