Could this OpenAI CEO backed chipmaker stock rival Nvidia?

Nvidia’s (NASDAQ: NVDA) dominance in the profitable AI chipmaking sector has attracted the attention of would-be competitors. 

One of those is Cerebras — a chipmaker founded in 2015 by SeaMicro veterans and two of its founders, Andrew Feldman and Gary Lauterbach, after SeaMicro was acquired by Advanced Micro Devices (NASDAQ: AMD) for $334 million in 2012.

In addition to this, OpenAI CEO Sam Altman and Sun Microsystems co-founder Andy Bechtolsheim are backing the company. 

Semiconductor stocks have seen record performance since the advent of the AI revolution, and no company has seen a bigger increase in valuation than Nvidia, which is up 158.36% year-to-date (YTD) at press time.

Cerebras stands out for more than just its experienced staff — the chipmaker’s technology provides an alternative method for training AI models, and the company has even filed a preliminary IPO prospectus.

Cerebras IPO and institutional backing

In a year that has seen limited recovery in terms of tech IPOs, Cerebras is a pretty unique case — the unicorn startup has raised more than $720 million in venture capital funding and was valued at $4.25 billion as of 2021. 

Notable investors include Foundation Capital, Benchmark, Eclipse Ventures, Alpha Wave, Coatue, and Altimeter — all of which own a 5% or larger stake in the company.

Cerebras is set to trade on the Nasdaq under the ticker ‘CBRS’, with Citigroup and Barclays as lead underwriters — significantly, neither Goldman Sachs nor Morgan Stanley, both leaders in tech IPOs, are involved. The company is aiming to raise between $750 million and $1 billion — which would place it at a valuation of $7 billion to $8 billion.

CBRS wafer-scale engine chips

At the core of Cerebras’ appeal is its unique Wafer-Scale Engine (WSE) technology, now in its third iteration with the WSE-3.

Nvidia’s chips, like the H100, are based on traditional GPU architecture designed to handle a variety of computing tasks, making them adaptable across different applications. 

Cerebras’ WSE-3, by contrast, is an application-specific processor optimized purely for AI workloads, especially in inference and training tasks.

At roughly 46,000 square millimeters, the WSE-3 is 56 times larger than Nvidia’s largest chip, and its integrated design eliminates the need for multiple processors linked by interconnects. 

This massive chip holds 4 trillion transistors and 900,000 cores, as well as 44GB of on-chip memory, enabling data to be processed within the chip rather than requiring frequent off-chip memory access.

CBRS uphill battle against Nvidia

Although CBRS chips can provide better performance for certain tasks, Nvidia has the benefit of a well-developed, versatile, and entrenched software ecosystem in CUDA — something that Cerebras currently lacks. The company’s chips are also designed to handle a broader range of tasks, which include graphics rendering, while CBRS specializes solely in AI inference.

NVDA also has a first-mover advantage — having already secured various partnerships with hyperscalers. Nvidia also has a massive R&D budget at its disposal, as well as the means to attract top talent in the space. Even with broad-based funding, it’s very unlikely that CBRS will be able to compete at scale.

Profitability and revenue concentration issues

In terms of financials, the company recorded a net loss of $77.8 million and $8.7 million in sales in the first six months of 2023. In the first six months of 2024, the company had $66.6 million in losses, but sales expanded rapidly to $136.4 million.

That’s good progress — however, the company is still far from profitability. These issues are further exacerbated by the fact that 87% of the company’s revenue comes from a single source — Abu Dhabi-based AI company G42.

Cerebras has identified the overreliance on G42 as a weakness in the prospectus. Regulatory complications and shifts in G42’s AI infrastructure could severely hamper the IPO. The usual issue of long, complex, international supply chains is also present.

However, it should be noted that CBRS could very well secure a niche for itself — albeit a much smaller, more specialized one.

Investors and traders should keep an eye out for any press releases as we move closer to the public listing, which is expected to occur in late 2024 or early 2025. In the meantime, CBRS’s fate is completely tied to G42.