Cerebras Systems has introduced a massive new chip designed to handle the explosive growth in AI inference workloads, marking a significant shift in how companies approach artificial intelligence deployment. The company unveiled the Wafer Scale Engine 3 (WSE-3), a wafer-sized semiconductor containing 4 trillion transistors and 900,000 AI cores, alongside 44 gigabytes of on-chip SRAM. When integrated into the Cerebras CS-3 system, the chip delivers 125 peak AI petaflops, a measure of computing speed that can be scaled up to 2,048 CS-3 systems in a single cluster for a total of 256 exaflops of AI compute.

What Makes Cerebras' New Chip Different From Competitors?

The WSE-3 represents a dramatic leap in scale compared to traditional graphics processing units (GPUs) used for AI work. According to Cerebras leadership, the chip is 52 times larger than an Nvidia GPU, with 52 times the number of cores, more than 800 times the on-chip memory, and orders of magnitude more memory bandwidth and communication bandwidth. The company achieved this performance increase while maintaining the same power envelope and price as its previous generation, the WSE-2.

"We didn't build it big because big was easy or big was sexy, but we bring 52 times the number of cores, more than 800x on-chip memory, and more than three orders of magnitude more memory bandwidth and communication bandwidth," said Dr. Andy Hock, VP of Product at Cerebras.

Dr. Andy Hock, VP of Product at Cerebras Systems

The CS-3 system consumes 23 kilowatts of power, making it relatively efficient for its computational capacity. While Cerebras has not officially disclosed pricing, industry estimates suggest the chips cost between $2 million and $3 million each.

How Is Cerebras Positioning Itself in the AI Inference Market?

The timing of the WSE-3 launch reflects a broader industry trend toward inference, the process of running trained AI models to generate predictions or responses. Cerebras has completed construction of Condor Galaxy 2, its second cloud supercomputer built as part of a $900 million partnership with G42, and is now building a third Condor Galaxy machine featuring the new CS-3 chips. This expansion demonstrates the company's confidence in sustained demand for specialized AI hardware.

Beyond its own chips, Cerebras has partnered with Qualcomm to deploy the latter company's Cloud AI 100 Ultra chips for inference work alongside its training supercomputers. The company is specifically training models to perform efficiently on these inference chips, optimizing them to eliminate unnecessary computations such as multiply-by-zero operations. This approach reflects a strategic focus on making AI systems more practical and cost-effective to operate at scale.

Steps to Understanding Cerebras' Competitive Strategy

• Wafer-Scale Design: Cerebras manufactures chips at the size of an entire silicon wafer, maximizing the number of transistors and cores on a single piece of silicon, unlike competitors who use smaller chip designs.
• On-Chip Memory Focus: The WSE-3 includes 44 gigabytes of on-chip SRAM, dramatically reducing the need to move data between the chip and external memory, which is a major bottleneck in traditional AI accelerators.
• Inference Optimization: The company is training models specifically for inference chips, reducing computational waste and making AI systems cheaper to operate after the initial training phase.

How Is the Market Reacting to Cerebras' Growth?

Wall Street has shown mixed sentiment toward Cerebras' progress. In early June 2026, Morgan Stanley initiated coverage with an overweight rating and a $250 price target, citing the company's potential to benefit from rising demand for low-latency AI inference. This bullish call sent Cerebras stock up approximately 19 to 20 percent as traders embraced the AI accelerator narrative.

However, the company's first-quarter earnings report delivered a reality check. Cerebras more than doubled its Q1 revenue to $193.4 million and narrowed its net loss year over year, but missed earnings per share (EPS) expectations. The market punished the miss sharply, with the stock falling approximately 16 to 17 percent on trading volume roughly double normal levels. This volatility highlights the tension between Cerebras' impressive revenue growth and investor expectations for profitability.

Despite the earnings stumble, Wall Street remains constructive. Wedbush reiterated its Outperform rating on Cerebras, flagging three key upside factors: early share gains in the rapidly expanding AI accelerator market, potential increases in TSMC wafer supply, and the future launch of the WSE-4 chip. The company's positive operating cash flow of $12.3 million in Q1 suggests that the path to sustained profitability is becoming clearer, even as free cash flow remained negative at approximately $119.6 million due to heavy capital spending.

For investors and industry observers, Cerebras represents a bet on specialized AI hardware designed specifically for inference workloads. The company's success depends on three critical factors: securing sufficient wafer supply from Taiwan Semiconductor Manufacturing Company (TSMC), executing flawlessly on next-generation hardware, and capturing market share as demand for AI inference accelerates across enterprises. The WSE-3 launch and Condor Galaxy expansion suggest Cerebras is moving aggressively to capitalize on this opportunity, even as the stock market demands proof that growth will eventually translate into profits.