Cerebras Systems is preparing to hit the public markets at what appears to be the perfect moment for an artificial intelligence infrastructure company. The AI chipmaker, which develops ultra-fast inference-focused processors designed to compete against NVIDIA GPUs, is expected to launch what could become the largest U.S. IPO of 2026 so far. Investor demand has reportedly exploded ahead of pricing, with the deal said to be more than 20 times oversubscribed as enthusiasm surrounding AI infrastructure, hyperscaler spending, and inference computing continues to accelerate.
Cerebras plans to sell 28 million shares in its IPO, with the original range set between $115 and $125 per share before reports emerged that the range could be increased to as high as $125 to $135 due to overwhelming demand. At the upper end of the revised range, the company could raise roughly $3.5 billion and command a valuation approaching $26.6 billion. Underwriters also hold an option to purchase an additional 4.2 million shares, potentially increasing total proceeds even further. Shares are expected to trade on the Nasdaq Global Select Market under the ticker CBRS.
The underwriting syndicate is stacked with major Wall Street firms, led by Morgan Stanley alongside Citigroup, Barclays, and UBS. According to reports, bankers are even asking institutional investors to submit limit orders specifying both the number of shares they want and the maximum price they are willing to pay — an unusual process that underscores just how aggressive demand for the deal has become. Early indications reportedly exceeded $10 billion in orders before the formal roadshow even began.
Founded in 2015, Cerebras was built around a bold idea: that traditional GPU architectures would eventually struggle to meet the communication and latency demands of advanced AI workloads. Instead of relying on many smaller chips connected together, Cerebras designed what it calls the Wafer-Scale Engine, or WSE, which uses an entire silicon wafer as a single giant processor. The company claims its flagship WSE-3 contains 4 trillion transistors, 900,000 AI-optimized cores, and dramatically higher memory bandwidth than competing AI chips. Cerebras says the WSE-3 is 58 times larger than NVIDIA’s B200 chip and delivers 2,625 times more memory bandwidth.
The company is positioning itself as a major player in the growing shift from AI training toward inference — the process where trained AI models generate real-time outputs for users. Cerebras argues that inference speed will become one of the defining battlegrounds in AI infrastructure, particularly as enterprises increasingly demand lower latency and faster response times from large language models and agentic AI systems. The company claims its architecture can run certain inference workloads up to 15 to 21 times faster than competing GPU-based systems.
One of the biggest reasons for investor excitement is Cerebras’ deepening relationship with OpenAI. In January 2026, Cerebras announced a multi-year agreement with OpenAI valued at more than $20 billion that will provide up to 750 megawatts of AI compute capacity through 2028. Under the arrangement, OpenAI and Cerebras will co-design future AI models optimized for future Cerebras hardware. OpenAI has also provided Cerebras with a $1 billion working capital loan tied to deployment milestones and holds warrants allowing it to purchase more than 33 million shares at a fraction of a penny per share.
The OpenAI relationship extends beyond business partnerships and into the shareholder base itself. OpenAI CEO Sam Altman and several prominent OpenAI executives and affiliates are reportedly investors in Cerebras. The relationship became notable enough to appear in legal filings tied to Elon Musk’s lawsuit involving OpenAI governance and investments. At one point, OpenAI reportedly even considered acquiring Cerebras outright before ultimately pursuing a commercial partnership instead.
Cerebras has also landed a potentially transformative partnership with Amazon Web Services. In March 2026, the company signed a binding term sheet under which AWS will become the first hyperscaler to deploy Cerebras systems inside its own data centers. The arrangement includes minimum capacity commitments, exclusivity protections, and warrants allowing AWS to purchase nearly 2.7 million shares tied to future deployment volumes. The partnership could significantly expand Cerebras’ reach by making its inference capabilities available through AWS’ global cloud distribution network.
Financially , the company is showing explosive growth, though the story is more nuanced beneath the surface. Revenue surged from $24.6 million in 2022 to $78.7 million in 2023, then to $290.3 million in 2024 before climbing another 76% year-over-year to $510 million in 2025. Cerebras reported GAAP net income of $237.8 million in 2025 after posting a net loss of $481.6 million in 2024. However, adjusted figures paint a less profitable picture, with the company still generating a non-GAAP net loss of roughly $75.7 million after excluding items tied to stock compensation and financial instruments.
Perhaps the most eye-catching financial metric is Cerebras’ backlog. Remaining performance obligations stood at approximately $24.6 billion at the end of 2025, with a significant portion tied to the OpenAI master relationship agreement. The company expects only about 15% of that revenue to be recognized over the first two years, with the majority spread over later years as infrastructure deployments ramp. That enormous backlog is one of the major reasons investors are willing to overlook near-term profitability concerns.
Still, risks remain substantial. Customer concentration is one of the largest concerns. The company disclosed that a significant portion of revenue and receivables are tied to a small number of customers including OpenAI, G42, MBZUAI, and AWS. A deterioration in any of those relationships could materially damage the business. OpenAI also holds substantial leverage under the agreement, including the ability to terminate portions of the deal if Cerebras fails to deliver promised compute capacity on schedule.
Supply chain risks are another issue. Cerebras acknowledged that its manufacturing and hardware supply chain is highly complex and globally interconnected. Any disruption in semiconductor fabrication, cooling infrastructure, networking equipment, or power availability could delay deployments and hurt growth. Export controls also represent a meaningful risk given the company’s global footprint and relationships tied to the Middle East.
Even with those concerns, the backdrop for the IPO appears exceptionally favorable. AI infrastructure remains the market’s hottest investment theme, semiconductor stocks have rallied sharply, and investors continue rewarding companies tied to hyperscaler AI spending. The oversubscription levels suggest institutions are aggressively chasing exposure to next-generation AI compute platforms, particularly as the market begins focusing more heavily on inference rather than just training workloads. Whether Cerebras ultimately justifies the valuation remains an open debate, but the early signals suggest CBRS could become one of the defining AI IPOs of this cycle.