Nvidia's latest annual revenue reached $216 billion, while Huawei's comparable period revenue was $126 billion. However, Nvidia's AI chip sales in China have stalled as the company faces significant challenges from domestic competitors. As reported by Counterpoint Research, Nvidia engineered its H20 chips by stripping down their computing power to comply with U.S. export restrictions, but up to last year, it was still selling H20 chips in China, though shipments were gradually declining. At Nvidia's recent shareholders meeting, CEO Jensen Huang acknowledged that the U.S. has lost its edge in China's advanced AI chips market as Chinese competitors have become 'giants.'
OpenAI has undergone a dramatic transformation from model lab to vertically integrated empire, securing 40% of the global raw undiced DRAM wafer output until 2029 and launching its first custom chip with Broadcom. As reported by The Information, OpenAI engineers have cut inference costs by more than half through new optimization methods, though the company has not published technical details. The company's Jalapeño chip, described as 'purpose-built for LLM workloads powering ChatGPT, Codex API, future agentic products', represents what analysts call 'the first chip designed with AI agents in the loop' - where the model helps design the chip that runs it. This strategic pivot reflects Sam Altman's vision to control the entire AI stack, from silicon design to memory supply and even human biology through the $500 million Intercept initiative to eliminate respiratory viruses.
Huawei has been expanding its global chip operations, operating in 170 countries and regions with a mission of 'bringing digital to every person, home and organization for a fully connected, intelligent world.' The company's semiconductor business head He Tingbo stated that 'We have found pretty good solutions' when asked about competing with global rivals. Huawei has been rolling out some of the world's most powerful AI computing clusters, combining the power of thousands of chips despite relying on Chinese-made semiconductors due to U.S. export controls. DeepSeek, the fast-growing Chinese rival of OpenAI's ChatGPT, adapted its latest V4 AI model rolled out in April for Huawei's advanced Ascend chips, with analysts noting 'significant effort going into collaboration between DeepSeek and Huawei' to train future models on domestic hardware.
Broadcom recently unveiled its Jalapeño chip, designed specifically for OpenAI's large language model workloads and agentic AI applications. According to Broadcom, while the chip is still in testing, Jalapeño's per-watt performance is 'substantially better than current state-of-the-art'. This breakthrough addresses the growing importance of inference-focused AI workloads as tech companies develop cutting-edge models capable of multi-step processes. The chip's performance advantage is particularly significant for OpenAI, which owns the popular ChatGPT chatbot, but also for other companies investing heavily in AI infrastructure seeking greater efficiency. This development could unlock significant revenue opportunities for Broadcom and potentially help the company develop similar chips for other chatbot companies.
Despite these developments, Nvidia stock rose nearly 2% on June 30, reaching a market value near $4.8 trillion. As reported by The Information, Nvidia's latest results showed data-center revenue increased 75% to a record $62.3 billion in a single quarter. However, investors have now rotated funds into AI infrastructure stocks perceived to have greater upside potential, with Nvidia benefiting from this broader market shift toward AI-focused investments. Nvidia expects around $91 billion of revenue in May-July, up from nearly $82 billion in the previous quarter, excluding any data center compute revenue from China. Most pressure from the inference shift is concentrated at the inference layer, where Nvidia still maintains dominance through its CUDA software that has locked in developers since 2006.
OpenAI's 40% DRAM wafer lock-up through 2029 creates a direct impact on Micron Technology, whose Q3 FY2026 revenue more than quadrupled year-over-year to $41.46 billion with gross margins of 84.6%. As reported by The Information, Micron's CEO Sanjay Mehrotra noted that record fiscal Q3 results reflect the strategic value of memory in the AI era. Wall Street targets from TD Cowen, Cantor Fitzgerald, and Susquehanna run up to $1,750, with analysts arguing AI has created a 'permanent, higher demand baseline, fundamentally changing the historically cyclical memory business.' On Reddit, sentiment swung from pre-earnings bearishness at 38 to post-earnings bullishness at 81, with the stock up roughly 800.86% over the past year.
