The Wire
Dispatch
Latest Daily
AMD Crossed $700 Billion Because AI Agents Need CPUs Too
AMD beat on earnings and revenue, jumped 18.6% in a day, and crossed $700B market cap. Lisa Su says agentic AI is driving "tremendous demand" for CPUs. The AI trade just expanded beyond GPUs. Micron also hit $700B. The semiconductor cycle isnt slowing down.
What Happened
AMD reported Q1 2026 earnings on Tuesday that beat on both lines. Earnings per share came in at $1.37 versus the $1.29 consensus. Revenue hit $10.25 billion against $9.9 billion expected, up 38% year over year. The data center business was the primary driver.
The stock surged 18.6% on Wednesday, pushing AMD's market cap above $713 billion. Shares are up 20% on the week and roughly 90% over the past month.
CEO Lisa Su told CNBC the demand picture became clearer over the last 90 days. The catalyst isnt what most people expected. "Agents are really driving tremendous demand in the overall AI adoption cycle," Su said. "We're seeing a shifting of the workload."
The move lifted the entire chip sector. The VanEck Semiconductor ETF jumped 5%. Intel climbed 4.5%. Micron crossed $700 billion in market cap on the same day, capping a 700% rally over the past year. The S&P 500's technology sector was the best performer of the day, up over 2%.
The Real Story
The AI trade just changed and most investors havent noticed
For the past three years, the AI investment thesis was simple: buy Nvidia. GPUs train models. Nvidia makes the best GPUs. Everything else is secondary.
Sus earnings call rewrote that thesis. Agentic AI doesnt just train models. It runs them continuously. Agents perform tasks, make decisions, call APIs, manage workflows, and operate around the clock. That runtime is CPU-heavy, not GPU-heavy. Training needs massive parallel processing. Inference and orchestration at scale need fast, efficient CPUs and massive memory bandwidth.
This is why AMD and Micron both crossed $700 billion on the same day. The market is pricing in a shift from "AI training" to "AI deployment." Training was Nvidias monopoly. Deployment spreads the demand across CPUs, memory, networking, and storage. The beneficiary list just got a lot longer.
90% in a month is not normal, even for semiconductors
AMD was trading around $375 a month ago. Its now above $700. Micron has rallied 700% in twelve months. SK Hynix gained 60% in April alone. These are not small-cap speculative names. These are some of the largest companies in the world moving at meme stock speed.
The earnings justify a lot of it. AMD's 38% revenue growth is real. The data center ramp is real. The agentic AI demand shift Su described is visible in hyperscaler capex plans. Microsoft, Google, Amazon, and Meta are all spending record amounts on AI infrastructure in 2026.
But 90% in a month prices in several quarters of perfect execution with no slowdown. The moment revenue growth decelerates from 38% to 30%, or even to 25%, the stock reprices. At $713 billion, AMD trades at roughly 70x trailing earnings. That multiple only works if the growth rate holds. It doesnt work if the growth rate merely stays good instead of exceptional.
Nvidia reports May 20 and that date now matters more than ever
AMD just told the market that agentic AI is the new demand driver. Nvidia's earnings will either confirm or deny that thesis at scale. If Jensen Huang echoes Su's framing and shows inference revenue accelerating, the entire semiconductor sector gets another leg up. If Nvidia's guidance disappoints or focuses only on training demand, the market questions whether the "agentic" narrative is AMD-specific rather than industry-wide.
The other variable is China. Huang recently said Nvidia's China revenue has fallen to zero due to export restrictions. AMD faces similar constraints. The AI chip boom is currently a US and allied-country story. Any change in export policy, in either direction, reshapes the revenue outlook for every name in the sector.
Market Impact
Bull case
Agentic AI creates demand across CPUs, memory, and networking beyond GPUs. AI capex is accelerating with hyperscalers spending record amounts in 2026. AMD, Micron, Broadcom, Marvell all benefit.
If AMD sustains 38% growth, a $1T market cap comes into view. If Su's "workload shift" is right, the CPU demand cycle is just beginning.
SMH up 5% in a single day. The entire sector moving together means this is an industry cycle, not a single-stock event.
Bear case
AMD up 90% in a month and Micron up 700% in a year are historically unsustainable rates. Even good earnings can become sell events when expectations are too high.
70x trailing earnings assumes 40%+ growth persists indefinitely. Semiconductors are cyclical. Cycles end.
Zero China revenue creates a growth ceiling. Getting the same multiple without access to the worlds largest chip market is a contradiction.
Already priced in?
Q1 earnings are priced. What isnt priced is how big agentic AI demand actually is and how long it lasts. Nvidia on May 20 answers that.
What's Next
May 20 is the date. Nvidia's earnings will either validate the agentic AI thesis across the entire sector or expose it as AMD-specific optimism. If Huang reports accelerating inference revenue and mentions agentic workloads, every chip stock gets rerated higher. If the guidance is cautious, the market asks whether AMD's 90% monthly gain was the cycle peak.
Watch hyperscaler earnings for capex signals. Microsoft reports May 12, Google May 14. Both will disclose AI infrastructure spending for Q2 and beyond. If the capex numbers keep rising, the semiconductor demand case strengthens. If any hyperscaler signals a pause or slowdown, the sector corrects fast because the valuations leave zero room for deceleration.
The Korea connection matters. Samsung and SK Hynix are roughly 44% of the KOSPI. AMD and Micron crossing $700 billion on the same day validates the memory and AI chip cycle that is driving the Korean market. If this cycle holds, KOSPI stays near records. If it breaks, the most concentrated major market in the world breaks with it.
For anyone buying semiconductor stocks today, the question isnt whether AI demand is real. It is. The question is whether 70x earnings and 90% monthly gains leave any margin of safety. The answer is no. That doesnt mean the stocks go down. It means if they do, they go down fast.
Latest Weekly
Market analysis and trade ideas. Have a ticker in mind? Send us your request.
