Highlights:

  • Recently, DreamBig launched its MARS platform, aiming to mass-produce a new generation of AI chips built on open standard chiplets.
  • DreamBig asserts that the MARS chiplet standard tackles two significant challenges in constructing AI servers and accelerators: scaling up computing power and scaling out networking capabilities.

Recently, an artificial intelligence chip startup, DreamBig Semiconductor Inc., secured a USD 75 million equity funding round, raising its total funds to over USD 93 million, according to sources.

The recent funding round was co-led by the Samsung Catalyst Fund and the Sutardja Family, with participation from new investors, including Hanwha, Event Horizon, and Raptor. Existing investors such as UMC Capital, BRV, Ignite Innovation Fund, and Grandfull Fund also took part.

Founded in 2019, DreamBig developed the “MARS Platform,” an open chiplet platform designed for creating high-performance semiconductors using advanced 3D high-bandwidth memory technology.

In January, DreamBig announced the launch of its MARS platform, aiming to introduce a new generation of AI chips to the mass market. These chips will be constructed using open standard chiplets, which are modular integrated circuits.

The startup stated that the MARS platform allows customers to concentrate their silicon investments on specialized capabilities that set them apart from competitors while benefiting from the reduced costs of open standard chiplets. This enables customers, including chipmakers and server hardware manufacturers, to develop computer chips tailored for specific tasks such as processing generative artificial intelligence, automotive, or edge applications.

DreamBig claims that the MARS chiplet standard addresses two major challenges in building AI servers and accelerators: scaling up computing and scaling out networking. According to the company, MARS overcomes these bottlenecks by being the first chiplet to offer direct access to both static random-access memory (SRAM) and dynamic random-access memory (DRAM) tiers. This capability allows it to function as a platform for computing, accelerator, or networking chips, enabling rapid data movement, data caching, or data processing based on the specific application.

DreamBig announced that the funds from the recent round would be used to speed up the development of its chiplet standard and bring it to market, along with its Chiplet Hub product—a development platform for companies interested in experimenting with its chiplets.

Co-founder and Chief Executive Sohail Syed stated that recent investment highlights the market’s increasing recognition of his company as a “transformative force” in the AI and data center infrastructure sector. He said, “Our open MARS Chiplet Platform enables unparalleled scale-up and scale-out solutions so customers can achieve the highest levels of performance and energy efficiency at lowest cost and fastest time-to-market.”

Marco Chisari, Head of Samsung Electronics Co Ltd.’s Semiconductor Innovation Center, expressed his belief that “chiplet-based AI solutions” represent the most practical path forward for the AI industry. He said,  “The ever-growing demands for intensive workloads and memory-bound applications – from generative AI to automotive – are fueling the need for more advanced chiplet-based designs with 3D HBM stacking.”

DreamBig joins a series of AI chip startups that have secured significant funding this year. Recently, California-based Halo Industries Inc., specializing in a laser manufacturing platform for the semiconductor sector, closed a USD 80 million Series B round led by the U.S. Innovative Technology Fund. This followed closely on the heels of another U.S. startup, Etched.ai, which announced a USD 120 million funding round less than a month ago aimed at challenging Nvidia Corp., the current leader in AI chips.

The largest chip-related funding round of the year went to Black Semiconductor GmbH, which announced last month that it had secured USD 274 million. This funding will bolster its research into graphene-based chip-to-chip interconnects. On the other end of the spectrum, Innatera Nanosystems B.V., a Dutch chip startup focused on developing energy-efficient semiconductor hardware for sensor-based edge applications, raised USD 21 million in a Series A round of funding last month.