AMD Q2 Revenue Surges 50% Beating Expectations, but Q3 Guidance Disappoints Some Investors
Claire Weston
AMD posted Q2 revenue of $11.54 billion, up 50% year-over-year and above estimates, with data-center sales doubling; yet its Q3 guidance of ~$13 billion fell short of the most bullish forecasts, sending shares down ~7% after hours — the market is repricing whether growth is fast enough.
How good was this quarter, really?
Revenue hit $11.54 billion, beating the Street's $11.28 billion consensus. Adjusted EPS came in at $1.66 versus the expected $1.62.
Data-center revenue reached $6.72 billion — more than double year-over-year and above the $6.48 billion estimate. This means → AI chip demand is still accelerating, and AMD is no longer just "the Nvidia alternative."
PC and gaming revenue rose 6% to $3.8 billion — modest growth, but it held the base steady.
It beat estimates — so why did the stock drop?
AMD guided Q3 revenue to ~$13 billion (± $300 million), above the Street average of $12.52 billion.
But per Bloomberg, some firms had forecasts well above $13 billion. This means → the most bullish money on the Street was betting on a bigger number.
AMD stock has already more than doubled this year, far outpacing the broader market. In plain terms = the bigger the run-up, the higher the bar for the next quarter — "good" isn't enough; it has to be "great" to justify the valuation. Shares fell ~7% after hours to $518.58.
What drove the data-center doubling?
The core driver is a global wave of hyperscale data-center build-outs. Per Reuters, AMD has shifted from selling standalone chips to offering integrated AI system packages — GPU, CPU, and networking bundled together — competing directly with Nvidia's rack-scale products.
In CPUs, AMD keeps taking share from Intel. As GPUs dominate AI training, servers need more CPUs to run alongside expensive GPUs, giving AMD an additional revenue stream.
On the supply side, a bottleneck: AMD relies heavily on TSMC's advanced packaging — a process that tightly assembles multiple chips together — and that capacity remains tight. This reflects a key dynamic: demand isn't the constraint; capacity is the ceiling.
Can the new product cycle sustain the second half?
CEO Lisa Su said in July that the second-generation Helios AI server — featuring the MI455X AI accelerator and "Venice" processor — has entered full production and will begin shipping in the coming months.
Previously disclosed customers include Meta and OpenAI, with procurement agreements with Microsoft, Oracle, and Anthropic. This means → the orders are signed; what matters now is the actual shipping pace.
Per Bloomberg, Bernstein analysts estimate that the world's largest data-center operators will spend over $1 trillion in capex next year. In plain terms = the downstream money is already committed; AMD's job is to match capacity to demand and ship — whether Q3 guidance converts into real deliveries is the market's core checkpoint for re-rating AMD.
Content is for reference only, not financial advice.