CBRS Volatile IPO: Addressing NVDA Competition, Supply Constraints & Partnerships

**Rick Ducat** (0:04)
Welcome to Tech Corner. I'm Rick Ducat. Today, we're looking at Cerebras Systems, a notable player in the AI infrastructure ecosystem. Cerebras delivered earnings for the second time as a public company on Wednesday, August 12th, reporting a net loss of about $450 million and a revenue miss that sent shares plunging lower. Cerebras competes in the AI hardware and hyperscaler business, building unique dinner plate size computer chips and supercomputers designed to train and run AI models at fast speeds. The company operates primarily through two key business segments, hardware and cloud and services.
Hardware, which accounted for 30 percent of total revenue, focuses on designing, manufacturing, and selling massive wafer-scale processors and on-premises AI systems that are designed to accelerate AI workloads without relying on traditional supply limited components like pie bandwidth memory. Meanwhile, cloud and services, which was just under 70 percent of total gap revenue, operates dedicated AI data center infrastructure to deliver high-speed inference as a service, inference meaning when a pre-trained AI model uses its learned patterns to process new information and give new answers. This segment also includes supporting enterprise agentic workflows, coding tools and large language model deployments. As we mentioned, Cerebras is an unusual player in this field. There are no direct competitors in terms of other companies that make 8.5 inch square computer chips. However, its primary competitor in the hardware space is Nvidia, which is the market leader in AI training and inference. Other notable rivals include the AI chip startup Grok and AMD.
In the cloud and hyperscaler field, its main competition is Google's custom tensor processing unit systems designed for internal cloud workloads. Other names in the space include Amazon's AWS, which produces custom training and inference chips, as well as Microsoft and Meta, which build custom internal accelerators called Maya and MTIA respectively.
Cerebras is unique in its chip manufacturing methods. Other chip makers typically cut their chips out of a silicon wafer, which is a very thin and very flat circular section of extremely pure crystalline silicon. This serves as the foundational base of computer chips and other types of electronic circuits. However, Cerebras makes its chips out of the entire wafer, giving it the largest processors ever created. A Cerebras wafer scale engine 3 chip is about 58 times larger than an Nvidia Blackwell B200 chip. So picture something about the size of a dinner plate with 4 trillion transistors versus the B200's 208 billion transistors. In addition to the processing speed advantages with these larger chips, Cerebras' product size provides enough physical space to directly embed RAM onto the chip itself. Other companies have to use off-chip memory, which is slower because the data has to travel farther distances. Cerebras says these unusual processes mean its systems can be substantially faster than traditional supercomputer clusters.
Cerebras reported its second quarterly results which missed Wall Street expectations. Total gap revenue for the quarter came in at $180 million, but that was below the Street's estimates of $194 million.
Cloud and services revenue for the quarter was $126 million, showing growth of 281% year-over-year.
Additionally, hardware sales actually declined 23% year-over-year to $51.1 million.
Co-founder and CEO, Andrew Feldman, said the demand for fast inference is enormous and Cerebras is calling to meet it. Feldman said the company is looking to triple its revenue in 2027 Let's consider some of the more positive highlights for Cerebras. The company raised $6.4 billion in gross proceeds through its IPO and closed a revolving credit facility for up to $850 million to accelerate the pace of its data center acquisitions. It also has $25.4 billion remaining in performance obligations as of June 30th, 2026
Cerebras also has separate partnerships with both AMD and Amazon's AWS to pioneer faster forms of AI inference. The partnership with AMD, for example, aims to deliver Cerebras' high speeds while increasing throughput by up to five times and is scheduled to go into production in the fourth quarter of 2026
Cerebras expects core revenue for the third quarter of $214 to $216 million while raising its full year outlook to $880 to $890 million from the prior guidance of $855 to $865 million.
So it's now time to discuss potential concerns for investors about Cerebras. Execution risk is a primary concern for this name. Like many other semiconductor companies, Cerebras does not actually manufacture its own chips. Rather, it designs the chips and outsources the manufacturing to other companies like Taiwan Semiconductor. This represents a general bottleneck in that the manufacturers only have so much production capacity to go around. However, Cerebras' CEO recently said as well that the primary constraint on AI growth is data centers referring to both their physical scarcity and slow deployment as well as the power infrastructure needed to scale AI systems. Another area of concern is customer concentration. Cerebras has a multi-year agreement with OpenAI valued at over $20 billion, which represents roughly 80% of their performance obligations. This issue is even more severe given Cerebras' unusual geopolitical concentration. Much of their revenue is derived from entities affiliated with the United Arab Emirates, which presents potential risks in the event of foreign policy changes or tariff and export news. To consider the technical picture for Cerebras, we would first begin by looking at a daily chart of the total span of Cerebras' trading activity. The company began trading here on May 15th, so we don't have too terribly much to go on, but we can see a few key things that stand out. First, we can see that the company bottomed out near 160.81. From there, we formed a kind of a triple type of bottom activity there, where we hit that level and rebounded several times. For the most part, activity has been range bound between this floor near 160 and a ceiling near about 243 or so. Price also has begun to trend upward recently heading into earnings. We had a white trend line in play there, but after a pop before the earnings report was delivered, we could see that that fizzled out and now price remains below that short term white trend line. We can see that our high point was roughly 266, which also stands out as a high point from back in late May shortly after shares began trading. When we think about RSI, our measure of momentum, we can see that we don't have a ton of data to go on once again. The RSI needs at least 14 trading days to consider the pace of the gains or losses in the case of a declining stock for traders to consider for this momentum indicator. We can see that our green upward sloping trend line has been broken once again, much like our price trend line here. However, RSI remains above the 50 midline, so we still have a slightly more bullish outlook on this momentum perspective here. In terms of our moving averages, oftentimes I use exponential moving averages, but because we have so little data, in this case, we have a 20-day simple moving average in yellow and a 50-day simple moving average in blue. Those come together almost right on top of each other between about 208 to 210 or so. So this gives traders another potentially supportive confluence to consider going forward. Volume profile shows that we have a few nodes to consider. 181 to 191 is a smaller node that comes in near the extreme lows that we have seen so far. But the bulk of the trading activity since Cerebras has began trading was between 200 to 233 with the point of control, the thick red line there that connotates the heaviest trading area, coming in at 221.50. Things start to thin out above 233 and then we pick up again between about 275 to 307 or so. Beyond that, the activity is relatively light because there wasn't a whole lot of time spent in that area for people to trade in. Finally, the expected move situation for this company shows some interesting information. First, we can see our green box, September 18th, 36 days out. That's our monthly expiration for September. Shows that the market is not expecting things to move too terribly far beyond our low points that we established there. We bottom out in that area near about 170 So, that could be an interesting area to watch out for if things do start to decline further. That could represent a major supportive area or a major breakdown point. Meanwhile, to the upside, even going out as far as November, at this time, the market does not expect us to reach those new highs that we established near 386