AMD’s 16% Rout: When Gigawatt Deals Meet Reality

AMD plunged 16% after Broadcom's AI miss sparked sector rout, but $100B OpenAI and Meta deals underscore long-term AI growth.

The slide happened fast. Seven days after touching a record €471, AMD’s stock had shed 16%, landing at €393.85. The decline was brutal, but the trigger had nothing to do with the company’s own performance. It came from across the chip aisle: Broadcom.

Broadcom’s forecast for third-quarter AI chip revenue fell short of analyst expectations — $16 billion versus the $17.2 billion consensus — and the company declined to raise its full-year outlook. That was enough to ignite a sector-wide rout. On June 5, 2026, US chipmakers collectively lost $1.3 trillion in market value. Nvidia, Micron and AMD all plunged. The Philadelphia Semiconductor Index tumbled 10%, a drop not seen since March 2020.

The sell-off is a sharp reminder that even the most formidable growth stories are priced for perfection. AMD’s annualized 30-day volatility now stands at 85%, and while the relative strength index has cooled to 51–54 — signalling a more neutral stance — the stock still trades well above its 50-day moving average. The 106–112% year-to-date gain leaves little room for disappointment.

Yet behind the noise of the past week lies a fundamental transformation that few companies can match. AMD’s first-quarter 2026 revenue hit $10.3 billion, up 38% year-over-year. Data centre revenue surged 57% to $5.8 billion, accounting for the majority of turnover for the first time in the company’s history. The era of AMD as solely a PC chip maker is over.

What has driven that shift is a pair of contracts that redefine the scale of AI infrastructure. AMD has signed agreements with OpenAI and Meta, each covering six gigawatts of AI computing capacity, financed through multiple generations of Instinct GPUs. The first tranche — one gigawatt per partner — is slated for delivery from the second half of 2026 using the new MI450. Combined with an earlier OpenAI deal from October 2025, AMD now has twelve gigawatts of binding GPU deployments from two of the most important AI companies in the world.

The OpenAI contract alone could generate roughly $100 billion in revenue over its lifespan, structured as a “compute-for-upside” model in which AMD grants warrants for up to 160 million shares that only vest if specific deployment and stock-price targets are met.

Those commitments rest on the MI400 platform, whose flagship MI450 chip and Helios rack system (the MI455X) are expected to begin production in the second half of the year. CEO Lisa Su has said the Helios rack will deliver up to three AI exaflops in a single unit by the third quarter. Analysts forecast that the MI400 series could generate around $7.2 billion in revenue in its debut year, representing roughly 25% of expected data centre revenue. For the full year, data centre revenue is projected to rise 73% to $28.7 billion.

A less heralded driver is the rise of agentic AI. CFO Jean Hu recently described it as “the biggest change in recent months,” generating “very significant and incremental demand” for AMD’s server CPUs. Server CPU revenue grew more than 50% in Q1, and management expects over 70% growth in Q2. AMD has doubled its addressable server CPU market estimate from $60 billion to more than $120 billion by 2030.

Still, the ramp is not without friction. AMD has warned that MI450 production will depress gross margins in the second half — the new GPUs come in below the corporate average. Supply constraints remain acute: advanced manufacturing nodes and packaging capacity are tight, and the company has said availability will stay constrained through the end of 2026. Meeting the gigawatt delivery promises depends heavily on partners such as TSMC adding capacity.

Investors are watching closely. The consensus price target sits at €417–418, offering only modest upside from current levels. After a 302% rally from the 52-week low of €100.58, the market is essentially saying the thesis holds but the near-term valuation is fair. The next major checkpoint comes on July 23, 2026, when AMD holds its “Advancing AI” event at San Francisco’s Moscone Center — the moment when MI450 production is supposed to begin, and when execution must meet enthusiasm.