SAN JOSE, California, April 27, 2026, 14:04 PDT
- Rambus shares tumbled about 10.8% Monday, after the company released its first-quarter results.
- Product revenue jumped 15% from a year earlier, driven by heightened demand for data-center and AI memory systems.
- The company signaled stronger product revenue for the second quarter, but warned investors that it’s all contingent on securing customer agreements.
Shares of Rambus Inc. plunged 10.79% on Monday, closing at $141.31. Earlier in the session, the stock had hit a high of $161.70 before retreating sharply. The chipmaker reported first-quarter revenue of $180.2 million and pointed to robust data-center demand driven by AI, but that wasn’t enough to calm investors. Growth in the company’s product business fell short of reversing the selloff.
This shift stands out—Rambus has become something of a stand-in for the memory bottleneck that’s plaguing AI systems. Unlike companies building core AI processors, Rambus focuses on memory interface chips and silicon IP, which are licensable chip designs crucial for moving data between memory and processors.
Rambus posted a 15% jump in product revenue, reaching $88.0 million. Cash from operations totaled $83.2 million. GAAP revenue increased as well, up from $166.7 million a year earlier. Net income was $59.9 million, or 55 cents per diluted share.
Luc Seraphin, the company’s chief executive, called it a “solid first quarter.” He flagged AI inference and agentic workloads as drivers lifting demand for higher memory bandwidth and accelerated data movement. The takeaway for Rambus: big AI models demand faster memory access, since processors by themselves can’t keep pace. Business Wire
Rambus set its sights on second-quarter product revenue of $95 million to $101 million, with licensing billings projected between $76 million and $82 million. Royalty revenue guidance landed in the $72 million to $78 million range. Licensing billings, which reflect invoices tied to Rambus technology, serve as a demand gauge, but not all amounts billed are booked as revenue right away.
Just days ahead of the latest report, Rambus introduced its SOCAMM2 server module chipset, targeting AI server workloads with a low-power memory solution. According to the company, these chipsets support LPDDR5X-based SOCAMM2 modules and push speeds up to 9.6 gigabits per second. The key difference: memory modules here are replaceable and serviceable, instead of being soldered onto the board.
Rambus edges a bit closer to the ranks of major memory vendors, but it’s not directly competing with them just yet. “The industry needs a robust ecosystem around LPDDR-class server memory,” said Praveen Vaidyanathan, Micron’s vice president and general manager of Cloud Memory Products. Over at IDC, Soo Kyoum Kim called SOCAMM2 “an important evolution” for balancing performance and efficiency in AI data centers. Nasdaq
Rambus is touting its HBM4E memory controller IP, a design set up specifically for high-bandwidth memory in AI accelerators and high-performance computing. The company claims the controller supports data speeds hitting 16 gigabits per second per pin and can push up to 4.1 terabytes each second per memory device. HBM puts memory close to processors, targeting quicker data transfer.
No letup in the fight. Micron, Samsung, and plenty of AI-chip names are all chasing solutions to bandwidth and power issues—no one’s cracked it yet. “HBM bandwidth,” as MatX boss Reiner Pope put it, remains a stubborn roadblock for large language models. Samsung’s Ben Rhew says the sector still needs HBM4E IP to really get moving on adoption. Rambus
Rambus hovered near its 52-week high before earnings, trading above the majority of analyst targets. Investing.com data showed eight of nine analysts rated the stock a buy. Put simply, expectations were lofty—so even solid product revenue didn’t move the needle much this quarter.
One catch: Rambus flagged that its Q2 revenue projection isn’t set in stone—it depends on sealing fresh customer agreements for both products and licenses. The company also pointed to possible downside: softer AI server demand, memory supply tightening in the wrong spots, or customer rollouts falling short could all weigh on results.
Rambus delivered growth in key AI infrastructure segments, but that wasn’t enough for the market. After a strong run-up, shares slipped as investors waited for more than just decent quarterly numbers.
