Huang Renxun once said: The demand for AI inference in the future will increase by 1 billion times. And just as the era of inference truly began to erupt, a dark horse of OpenAI based AI chips was about to go public. Cerebras Systems（ CBRS ） The latest issuance price range has reached $150-160, and the final pricing is expected to be set after the market closes on May 13th Eastern Time. It will be officially listed on May 14th. What's even more interesting is: It is actually the first Pre IPO project of @ MSX_CN. ： https://(msx.com)/? code=jUTs61 The first batch of MSX participating users had a buying price of only $100.35 at that time. That is to say, when many people are just starting to study Cerebras, the first batch of people have already been ambushing for several months in advance. I think this incident can illustrate a problem: The market will always be: Doubt first, then acknowledge, and finally chase higher. At the beginning of March this year, many people still thought that 'Pre IPO is a fake demand'. As a result, more and more people in the market are starting to study it MSX is one of the platforms that laid out this direction relatively early, and the first phase was directly opened at that time: SpaceX, ByteDance, Lambda Labs, and Cerebras are four unlisted targets. Essentially, Pre IPO is not something new. Traditional finance has always existed. Previously, these opportunities mostly belonged to institutional, high net worth, and VC circles. And what Crypto does is actually to lower the threshold and give ordinary users the first opportunity to access these primary market opportunities. Let's talk about Cerebras itself. Why has it suddenly become so popular recently? Because it hits not just AI. But rather the hottest direction in AI: inference. Now the entire industry has discovered: Training models is certainly important, but what really consumes computing power on a large scale is actually inference. including: Long text generation AI Agent、 Real time interaction Code generation Search reasoning These are all reasoning requirements. And Cerebras' route is completely different from NVIDIA's. The core of NVIDIA is GPU cluster+CUDA ecosystem. Cerebras is another route: Directly turn the entire wafer into a super large chip. Its core product WSE-3: The area is 57 times that of H100 52 times the number of cores 880 times on-chip memory • Memory bandwidth 7000 times It looks exaggerated. But what really matters is not 'big'. But rather: it has lower latency, higher throughput, and lower energy consumption in inference scenarios. Especially suitable for future needs such as AI agents, long texts, and real-time interaction. Official and some estimates even show that in some inference tasks, Cerebras' CS-3 system is much faster than NVIDIA DGX B200, while having lower cost and power consumption. That's also why OpenAI has recently started to focus on supporting Cerebras. Many people have not noticed that OpenAI is currently restructuring its computing power supply chain. Previously: Training, reasoning, networking, cloud, Almost all of them are covered by NVIDIA alone. But now OpenAI is starting to split: Who is used for training Who is used for reasoning How to do network protocol How to allocate cloud services Let's start redefining everything. And Cerebras, Essentially, it is an important layout for OpenAI's inference side. Even OpenAI core executives, including Sam Altman and others, have investment relationships in Cerebras. besides: Cerebras has already taken it Collaboration with OpenAI, Amazon Web Services, and others. AWS has even officially connected Cerebras' CS-3 system to the cloud. Besides, financial data is actually quite impressive. The revenue for 2025 is approximately 510 million US dollars, A year-on-year increase of about 76%, And it has already achieved profitability. Placed in the current AI inference arena, The market is willing to give it a high valuation, It's actually not difficult to understand. Of course Risks must also be considered. for example At present, the concentration of customers is still relatively high, Moreover, the global risk market itself is currently unstable. Last night's CPI data was average, and the situation in the Middle East has not fully settled yet. So even if the market is hot, the final stock price performance will still be affected by the overall market sentiment. But from an industry perspective, Cerebras' IPO may not just be an AI chip company going public. But rather: OpenAI is starting to price "inference computing power" separately. When the inference market begins to take shape independently, The AI computing power market will gradually shift from being dominated by NVIDIA to having different chip solutions for different scenarios. That's also why the market has suddenly started to refocus on Cerebras recently.