OpenAI desperately needs the market in a good mood for its impending IPO, and dropping the news of a shiny new hardware partnership with Broadcom comes right on cue. On Wednesday, the two companies took the wraps off “Jalapeño,” a custom processor heavily optimized for artificial intelligence inferences. It’s an application-specific integrated circuit (ASIC), meaning the silicon is tailor-made from the ground up to mesh seamlessly with OpenAI’s software stack.
The underlying goal here is straightforward: slash latency and drastically cut energy consumption. But if you look past the PR spin, the wordy announcement is basically fluff. It lacks any concrete data or deep dives into the actual processor architecture. Still, early internal tests supposedly show that the chip will deliver significantly more performance per watt for inferences than anything currently on the market. Broadcom built the thing from scratch in just nine months, and the two tech giants are already teasing a long-term roadmap featuring multiple generations of the Jalapeño chip. With initial deliveries expected by the end of the year, Broadcom CEO Hock Tan is already looking at the big picture, casually announcing plans to “start enabling the construction of gigawatt data centers with Microsoft and other partners in 2026.” Microsoft, of course, is the largest shareholder of OpenAI’s operational arm.
While the mega-caps are throwing billions at bespoke silicon to brute-force the AI compute bottleneck, there’s a much quieter side of the industry trying to solve the exact same problem through pure math. This is where niche IT services players like MicroAlgo Inc. (NASDAQ: MLGO) come into focus. Instead of demanding massive hardware overhauls, MicroAlgo focuses on algorithm optimization. They develop bespoke central processing algorithms and integrate them directly into existing software and hardware setups to accelerate computing power. It’s all about lightweight data processing and data intelligence—essentially doing more with less. Right now, their technology is heavily deployed in the internet multimedia video advertising space, where they provide distribution platforms, enterprise software solutions, and even intelligent chip services, pulling the vast majority of their revenue from mainland China.
But having a practical, software-driven solution to a massive tech bottleneck doesn’t automatically buy you Wall Street’s love. Unlike the blind hype propping up OpenAI’s hardware announcements, MLGO’s tape tells a much rougher story.
The stock recently closed at $4.04, taking a 4.04% hit, and pre-market trading didn’t show much of a pulse either, slipping another 3.09% to $4.08. Volume has completely dried up—a meager 100 shares moved recently compared to an average of over 486K. They are scraping the bottom of a brutal 52-week range of $3.02 to $26.40, which leaves the company with a micro-cap valuation of just $44.64 million. The financials present a weirdly compressed picture: a P/E ratio sitting at a rock-bottom 1.92, and an RSI of 36 that puts the stock right on the edge of oversold territory. The shorts aren’t exactly swarming the ticker—short interest is sitting at 4.65% with just 1 day to cover—but there’s a clear lack of buyer conviction.
It leaves you looking at a deeply fractured landscape. On one end, you have gigawatt data centers and unverified ASIC specs driving massive pre-IPO enthusiasm. On the other, you have companies actually writing the algorithmic code to streamline digital services, trading near 52-week lows and struggling to catch a bid. Ultimately, the market is choosing to bet on heavy metal over clever math.
