The AI plumbing trade took the wheel over the last eight hours after AMD raised the stakes at its analyst day, projecting data center revenue to jump 60% over the next three to five years from a $16 billion base in 2025. Wall Street never met a trillion-dollar TAM it didn’t fall in love with, and AMD just hung a $1 trillion sign over AI data centers. Volume chased chips and servers, while hand-wringing over power and financing simmered in the background.
What drove attention today: AMD said it expects data center revenue to climb about 60% over three to five years as MI450 accelerators, Helios rack-scale systems, and future MI500 parts ramp. The company pegs the AI data center total addressable market at $1 trillion across GPUs, CPUs, and networking. It touted a 6-gigawatt OpenAI build and 50,000 chips for Oracle starting in 2026. Before the bell, AMD stock jumped roughly 5% as this script hit traders’ screens.
Trading profile: High-beta AI proxy with options volume blasting through normals. Gross margin guide of 55% to 58% and operating margins above 35% frame a profitable path if execution holds. Data center CPUs also matter here, with an ambition to take server revenue share toward 50% from roughly 40% now. It is a growth story priced like one.
Key takeaway for investors: AMD isn’t just chasing Nvidia anymore; it is selling a roadmap to share gains in both accelerators and server CPUs. The two wildcards are power availability for gigawatt-scale projects and the financing cadence for AI-native customers. If the GPU pipeline arrives on time, the numbers have room to work.
What drove attention today: Sympathy and friction. While AMD grabbed the megaphone, Nvidia drew the money. In heavy action, about 158.6 million shares traded with the stock slipping roughly 2.9% to near 193, a classic profit-taking move after a big run and a reminder that even market darlings need to breathe. No new product bombshells today, just recalibration around competitive intensity.
Trading profile: The liquidity behemoth in AI with deep options markets and relentless dip-buying tendencies. CUDA software remains a moat, supply is scaling, and hyperscaler budgets are still pointed at Nvidia clusters. When the group wobbles, NVDA typically gets sold first and bought first.
Key takeaway for investors: AMD’s ambition sharpens the competitive lens, but Nvidia still owns the stack and the developer mindshare. Short-term red on big volume is noise in a structurally strong tape. Watch gross margin durability and allocation dynamics as rivals push MI-series accelerators into more racks.
What drove attention today: AMD’s server share target is a direct shot across Intel’s data center bow. AMD said it sees a path to capture as much as 50% of server revenue share, up from about 40% today, while also calling for double-digit client growth over five years. That implies ongoing pressure on Intel’s high-margin franchise as the foundry comeback grinds on.
Trading profile: A turnaround that trades on milestones: process nodes, foundry wins, and signs of data center stabilization. INTC often fades when the discourse tilts to CPU share losses or when an AI GPU arms race diverts spending away from legacy CPU-heavy builds. Volatility is lower than the GPU pure-plays but spikes around roadmap events.
Key takeaway for investors: The investment case hinges on delivering next-gen nodes on time and recapturing server credibility with competitive performance-per-watt. If AMD inches toward 50% server share, Intel must prove Granite Rapids-class parts can change the narrative and that foundry economics won’t kneecap margins along the way.
What drove attention today: AMD flagged a plan to supply Oracle with 50,000 chips starting in 2026, a notable tranche for a cloud provider leaning hard into AI infrastructure. Oracle already has deep AI-wins with select customers, and fresh visibility into GPU procurement keeps the story in the AI data center conversation without relying on a single vendor.
Trading profile: Lower beta than chipmakers, with the stock reacting to cloud backlog growth, capex cadence, and headline flow around GPU access. Options are active but not manic. The narrative revolves around AI instance availability and the speed of onboarding large AI-native tenants.
Key takeaway for investors: Oracle’s AI lane is simple: secure supply, stand up clusters, monetize capacity. If 2026 hardware arrives on schedule, revenue mix can skew more toward high-performance cloud. The risk is timing slippage in chip deliveries or power constraints that push out the ramp.
What drove attention today: AMD’s Helios rack-scale shout-out puts the spotlight on server integrators that can turn GPU pallets into deployable capacity at speed. SMCI lives at that intersection. When traders hear gigawatt-scale and multi-vendor accelerator roadmaps, they reach for the purest rack-scale beta on the board.
Trading profile: High-volatility, high-liquidity, and frequently gapped. SMCI’s tape is hypersensitive to GPU allocation chatter from Nvidia and AMD, and to hyperscaler capex signals. It often overshoots both ways and attracts momentum funds and retail alike.
Key takeaway for investors: As long as accelerators ship and power comes online, SMCI can outrun consensus on volume and mix. Miss a quarter on supply, qualification, or logistics, and the air pocket is brutal. Position sizing is a risk management exercise, not a slogan.
AI data centers were the market’s most active arena because the thesis got a fresh catalyst: AMD’s multi-year guide, reinforced by real customers and credible margin targets. Nvidia remains the liquidity sun, Intel is fighting share math, Oracle needs timely chips and power, and Super Micro amplifies every twist in the supply chain. Outside the AI plumbing, names like BigBear.ai and Opendoor soaked up their own volume, but the loudest cash register today rang in silicon and servers.
