A New Phase Of The AI Trade
Chipmaker stocks have surged over the past month as investors reassess which companies may benefit from the next stage of artificial intelligence infrastructure. Micron has risen 80%, SanDisk has gained 52% and Intel has advanced 85%, highlighting a rally that is expanding beyond the companies most closely tied to graphics processors.
The driver is a changing view of AI system architecture. Wall Street is increasingly focused on orchestration, a model in which workloads are distributed across multiple processing channels instead of being concentrated in larger, centralized computing blocks.
CPUs Gain Importance Alongside GPUs
During the first phase of the AI buildout, graphics processing units dominated investor attention because they powered model training and inference workloads. Nvidia became the main beneficiary of that cycle, as GPUs became essential to the rapid expansion of AI computing.
Orchestration does not eliminate the need for GPUs. Instead, it increases the importance of central processing units, networking and memory because more tasks must be coordinated across the system. That shift broadens the investment case across the semiconductor supply chain.
Agentic AI Changes Infrastructure Demand
Morgan Stanley analysts said agentic AI could increase the CPU-to-GPU mix in AI systems by adding more orchestration, memory and tool-use work. Agentic AI refers to software that can handle more generalized instructions and coordinate multiple steps or tools.
The analysts noted that this should not reduce GPU demand, but it does increase overall system complexity. As a result, incremental infrastructure spending may move toward CPUs, networking and memory, creating new winners beyond the most visible AI chip leaders.
Meta And AMD Highlight The Same Trend
Major technology companies are also emphasizing orchestration. Meta said in April that no single chip architecture can efficiently serve every workload and that its work with agentic AI is increasing demand for CPU capacity.
AMD made a similar point when announcing a deal with Meta in February. The company said CPUs are becoming a strategic pillar of the AI compute stack, supporting efficiency, scalability and orchestration alongside GPUs.
Cybersecurity Shows The Model In Practice
Orchestration is already appearing in cybersecurity research, where coordinated models can reproduce some advanced capabilities at lower cost. After Anthropic’s Mythos raised concerns in the security community, several research groups said they were able to replicate similar results using less advanced public models arranged through structured workflows.
Vidoc Security Lab said it used GPT-5.4 and Claude Opus 4.6 with a standardized security-review process to reproduce public examples outside Anthropic’s internal stack. Cybersecurity firm Aisle said smaller and cheaper coordinated models could also identify similar software bugs.
Memory And Infrastructure Suppliers Benefit
The orchestration theme is also supporting companies across the broader data center supply chain. Beneficiaries may include electronic design automation, baseboard management control, substrates, DRAM and NAND memory systems.
Morgan Stanley highlighted companies such as KLA, Cadence Design Systems and Taiwan-based Gold Circuit Electronics, along with major memory suppliers including Samsung, SK Hynix, Micron, SanDisk and Kioxia. For investors, the key message is that the AI trade may be entering a broader phase where system coordination, memory depth and infrastructure complexity matter as much as raw GPU capacity.
